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 4.18.0-423.el8 (RHEL8.7)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 3.10.0-1062.18.1.el7 (RHEL7.7)
11 3.10.0-1127.19.1.el7 (RHEL7.8)
12 3.10.0-1160.76.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.18.0-348.23.1.el8 (RHEL8.5)
19 4.18.0-372.26.1.el8 (RHEL8.6)
20 4.4.120-92.70 (SLES12 SP2)
21 4.4.180-94.100 (SLES12 SP3)
22 4.4.0-131 (Ubuntu 16.04)
23 4.15.0-32 (Ubuntu 18.04)
24 5.4.0-48 (Ubuntu 20.04)
25 vanilla linux 5.4.0 (ZFS + ldiskfs)
26 vanilla linux 5.4.21 (ZFS + ldiskfs)
27 vanilla linux 5.4.136 (ZFS + ldiskfs)
28 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
29 * Client primary kernels built and tested during release cycle:
30 5.14.0-70.26.1.el9 (RHEL9.0)
31 4.18.0-423.el8 (RHEL8.7)
32 5.4.0-37 (Ubuntu 20.04)
33 5.14.21-150400.24.21 (SLES15 SP4)
34 * Other clients known to build on these kernels at some point (others may also work):
35 3.10.0-862.14.4.el7 (RHEL7.5)
36 3.10.0-957.27.2.el7 (RHEL7.6)
37 4.14.0-49.13.1.el7a (RHEL7.5)
38 4.14.0-115.2.2.el7a (RHEL7.6)
39 3.10.0-1062.18.1.el7 (RHEL7.7)
40 3.10.0-1127.19.1.el7 (RHEL7.8)
41 3.10.0-1160.76.1.el7 (RHEL7.9)
42 4.18.0-80.11.2.el8 (RHEL8.0)
43 4.18.0-147.8.1.el8 (RHEL8.1)
44 4.18.0-193.28.1.el8 (RHEL8.2)
45 4.18.0-240.22.1.el8 (RHEL8.3)
46 4.18.0-305.25.1.el8 (RHEL8.4)
47 4.18.0-348.23.1.el8 (RHEL8.5)
48 4.18.0-372.26.1.el8 (RHEL8.6)
49 4.4.120-92.70 (SLES12 SP2)
50 4.4.180-94.100 (SLES12 SP3)
51 4.12.14-95.48 (SLES12 SP4)
52 4.12.14-122.133 (SLES12 SP5)
53 4.12.14-197.75 (SLES15 SP1)
54 5.3.18-24.96 (SLES15 SP2)
55 5.3.18-150300.59.93 (SLES15 SP3)
56 4.4.0-131 (Ubuntu 16.04)
57 4.15.0-48 (Ubuntu 18.04)
58 5.8.0-53 (Ubuntu 20.04.2 HWE)
59 5.11.0-31 (Ubuntu 20.04.3 HWE)
60 5.11.0 (vanilla kernel.org)
61 * Recommended e2fsprogs version: 1.46.5.wc1 or newer
62 * Recommended ZFS version: 2.1.5
63 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
64 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
65 stack overflow. For more information, please refer to bugzilla 17630.
66 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
67 the re-exporting nodes to be mounted with "32bitapi" mount option
69 --------------------------------------------------------------------------------
74 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
75 for currently supported client and server kernel versions.
76 * Server primary kernels built and tested during release cycle:
77 4.18.0-348.2.1.el8 (RHEL8.5)
78 * Other server kernels known to build and work at some point (others may also work):
79 3.10.0-862.14.4.el7 (RHEL7.5)
80 3.10.0-957.27.2.el7 (RHEL7.6)
81 3.10.0-1062.18.1.el7 (RHEL7.7)
82 3.10.0-1127.19.1.el7 (RHEL7.8)
83 3.10.0-1160.49.1.el7 (RHEL7.9)
84 4.18.0-80.11.2.el8 (RHEL8.0)
85 4.18.0-147.8.1.el8 (RHEL8.1)
86 4.18.0-193.28.1.el8 (RHEL8.2)
87 4.18.0-240.22.1.el8 (RHEL8.3)
88 4.18.0-305.25.1.el8 (RHEL8.4)
89 4.4.120-92.70 (SLES12 SP2)
90 4.4.180-94.100 (SLES12 SP3)
91 4.4.0-131 (Ubuntu 16.04)
92 4.15.0-32 (Ubuntu 18.04)
93 5.4.0-48 (Ubuntu 20.04)
94 vanilla linux 5.4.0 (ZFS + ldiskfs)
95 vanilla linux 5.4.21 (ZFS + ldiskfs)
96 vanilla linux 5.4.136 (ZFS + ldiskfs)
97 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
98 * Client primary kernels built and tested during release cycle:
99 4.18.0-348.2.1.el8 (RHEL8.5)
100 5.4.0-37 (Ubuntu 20.04)
101 5.3.18-59.27 (SLES15 SP3)
102 * Other clients known to build on these kernels at some point (others may also work):
103 3.10.0-862.14.4.el7 (RHEL7.5)
104 3.10.0-957.27.2.el7 (RHEL7.6)
105 4.14.0-49.13.1.el7a (RHEL7.5)
106 4.14.0-115.2.2.el7a (RHEL7.6)
107 3.10.0-1062.18.1.el7 (RHEL7.7)
108 3.10.0-1127.19.1.el7 (RHEL7.8)
109 3.10.0-1160.49.1.el7 (RHEL7.9)
110 4.18.0-80.11.2.el8 (RHEL8.0)
111 4.18.0-147.8.1.el8 (RHEL8.1)
112 4.18.0-193.28.1.el8 (RHEL8.2)
113 4.18.0-240.22.1.el8 (RHEL8.3)
114 4.18.0-305.25.1.el8 (RHEL8.4)
115 4.4.120-92.70 (SLES12 SP2)
116 4.4.180-94.100 (SLES12 SP3)
117 4.12.14-95.48 (SLES12 SP4)
118 4.12.14-122.91 (SLES12 SP5)
119 4.12.14-197.75 (SLES15 SP1)
120 5.3.18-24.96 (SLES15 SP2)
121 4.4.0-131 (Ubuntu 16.04)
122 4.15.0-48 (Ubuntu 18.04)
123 5.8.0-53 (Ubuntu 20.04.2 HWE)
124 5.11.0-31 (Ubuntu 20.04.3 HWE)
125 5.11.0 (vanilla kernel.org)
126 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
127 * Recommended ZFS version: 2.0.7
128 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
129 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
130 stack overflow. For more information, please refer to bugzilla 17630.
131 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
132 the re-exporting nodes to be mounted with "32bitapi" mount option
134 --------------------------------------------------------------------------------
138 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
139 for currently supported client and server kernel versions.
140 * Server primary kernels built and tested during release cycle:
141 3.10.0-1062.18.1.el7 (RHEL7.7)
142 3.10.0-1127.19.1.el7 (RHEL7.8)
143 3.10.0-1160.6.1.el7 (RHEL7.9)
144 * Other server kernels known to build and work at some point (others may also work):
145 3.10.0-862.14.4.el7 (RHEL7.5)
146 3.10.0-957.27.2.el7 (RHEL7.6)
147 4.18.0-80.11.2.el8 (RHEL8.0)
148 4.18.0-147.8.1.el8 (RHEL8.1)
149 4.18.0-193.28.1.el8 (RHEL8.2)
150 4.18.0-240.1.1.el8 (RHEL8.3)
151 4.4.120-92.70 (SLES12 SP2)
152 4.4.180-94.100 (SLES12 SP3)
153 4.4.0-131 (Ubuntu 16.04)
154 4.15.0-32 (Ubuntu 18.04)
155 5.4.0-48 (Ubuntu 20.04)
156 vanilla linux 5.4.0 (ZFS + ldiskfs)
157 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
158 * Client primary kernels built and tested during release cycle:
159 3.10.0-1062.18.1.el7 (RHEL7.7)
160 3.10.0-1127.19.1.el7 (RHEL7.8)
161 3.10.0-1160.6.1.el7 (RHEL7.9)
162 4.12.14-95.48 (SLES12 SP4)
163 4.12.14-122.20 (SLES12 SP5)
164 4.12.14-197.75 (SLES15 SP1)
165 5.3.18-24.24 (SLES15 SP2)
166 4.15.0-48 (Ubuntu 18.04)
167 * Other clients known to build on these kernels at some point (others may also work):
168 3.10.0-862.14.4.el7 (RHEL7.5)
169 3.10.0-957.27.2.el7 (RHEL7.6)
170 4.14.0-49.13.1.el7a (RHEL7.5)
171 4.14.0-115.2.2.el7a (RHEL7.6)
172 4.18.0-80.11.2.el8 (RHEL8.0)
173 4.18.0-147.8.1.el8 (RHEL8.1)
174 4.18.0-193.28.1.el8 (RHEL8.2)
175 4.18.0-240.1.1.el8 (RHEL8.3)
176 4.4.120-92.70 (SLES12 SP2)
177 4.4.180-94.100 (SLES12 SP3)
178 4.4.0-131 (Ubuntu 16.04)
179 5.4.0-37 (Ubuntu 20.04)
180 5.4.0 (vanilla kernel.org)
181 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
182 * Recommended ZFS version: 2.0.0
183 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
184 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
185 stack overflow. For more information, please refer to bugzilla 17630.
186 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
187 the re-exporting nodes to be mounted with "32bitapi" mount option
189 --------------------------------------------------------------------------------
193 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
194 for currently supported client and server kernel versions.
195 * Server primary kernels built and tested during release cycle:
196 3.10.0-1062.1.1.el7 (RHEL7.7)
197 * Other server kernels known to build and work at some point (others may also work):
198 3.10.0-862.14.4.el7 (RHEL7.5)
199 3.10.0-957.27.2.el7 (RHEL7.6)
200 4.18.0-80.7.1.el8 (RHEL8.0)
201 4.4.120-92.70 (SLES12 SP2)
202 4.4.180-94.100 (SLES12 SP3)
203 4.4.0-131 (Ubuntu 16.04)
204 4.15.0-32 (Ubuntu 18.04)
205 vanilla linux 4.6.7 (ZFS only)
206 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
207 * Client primary unpatched kernels built and tested during release cycle:
208 3.10.0-1062.1.1.el7 (RHEL7.7)
209 4.12.14-95.29 (SLES12 SP4)
210 4.15.0-48 (Ubuntu 18.04)
211 * Other clients known to build on unpatched kernels at some point (others may also work):
212 3.10.0-862.14.4.el7 (RHEL7.5)
213 3.10.0-957.27.2.el7 (RHEL7.6)
214 4.14.0-49.13.1.el7a (RHEL7.5)
215 4.14.0-115.2.2.el7a (RHEL7.6)
216 4.18.0-80.7.1.el8 (RHEL8.0)
217 4.4.120-92.70 (SLES12 SP2)
218 4.4.180-94.100 (SLES12 SP3)
219 4.4.0-131 (Ubuntu 16.04)
220 vanilla linux 4.15.0 (ZFS only)
221 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
222 * Recommended ZFS / SPL version: 0.7.13
223 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
224 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
225 stack overflow. For more information, please refer to bugzilla 17630.
226 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
227 the re-exporting nodes to be mounted with "32bitapi" mount option
228 * Note for Ubuntu 14.04.5:
229 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
230 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
232 --------------------------------------------------------------------------------
236 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
237 for currently supported client and server kernel versions.
238 * Server known to build on the following kernels (others may also work):
239 2.6.32-696.18.7.el6 (RHEL6.9)
240 3.10.0-862.14.4.el7 (RHEL7.5)
241 3.10.0-957.el7 (RHEL7.6)
242 4.4.120-92.70 (SLES12 SP2)
243 4.4.162-94.69 (SLES12 SP3)
244 4.4.0-131 (Ubuntu 16.04)
245 4.15.0-32 (Ubuntu 18.04)
246 vanilla linux 4.6.7 (ZFS only)
247 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
248 * Client known to build on unpatched kernels (others may also work):
249 2.6.32-696.18.7.el6 (RHEL6.9)
250 2.6.32-754.6.3.el6 (RHEL6.10)
251 3.10.0-862.14.4.el7 (RHEL7.5)
252 3.10.0-957.el7 (RHEL7.6)
253 4.14.0-49.13.1.el7a (RHEL7.5)
254 4.14.0-115.2.2.el7a (RHEL7.6)
255 4.4.120-92.70 (SLES12 SP2)
256 4.4.162-94.69 (SLES12 SP3)
257 4.4.0-131 (Ubuntu 16.04)
258 4.15.0-32 (Ubuntu 18.04)
259 vanilla linux 4.15.0 (ZFS only)
260 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
261 * Recommended ZFS / SPL version: 0.7.9
262 * Tested with ZFS / SPL version: 0.7.9
263 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
264 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
265 stack overflow. For more information, please refer to bugzilla 17630.
266 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
267 the re-exporting nodes to be mounted with "32bitapi" mount option
268 * Note for Ubuntu 14.04.5:
269 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
270 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
272 --------------------------------------------------------------------------------
274 2018-04-03 Intel Corporation
276 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
277 for currently supported client and server kernel versions.
278 * Server known to build on patched kernels:
279 2.6.32-431.29.2.el6 (RHEL6.5)
280 2.6.32-504.30.3.el6 (RHEL6.6)
281 2.6.32-573.26.1.el6 (RHEL6.7)
282 2.6.32-642.15.1.el6 (RHEL6.8)
283 2.6.32-696.18.7.el6 (RHEL6.9)
284 3.10.0-693.21.1.el7 (RHEL7.4)
285 3.0.101-0.47.71 (SLES11 SP3)
286 3.0.101-107 (SLES11 SP4)
287 3.12.74-60.64.40 (SLES12 SP1)
288 4.4.114-92.64 (SLES12 SP2)
289 4.4.114-94.11 (SLES12 SP3)
290 3.13.0-101 (Ubuntu 14.04, ZFS only)
291 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
292 4.4.0-116 (Ubuntu 16.04)
293 vanilla linux 4.6.7 (ZFS only)
294 * Client known to build on unpatched kernels:
295 2.6.32-431.29.2.el6 (RHEL6.5)
296 2.6.32-504.30.3.el6 (RHEL6.6)
297 2.6.32-573.26.1.el6 (RHEL6.7)
298 2.6.32-642.15.1.el6 (RHEL6.8)
299 2.6.32-696.18.7.el6 (RHEL6.9)
300 3.10.0-693.21.1.el7 (RHEL7.4)
301 3.0.101-0.47.71 (SLES11 SP3)
302 3.0.101-107 (SLES11 SP4)
303 3.12.74-60.64.40 (SLES12 SP1)
304 4.4.114-92.64 (SLES12 SP2)
305 4.4.114-94.11 (SLES12 SP3)
306 3.13.0-101 (Ubuntu 14.04)
307 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
308 4.4.0-116 (Ubuntu 16.04)
310 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
311 * Recommended ZFS / SPL version: 0.7.6
312 * Tested with ZFS / SPL version: 0.7.6
313 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
314 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
315 stack overflow. For more information, please refer to bugzilla 17630.
316 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
317 the re-exporting nodes to be mounted with "32bitapi" mount option
318 * Note for Ubuntu 14.04.5:
319 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
320 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
322 --------------------------------------------------------------------------------
324 2017-07-05 Intel Corporation
326 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
327 for currently supported client and server kernel versions.
328 * Server known to build on patched kernels:
329 2.6.32-431.29.2.el6 (RHEL6.5)
330 2.6.32-504.30.3.el6 (RHEL6.6)
331 2.6.32-573.26.1.el6 (RHEL6.7)
332 2.6.32-642.15.1.el6 (RHEL6.8)
333 2.6.32-696.3.1.el6 (RHEL6.9)
334 3.10.0-514.21.1.el7 (RHEL7.3)
335 3.0.101-0.47.71 (SLES11 SP3)
336 3.0.101-100 (SLES11 SP4)
337 3.12.74-60.64.40 (SLES12 SP1)
338 4.4.59-92.17 (SLES12 SP2)
339 vanilla linux 4.6.7 (ZFS only)
340 * Client known to build on unpatched kernels:
341 2.6.32-431.29.2.el6 (RHEL6.5)
342 2.6.32-504.30.3.el6 (RHEL6.6)
343 2.6.32-573.26.1.el6 (RHEL6.7)
344 2.6.32-642.15.1.el6 (RHEL6.8)
345 2.6.32-696.3.1.el6 (RHEL6.9)
346 3.10.0-514.21.1.el7 (RHEL7.3)
347 3.0.101-0.47.71 (SLES11 SP3)
348 3.0.101-100 (SLES11 SP4)
349 3.12.74-60.64.40 (SLES12 SP1)
350 4.4.59-92.17 (SLES12 SP2)
352 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
353 * Recommended ZFS / SPL version: 0.6.5.9
354 * Tested with ZFS / SPL version: 0.6.5.9
355 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
356 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
357 stack overflow. For more information, please refer to bugzilla 17630.
358 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
359 the re-exporting nodes to be mounted with "32bitapi" mount option
361 --------------------------------------------------------------------------------
363 12-07-2016 Intel Corporation
365 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
366 for currently supported client and server kernel versions.
367 * Server known to build on patched kernels:
368 2.6.32-431.29.2.el6 (RHEL6.5)
369 2.6.32-504.30.3.el6 (RHEL6.6)
370 2.6.32-573.26.1.el6 (RHEL6.7)
371 2.6.32-642.11.1.el6 (RHEL6.8)
372 3.10.0-514.el7 (RHEL7.3)
373 3.0.101-0.47.71 (SLES11 SP3)
374 3.0.101-80 (SLES11 SP4)
375 3.12.62-60.64.8 (SLES12 SP1)
376 vanilla linux 4.5.7 (ZFS only)
377 * Client known to build on unpatched kernels:
378 2.6.32-431.29.2.el6 (RHEL6.5)
379 2.6.32-504.30.3.el6 (RHEL6.6)
380 2.6.32-573.26.1.el6 (RHEL6.7)
381 2.6.32-642.11.1.el6 (RHEL6.8)
382 3.10.0-514.el7 (RHEL7.3)
383 3.0.101-0.47.71 (SLES11 SP3)
384 3.0.101-80 (SLES11 SP4)
385 3.12.62-60.64.8 (SLES12 SP1)
387 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
388 * Recommended ZFS / SPL version: 0.6.5.8
389 * Tested with ZFS / SPL version: 0.6.5.8
390 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
391 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
392 stack overflow. For more information, please refer to bugzilla 17630.
393 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
394 the re-exporting nodes to be mounted with "32bitapi" mount option
396 --------------------------------------------------------------------------------
398 02-29-2016 Intel Corporation
400 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
401 for currently supported client and server kernel versions.
402 * Server known to build on patched kernels:
403 2.6.32-431.29.2.el6 (RHEL6.5)
404 2.6.32-504.30.3.el6 (RHEL6.6)
405 2.6.32-573.12.1.el6 (RHEL6.7)
406 3.10.0-327.3.1.el7 (RHEL7.2)
407 3.0.101-0.47.71 (SLES11 SP3)
408 3.0.101-68 (SLES11 SP4)
409 vanilla linux 4.2.1 (ZFS only)
410 * Client known to build on unpatched kernels:
411 2.6.32-431.29.2.el6 (RHEL6.5)
412 2.6.32-504.30.3.el6 (RHEL6.6)
413 2.6.32-573.12.1.el6 (RHEL6.7)
414 3.10.0-327.3.1.el7 (RHEL7.2)
415 3.0.101-0.47.71 (SLES11 SP3)
416 3.0.101-68 (SLES11 SP4)
419 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
420 * Recommended ZFS / SPL version: 0.6.4.2
421 * Tested with ZFS / SPL version: 0.6.4.2
422 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
423 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
424 stack overflow. For more information, please refer to bugzilla 17630.
425 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
426 the re-exporting nodes to be mounted with "32bitapi" mount option
428 --------------------------------------------------------------------------------
430 03-10-2015 Intel Corporation
432 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
433 for currently supported client and server kernel versions.
434 * Server known to build on patched kernels:
435 2.6.32-431.29.2.el6 (RHEL6.5)
436 2.6.32-504.8.1.el6 (RHEL6.6)
437 3.0.101-0.46 (SLES11 SP3)
438 * Client known to build on unpatched kernels:
439 2.6.32-431.29.2.el6 (RHEL6.5)
440 2.6.32-504.8.1.el6 (RHEL6.6)
441 3.10.0-123.20.1.el7 (RHEL7)
442 3.0.101-0.46 (SLES11 SP3)
443 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
444 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
445 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
446 stack overflow. For more information, please refer to bugzilla 17630.
447 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
448 the re-exporting nodes to be mounted with "32bitapi" mount option
450 Severity : enhancement
452 Description: control OST-index in IDIF via ROCOMPAT flag.
453 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
454 last_rcvd file. For new formatted OST device, it will be auto set;
455 for the case of upgrading from old OST device, you can enable it
456 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
457 enabled, for new created OST-object, its IDIF-in-LMA will contain
458 the OST-index; for the existing OST-object, the OSD will convert
459 old format IDIF as new format IDIF with OST-index stored in the
460 LMA EA when accessing such OST-object or via OI scrub. Once such
461 flag is enabled, it cannot be reverted back, so the system cannot
462 be downgraded to the orignal incompatible version.
464 --------------------------------------------------------------------------------
466 07-30-2014 Intel Corporation
468 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
469 for currently supported client and server kernel versions.
470 * Server known to build on patched kernels:
471 2.6.32-431.20.3.el6 (RHEL6.5)
472 3.0.101-0.31 (SLES11 SP3)
473 * Client known to build on unpatched kernels:
474 2.6.32-431.20.3.el6 (RHEL6.5)
475 3.0.101-0.31 (SLES11 SP3)
478 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
479 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
480 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
481 stack overflow. For more information, please refer to bugzilla 17630.
482 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
483 the re-exporting nodes to be mounted with "32bitapi" mount option
485 --------------------------------------------------------------------------------
487 10-31-2013 Intel Corporation
489 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
490 * Server builds for kernels:
491 2.6.32-358.18.1.el6 (RHEL6.4)
492 * Client builds for unpatched kernels:
493 2.6.32-358.18.1.el6 (RHEL6.4)
494 3.0.92-0.8 (SLES11 SP3)
496 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
497 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
498 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
499 stack overflow. For more information, please refer to bugzilla 17630.
500 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
501 the re-exporting nodes to be mounted with "32bitapi" mount option
503 --------------------------------------------------------------------------------
505 05-30-2013 Intel Corporation
507 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
508 * Server builds for kernels:
509 2.6.32-358.6.2.el6 (RHEL6.4)
510 * Client builds for unpatched kernels:
511 2.6.32-358.6.2.el6 (RHEL6.4)
512 3.0.74-0.6.6 (SLES11 SP2)
514 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
515 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
516 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
517 stack overflow. For more information, please refer to bugzilla 17630.
518 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
519 the re-exporting nodes to be mounted with "32bitapi" mount option
521 --------------------------------------------------------------------------------
523 9-30-2012 Whamcloud, Inc.
525 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
526 * Server support for kernels:
527 2.6.32-279.5.1.el6 (RHEL6)
528 * Client support for unpatched kernels:
529 2.6.32-279.5.1.el6 (RHEL6)
531 2.6.18-238.19.1.el5 (RHEL5)
532 2.6.16.60-0.69.1 (SLES10)
533 * Recommended e2fsprogs version: 1.41.90.wc4
534 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
535 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
536 stack overflow. For more information, please refer to bugzilla 17630.
537 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
538 the re-exporting nodes to be mounted with "32bitapi" mount option
540 Severity : enhancement
542 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
544 Severity : enhancement
546 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
548 Severity : enhancement
550 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
552 Severity : enhancement
554 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
556 Severity : enhancement
558 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
560 --------------------------------------------------------------------------------
562 04-30-2012 Whamcloud, Inc.
564 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
565 * Server support for kernels:
566 2.6.32-220.4.2.el6 (RHEL6)
567 * Client support for unpatched kernels:
568 2.6.32-220.4.2.el6 (RHEL6)
570 2.6.18-238.19.1.el5 (RHEL5)
571 2.6.16.60-0.69.1 (SLES10)
572 * Recommended e2fsprogs version: 1.41.90.wc4
573 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
574 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
575 stack overflow. For more information, please refer to bugzilla 17630.
576 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
577 the re-exporting nodes to be mounted with "32bitapi" mount option
579 Severity : enhancement
581 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
583 Severity : enhancement
585 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
587 Severity : enhancement
589 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
591 Severity : enhancement
593 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
595 --------------------------------------------------------------------------------
597 09-30-2011 Whamcloud, Inc.
599 * Server support for kernels:
600 2.6.18-238.19.1.el5 (RHEL5)
601 2.6.32-131.6.1.el6 (RHEL6)
602 * Client support for unpatched kernels:
603 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
604 2.6.18-238.19.1.el5 (RHEL5)
605 2.6.32-131.6.1.el6 (RHEL6)
606 2.6.29.4-167.fc11 (FC11)
607 2.6.16.60-0.69.1 (SLES10)
609 * Recommended e2fsprogs version: 1.41.90.wc3
610 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
611 Lustre filesystem with 4K stack may cause a stack overflow. For more
612 information, please refer to bugzilla 17630.
613 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
614 mounted with "32bitapi" mount option
616 Severity : enhancement
617 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
619 Severity : enhancement
620 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
622 Severity : enhancement
623 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
626 Frequency : only with SLES10
628 Description: Use OFED "KMP" provided by Novell
629 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
630 Lustre is now built against this package. That means you need to
631 install the ofed-kmp package from Novell for the patchless client.
632 Note that the ofed-kmp that Novell ships may not exactly match the
633 kernel version but should still be compatible.
635 Severity : enhancement
637 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
638 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
640 --------------------------------------------------------------------------------
642 2010-07-15 Oracle, Inc.
644 * Server support for kernels:
645 2.6.18-164.11.1.el5 (RHEL 5)
646 2.6.18-164.11.1.0.1.el5 (OEL 5)
647 * Client support for unpatched kernels:
648 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
649 2.6.18-164.11.1.el5 (RHEL 5),
650 2.6.18-164.11.1.0.1.el5 (OEL 5)
651 2.6.16.60-0.42.8 (SLES 10),
653 2.6.29.4-167.fc11 (FC11)
654 * Recommended e2fsprogs version: 1.41.10-sun2
656 Severity : enhancement
658 Description: Distribute lfs_migrate script with lustre
659 Details : The lfs_migrate mentioned in the manual is now included in the
660 lustre tree in lustre/scripts and will distributed in lustre RPMs.
662 Severity : enhancement
664 Description: Failover nid list added to import proc file.
666 Severity : enhancement
668 Description: "som_preview" mount option required to use SOM
669 Details : The clients will need to use som_preview mount option
673 Description: Lustre does not do 1MB IOs to HW RAID
674 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
676 Severity : enhancement
678 Description: lfs quota output is not very convenient for awk/sed-parsing
679 Details : Some positions in lfs quota output table could be empty or
680 non-empty which made it hard to parse it with scripts, now
681 a dash is put instead of space where there is not supposed
686 Description: Update to OFED 1.5.1
687 Details : OFED 1.5.1 will be used in environments that don't use
688 vendor supplied OFED.
692 Description: Remove lnet. prefix from lctl params display
693 Details : Also renamed lnet.memused to lnet_memused to
698 Description: Use conf_param -d to remove permanant settings.
702 Description: Add list_param -R
703 Details : lctl list_param now supports recursive mode.
706 Bugzilla : 21829, 22312
707 Description: Fix llobdstat.
708 Details : 1. Now updating the write_bytes counter
709 2. Port script fixes from 1.8.
713 Description: decrease the usage of memory on clients.
714 Details : 1. On clients, recycle dentries and inodes unused.
715 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
720 Description: llobdstat fix and enhancement
721 Details : add a counter to set a limit to how many samples will be returned
722 fix a wildcard in the path to limit to obdfilter stats only
726 Description: lfs check only allowed for root
727 Details : Code cleanups and added a regresion test of non-root use of
732 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
733 Details : Fix hr->hr_index overload.
737 Description: Fix FIEMAP functionality.
741 Description: FC11 patchless client support added
745 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
746 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
750 Description: New acc-sm script and various test-framework enhancments.
754 Description: Previously landed patch was reverted.
758 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
762 Description: Add ldiskfs maxdirsize mount option.
766 Description: Fix mkfs.lustre support for 16TB LUNs.
770 Description: Print mmp_check_interval
771 Details : print mmp_check_interval and make it possible to abort
772 mount operation in case it takes too long.
776 Description: Add cascading_rw to lustre mpi tests.
780 Description: Store SOM data in EA.
781 Details : Also several SOM related cleanups.
785 Description: Add refcount to md_open_data
789 Description: Use refcount on lov_request_set to prevent memory corruption.
793 Description: Statahead: use parent dir's i_mutex for synchronization.
797 Description: loadgen improvements.
801 Description: Enable peer health detection in ksocklnd and ko2iblnd.
805 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
809 Description: Fix race on obd_devcie:md_stats
813 Description: Fix SLES 11 client builds.
814 Details : Resolved uninitalized pointer gcc warnings.
818 Description: Allow non-root access for "lfs check".
819 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
823 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
825 Severity : enhancement
827 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
828 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
832 Description: Sanity test_27p fails.
833 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
834 Sanity tests 27 were cleaned up and improved.
838 Description: Unify req format on client/servers
839 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
840 buffers and swabbers approach..
845 Description: drop join file stuff.
847 Severity : enhancement
849 Description: Add 2.6.27 and SLES11 (patchless client) support
853 Description: Fix in ptlrpc_expire_one_request() to print the signed time
856 Severity : enhancement
858 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
861 Severity : enhancement
863 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
865 Severity : enhancement
867 Description: correctly handle big reply message.
868 Details : send LNet event if reply is bigger than buffer and adjust this buffer
873 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
875 Severity : enhancement
877 Description: Add support for OEL5.
881 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
883 Severity : enhancement
885 Description: "weak-modules" support
886 Details : Implement "weak-modules" support which enables kernel modules
887 to be used with any kernel that implements the same kABI. In
888 order to achieve this modules are now installed in
889 /lib/modules/$(uname -r)/updates/kernel on all distributions.
893 Description: client could not reconnect to OST because of an active request.
894 Details : The client now retries to reconnect to the same server, if a
895 connect request failed with EBUSY or -EAGAIN. Test case 35b
896 conf-sanity.sh has been created.
900 Description: Don't allow make backward step on assignin osc next id.
901 Details : race between allocation next id and ll_sync thread can be cause
902 of set wrong osc next id and can be kill valid ost objects.
906 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
910 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
914 Description: Error handling in osc_statfs_interpret() has been improved.
915 Details : Check in osc_statfs_interpret() for EBADR.
919 Description: Do not send statfs() requests to OSTs disabled by administrator.
920 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
922 Severity : enhancement
924 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
926 Severity : enhancement
928 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
930 Severity : enhancement
932 Description: Enhancement for lfs(1) command to use numeric uid/gid.
936 Description: Improved cleanup procedure for conf-sanity test_46a.
940 Description: Do not start bulk transfer, if client tried to reconnect.
941 Details : Do not start bulk transfer, if client tried to reconnect,
942 but failed due to active rpcs. Check for exp_abort_active_req
943 has been implemented.
947 Description: Prevent inconsistences between linux and lustre mount structures.
948 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
949 Make the sleep interruptible.
951 Severity : enhancement
953 Description: provide server to client comms path
954 Details : server to client communications path via new LDLM_SET_INFO rpc
958 Description: truncate starts GFP_FS allocation under transaction causing deadlock
959 Details : ldiskfs_truncate calls grab_cache_page which may start page
960 allocation under an open transaction. This may lead to
961 calling prune_icache with consequent lustre reentrance.
965 Description: send correctly lsm on open replay
966 Details : MDS is trust to LSM size on replay open, but client can set wrong size
971 Description: enable client interface failover
972 Details : When a child reconnects from another NID, properly update export
973 nid hash position and ldlm reverse import.
975 Severity : enhancement
977 Description: Communicate OST degraded/readonly state via statfs to MDS
978 Details : Flags in the statfs returned from OSTs indicate whether the
979 OST is in a degraded RAID state, or if the filesystem has
980 turned read-only after a filesystem error is detected.
985 Description: don't panic if EPROTO was hit when reading symlink
986 Details : correctly handling request reference in error cases.
988 Severity : enhancement
990 Description: Add LustreNetLink, a kernel-userspace communcation path.
992 Severity : enhancement
994 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
999 Description: access to llog context before init.
1000 Details : move handling CATALOGS file at osc layer and forbid access to llog
1001 context before init.
1004 Frequency : rare, only when client is writing to OST on local node
1006 Description: Avoid deadlock for local client writes
1007 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1008 memory freeing context. This allows OST threads to set the
1009 PF_MEMALLOC flag on task structures in order to allocate memory
1010 from reserved pools and complete IO.
1011 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1012 so that the OST threads generate memory pressure and allow
1013 inactive pages to be reclaimed.
1015 Severity : enhancement
1017 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1022 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1023 Details : move ldlm namespace creation in setup phase to avoid grab
1024 _lprocfs_lock with cli_sem held.
1028 Description: Temporarily disable grant shrink.
1029 Details : Disable the feature for debugging.
1032 Frequency : only when formatting test filesystems
1034 Description: Unable to run several mkfs.lustre on loop devices at the same time
1035 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1036 formatting. The solution is to proper handle the error.
1038 Severity : enhancement
1040 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1042 Severity : enhancement
1044 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1049 Description: resolve race between obd_disconnect and class_disconnect_exports
1050 Details : if obd_disconnect will be called to already disconnected export he
1051 forget release one reference and osc module can't unloaded.
1053 Severity : enhancement
1055 Description: Allow tuning service thread via /proc
1056 Details : For each service a new
1057 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1058 created that can be used to set min/max thread counts, and get the
1059 current number of running threads.
1064 Descriptoin: don't return error if have partially created objects for file.
1065 Details : lov_update_create_set uses set->set_success as index for created
1066 objects, so if some requests will be failed, they will have hole
1067 at end of array and qos_shrink_lsm for allocate correct lsm.
1069 Severity : enhancement
1071 Description: Update OFED support to 1.4.1
1076 Description: Reduce small size read RPC
1077 Details : Set read-ahead limit for every file and only do read-ahead when
1078 available read-ahead pages are bigger than 1M to avoid small size
1081 Severity : enhancement
1083 Description: move AT tunable parameters for more consistent usage
1084 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1086 Severity : enhancement
1088 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1089 Details : allow skip disconnected ost for send statfs request and hide error
1095 Description: fix racy locking of mballoc block bitmaps causing BUG
1096 Details : The locking of the mballoc buddy bitmap and the in-memory
1097 block bitmap was using two different spin locks in some
1098 cases. This made it possible to incorrectly access the
1099 mballoc bitmap while another process was modifying it,
1100 causing a sanity assertion to fail. While no on-disk corruption
1101 was reported, there was some risk of this happening.
1104 Frequency : rare, on llog test 6
1106 Descriptoin: don't allow connect to already connected import
1107 Details : allowing connect to already connected import is hide connecting problem.
1112 Descriptoin: allow kill process which wait statahead result
1113 Details : for some reasons 'ls' can stick in waiting result from statahead,
1114 in this case need way for kill this process.
1116 Severity : enhancement
1118 Description: Add state history info file, enhance import info file
1119 Details : Track import connection state changes in a new osc/mdc proc file;
1120 add overview-type data to the osc/mdc import proc file.
1122 Severity : enhancement
1124 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1126 Severity : enhancement
1128 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1133 Descriptoin: OST grant too much space to client even there are not enough space.
1134 Details : Client will shrink its grant cache to OST if there are no write
1135 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1136 this grant cache if there are already not enough avaible space
1137 (left_space < total_clients * 32M).
1140 Frequency : start MDS on uncleanly shutdowned MDS device
1142 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1143 Details : stay in waiting mds<>ost recovery finished produce random bugs
1144 due race between two ll_sync thread for one lov target. send
1145 ACTIVATE event only if connect realy finished and import have
1149 Frequency : rare, connect and disconnect target at same time
1151 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1152 Details : don't call obd_disconnect under lov_lock. this long time
1153 operation and can block ptlrpcd which answer to connect request.
1158 Descriptoin: don't lose wakeup for imp_recovery_waitq
1159 Details : recover_import_no_retry or invalidate_import and import_close can
1160 both sleep on imp_recovery_waitq, but we was send only one wakeup
1164 Frequency : always with long access acl
1166 Descriptoin: mds can't pack reply with long acl.
1167 Details : mds don't control size of acl but they limited by reint/getattr
1170 Severity : enhancement
1172 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1174 Severity : enhancement
1176 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1179 Frequency : start MDS on uncleanly shutdowned MDS device
1181 Descriptoin: aborting recovery hang on MDS
1182 Details : don't throttle destroy RPCs for the MDT.
1185 Frequency : on remount
1187 Description: external journal device not working after the remount
1188 Details : clear dev_rdonly flag for external journal devices in
1194 Description: shutdown vs evict race
1195 Details : client_disconnect_export vs connect request race.
1196 if client will evicted at this time - we start invalidate
1197 thread without referece to import and import can be freed
1201 Frequency : rare, need acl's on inode.
1203 Description: client can't handle ost additional correctly
1204 Details : if ost was added after client connected to mds client can have
1205 hit lnet_try_match_md ... to big messages to wide striped files.
1206 in this case need teach client to handle config events about add
1207 lov target and update client max ea size at that event.
1209 Severity : enhancement
1211 Description: Changelogs
1212 Details : Changelogs are a lightweight mechanism to track filesystem
1213 metadata and namespace changes. The changelog is recorded
1214 permanently on the MDTs, and is periodically "consumed" / purged
1215 when records are no longer needed.
1217 Severity : enhancement
1219 Description: compact fld format with extents
1220 Details : Store range of seq rather than every seq in FLD. Seq
1221 controller update FLD rather than clients. In Case of CMD, mdt0
1222 has FLD, all other metadata server act as non persistent proxy
1223 for FLD queries and cache fld entries in fld cache.
1228 Description: don't skip ost target if they assigned to file
1229 Details : Drop slow OSCs if we can, but not for requested start idx.
1230 This means "if OSC is slow and it is not the requested
1231 start OST, then it can be skipped, otherwise skip it only
1232 if it is inactive/recovering/out-of-space.
1236 Description: more cleanup in mds_lov
1237 Details : not send LOV EA under replay, we can't know about they size at this
1238 time. Don't allow client connect to mds before any ost connected,
1239 for avoid problems with LOV EA size and returning EIO to client.
1241 Severity : enhancement
1243 Description: Interoperability at server side (Disk interoperability)
1245 Severity : enhancement
1247 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1249 Severity : enhancement
1251 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1253 Severity : enhancement
1255 Description: New client IO stack (CLIO).
1257 Severity : enhancement
1259 Description: Commit on sharing. Eliminate inter-client dependencies between
1260 uncommitted transactions by doing transaction commits.
1261 Thereby clients may recovery independently.
1264 Frequency : Create a symlink file with a very long name
1266 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1267 Details : If there is no extra space in the request for early cancels,
1268 ldlm_req_handles_avail() returns 0 instead of a negative value.
1270 Severity : enhancement
1272 Description: Add /proc entry for import status
1273 Details : The mdc, osc, and mgc import directories now have
1274 an import directory that contains useful import data for debugging
1275 connection problems.
1277 Severity : enhancement
1279 Description: Re-disable certain /proc logging
1280 Details : Enable and disable client's offset_stats, extents_stats and
1281 extents_stats_per_process stats logging on the fly.
1284 Frequency : Only on FC kernels 2.6.22+
1286 Description: oops in statahead
1287 Details : Do not drop reference count for the dentry from VFS when lookup,
1288 VFS will do that by itself.
1290 Severity : enhancement
1292 Description: Generic /proc file permissions
1293 Details : Set /Proc file permissions in a more generic way to enable non-
1294 root users operate on some /proc files.
1298 Description: Hitting mdc_commit_close() ASSERTION
1299 Details : Properly handle request reference release in
1300 ll_release_openhandle().
1304 Description: quota recovery deadlock during mds failover
1305 Details : This patch includes att18982, att18236, att18237 in bz14840.
1307 1. fix osts hang when mds does failover with quotaon
1308 2. prevent watchdog storm when osts threads wait for the
1313 Frequency : only patchless client
1314 Description: add workaround for race between add/remove dentry from hash
1316 Severity : enhancement
1318 Description: Allow OST glimpses to return PW locks
1322 Description: LBUG when llog conf file is full
1323 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1327 Description: Prevent import from entering FULL state when server in recovery
1331 Description: service mount cannot take device name with ":"
1332 Details : Only when device name contains ":/" will mount treat it as
1338 Description: replace ptlrpcd with the statahead thread to interpret the async
1339 statahead RPC callback
1343 Frequency : on recovery
1344 Description: I/O failures after umount during fail back
1345 Details : if client reconnected to restarted server we need join to recovery
1346 instead of find server handler is changed and process self
1347 eviction with cancel all locks.
1349 Severity : enhancement
1351 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1353 Severity : enhancement
1355 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1357 Severity : enhancement
1359 Description: Upcall on Lustre log has been dumped
1360 Details : Allow for a user mode script to be called once a Lustre log has
1361 been dumped. It passes the filename of the dumped log to the
1362 script, the location of the script can be specified via
1363 /proc/sys/lnet/debug_log_upcall.
1368 Description: avoid idr_remove called for id which is not allocated.
1369 Details : Move assigment s_dev for clustered nfs to end of initialization,
1370 to avoid problem with error handling.
1375 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1376 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1377 into CONN_USED_HASH and this prodice warning when put connection
1378 again in unused hash.
1381 Severity : enhancement
1383 Description: File striping can now be set to use an arbitrary pool of OSTs.
1385 Severity : enhancement
1387 Description: Export bytes_read/bytes_write count on OSC/OST.
1391 Description: Early reply size mismatch, MGC loses connection
1392 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1393 the connect flags are properly negotiated.
1398 Description: quotas are not honored with O_DIRECT
1399 Details : all writes with the flag O_DIRECT will use grants which leads to
1400 this problem. Now using OBD_BRW_SYNC to guard this.
1404 Description: add quota statistics
1405 Details : 1. sort out quota proc entries and proc code.
1406 2. add quota statistics
1408 Severity : enhancement
1410 Description: enable quota support for HEAD.
1414 Description: Properly propagate oinfo flags from lov to osc for statfs
1415 Details : restore missing copy oi_flags to lov requests.
1417 Severity : enhancement
1419 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1420 : l_getgroups(8), lst(8), routerstat(8)
1421 Details : included man pages for llobdstat(8), llstat(8),
1422 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1424 Severity : enhancement
1426 Description: configure's --enable-quota should check the
1427 : kernel .config for CONFIG_QUOTA
1428 Details : configure is terminated if --enable-quota is passed but
1429 : no quota support is in kernel
1433 Description: Remove portals compatibility
1434 Details : Remove portals compatibility, not interoperable with releases
1439 Description: Resolve device initialization race
1440 Details : Prevent proc handler from accessing devices added to the
1441 obd_devs array but yet be intialized.
1443 Severity : enhancement
1445 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1447 Severity : enhancement
1449 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1454 Description: Using wrong pointer in osc_brw_prep_request
1455 Details : Access to array[-1] can produce panic if kernel compiled with
1456 CONFIG_PAGE_ALLOC enabled
1460 Description: Client runs out of low memory
1461 Details : Consider only lowmem when counting initial number of llap pages
1465 Description: *optional* service tags registration
1466 Details : if the "service tags" package is installed on a Lustre node
1467 When the filesystem is mounted, a local-node service tag will
1468 be created. See http://inventory.sun.com/ for more information
1469 about the Service Tags asset management system.
1473 Description: Kernel BUG tries to release flock
1474 Details : Lustre does not destroy flock lock before last reference goes
1475 away. So always drop flock locks when client is evicted and
1476 perform unlock regardless of successfulness of speaking to MDS.
1480 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1484 Description: sanity test 65a fails if stripecount of -1 is set
1485 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1490 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1491 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1494 Severity : enhancement
1496 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1500 Description: do not process already freed flock
1501 Details : flock can possibly be freed by another thread before it reaches
1502 to ldlm_flock_completion_ast.
1506 Description: LBUG during stress test
1507 Details : Need properly lock accesses the flock deadlock detection list.
1511 Description: oops in page fault handler
1512 Details : kernel page fault handler can return two special 'pages' in error
1513 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1517 Description: timeout with invalidate import.
1518 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1519 should be handled by ptlrpcd. This produce long age waiting and
1520 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1522 Severity : enhancement
1524 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1528 Description: enable MGS and MDT services start separately
1529 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1530 a co-located MGS without starting the MGS, which is a complement
1531 to 'nosvc' mount option.
1535 Frequency : after recovery
1536 Description: precreate to many object's after del orphan.
1537 Details : del orphan st in oscc last_id == next_id and this triger growing
1538 count of precreated objects. Set flag LOW to skip increase count
1539 of precreated objects.
1543 Frequency : rare, on clear nid stats
1544 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1545 Details : when clean nid stats sometimes try destroy live entry,
1546 and this produce panic in free.
1550 Description: Stack overflow during MDS log replay
1551 ease stack pressure by using a thread dealing llog_process.
1555 Description: wait until IO finished before start new when do lock cancel.
1556 Details : VM protocol want old IO finished before start new, in this case
1557 need wait until PG_writeback is cleared until check dirty flag and
1558 call writepages in lock cancel callback.
1560 Severity : enhancement
1562 Description: using special macro for print time and cleanup in includes.
1566 Description: mds_mfd_close() ASSERTION(rc == 0)
1567 Details : In mds_mfd_close(), we need protect inode's writecount change
1568 within its orphan write semaphore to prevent possible races.
1572 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1576 Frequency : rare, on shutdown ost
1577 Description: don't hit live lock with umount ost.
1578 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1579 use shrink_dcache_sb instead.
1583 Description: don't panic with use echo client
1584 Details : echo client pass NULL as client nid pointer and this produce null
1585 pointer dereference.
1589 Description: fix build on ppc32
1590 Details : compile code with -m64 flag produce wrong object file for ppc32.
1594 Description: add message levels for liblustreapi
1598 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1599 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1600 will result in an -EAGAIN and in the exit call path we call
1601 alloc_rr() which will with increasing aggressiveness attempt to
1602 aquire precreated objects on the minimum number of required OSCs.
1606 Description: Use old size assignment to avoid deadlock
1607 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1608 the scheduling while holding a spinlock. We do not need locking
1609 for size in ll_update_inode() because size is only updated from
1610 the MDS for directories or files without objects, so there is no
1611 other place to do the update, and concurrent access to such inodes
1612 are protected by the inode lock.
1616 Description: resolve "_IOWR redefined" build error on SLES10
1620 Description: dump the memory debugging after all modules are unloaded to
1621 suppress false negative in conf_sanity test 39
1623 Severity : enhancement
1625 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1630 Description: fix SLES kernel versioning
1631 Details : the kernel version for our SLES 10 kernel did not include a "-"
1632 before the "smp" at the end. while this was not a problem in
1633 general, it did mean that software trying to use the kernel
1634 version to try to detect a vendor specific kernel would fail.
1635 this was most evident by the OFED build scripts.
1639 Description: Don't update lov_desc members until making sure they are valid
1640 Details : When updating lov_desc members via proc fs, need fix their
1641 validities before doing the real update.
1645 Description: don't put request into delay list while invalidate in flight.
1646 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1647 import in flight. this produce timeout for invalidate and sometimes
1648 can cause stale data.
1652 Frequency : on ppc only
1653 Description: not convert ost objects for directory because it's not exist.
1654 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1658 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1660 Severity : enhancement
1662 Description: Update to support 2.6.22.14 vanilla kernel.
1666 Frequency : rare, on recovery
1667 Description: read procfs can produce deadlock in some situation
1668 Details : Holding lprocfs lock which send rpc can produce block for destroy
1669 obd objects and this also block reconnect with -EALREADY. This
1670 isn't fix all lprocfs bugs - but make it rare.
1672 Severity : enhancement
1674 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1677 Frequency : frequent on X2 node
1679 Description: mdc_set_open_replay_data LBUG
1680 Details : Set replay data for requests that are eligible for replay.
1684 Description: lustre_mgs: operation 101 on unconnected MGS
1685 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1686 MGC, and late on MGC cannot successfully connect to MGS and a lot
1687 of the error messages complaining that MGS is not connected.
1691 Frequency : on network error
1692 Description: panic with double free request if network error
1693 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1694 but it's true only for synchronus enqueue, for async enqueue
1695 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1698 Severity : enhancement
1700 Description: client-side metadata stat-ahead during readdir(directory readahead)
1701 Details : perform client-side metadata stat-ahead when the client detects
1702 readdir and sequential stat of dir entries therein
1705 Frequency : on start mds
1707 Description: Implement get_info(last_id) in obdfilter.
1710 Frequency : occasional
1712 Description: Correctly check stale fid, not start epoch if ost not support SOM
1713 Details : open with flag O_CREATE need set old fid in op_fid3 because
1714 op_fid2 was overwritten with new generated fid, but mds can answer
1715 with one of these two fids and both is not stale. Setattr
1716 incorrectly started an epoch and assume will be called
1717 done_writing, but without SOM done_writing ever being called.
1720 Frequency : rare, depends on device drivers and load
1722 Description: MDS or OSS nodes crash due to stack overflow
1723 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1724 In some cases, in conjunction with device drivers that use a lot
1725 of stack the MDS (or possibly OSS) service threads could overflow
1726 the stack. One change which was identified to consume additional
1727 stack has been reworked to avoid the extra stack usage.
1730 Frequency : occasional
1732 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1733 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1734 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1735 oscc_has_objects and see OSC has no objests and call
1736 oscc_internal_create to resend create request.
1738 Severity : enhancement
1740 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1742 Severity : enhancement
1744 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1747 Frequency : very rare
1749 Description: Fix replay if there is an un-replied request and open
1750 Details : In some cases, older replay request will revert the
1751 mcd->mcd_last_xid on MDS which is used to record the client's
1752 latest sent request.
1754 Severity : enhancement
1756 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1758 Severity : enhancement
1760 Description: Add rhel5 support to HEAD.
1762 Serverity : enhancement
1764 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1769 Description: Don't allow skipping OSTs if index has been specified.
1770 Details : Don't allow skipping OSTs if index has been specified, make
1771 locking in internal create lots better.
1775 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1776 Details : access to bitfield in structure is always rounded to long
1777 and this produce problem with not atomic change any bit.
1781 Description: Lustre make rpms failed.
1782 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1783 builds Lustre rpms from tarball.
1786 Frequency : rare on shutdown ost
1788 Description: If llog cancel was not send before clean_exports phase, this can
1789 produce deadlock in llog code.
1790 Details : If llog thread has last reference to obd and call class_import_put
1791 this produce deadlock because llog_cleanup_commit_master wait when
1792 last llog_commit_thread exited, but this never success because was
1793 called from llog_commit_thread.
1797 Description: allow userland application know is lost one of stripes.
1798 Details : fill lvb_blocks with error code on ost and return it to
1799 application if error flag found.
1803 Description: NULL lov_tgts causing MDS oops
1804 Details : more safe checks for NULL lov_tgts for avoid oops.
1806 Severity : enhancement
1808 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1812 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1813 Details : If an OST is down the MDS will hang indefinitely in
1814 obd_statfs_rqset() waiting for the statfs data. While for
1815 MDS QOS usage of statfs, it should not stuck in waiting.
1817 Severity : enhancement
1819 Description: remote_acl support
1820 Details : Support ACL-based permission check for remote user.
1821 Support setfacl/getfacl for remote user with the utils
1822 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1823 the system "{s,g}etfacl" utils.
1825 Severity : enhancement
1827 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1829 Severity : enhancement
1831 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1835 Description: make mgs_setparam() handle fsname containing dash
1836 Details : fsname containing a dash does not work with lctl conf_param
1838 Severity : enhancement
1840 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1842 Severity : enhancement
1844 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1848 Frequency : rare, at shutdown
1849 Description: access already free / zero obd_namespace.
1850 Details : if client_disconnect_export was called without force flag set,
1851 and exist connect request in flight, this can produce access to
1852 NULL pointer (or already free pointer) when connect_interpret
1853 store ocd flags in obd_namespace.
1857 Frequency : only at startup
1858 Description: not alloc memory with spinlock held.
1859 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1860 if any spinlock held.
1862 Severity : enhancement
1864 Description: make lustre randomly fail allocating memory
1865 Details : Make lustre randomly failed allocating memory for testing purpose.
1867 Severity : enhancement
1869 Description: lost problems with lov objid file
1870 Details : Fixes some scability and access to not inited memory problems
1871 in work with lov objdid file.
1876 Description: lfs find does not continue on file error
1877 Details : Continue other files processing when a file/dir is absent.
1881 Description: Inconsistent usage of lustre_pack_reply()
1882 Details : Standardize the usage of lustre_pack_reply() such that it
1883 always generate a CERROR on failure.
1886 Frequency : occasional
1888 Description: MDS hang or stay in waiting lock
1889 Details : If client receive lock with CBPENDING flag ldlm need send lock
1890 cancel as separate rpc, to avoid situation when cancel request
1891 can't processed due all i/o threads stay in wait lock.
1895 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1899 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1903 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1907 Description: Update to RHEL4 latest kernel.
1909 Severity : enhancement
1911 Description: Build SLES10 patchless client fails
1912 Details : The configure was broken by run ./configure with
1913 --with-linux-obj=.... argument for patchless client. When the
1914 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1915 can't search header adequately. Use absolute path such as
1916 -I($LINUX)/include instead.
1920 Description: interrupt oig_wait produce painc on resend.
1921 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1922 private set, and this produce situation when rq_ptlrpcd_data not
1923 copyed to new allocated request and triggered LBUG on assert
1924 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1925 ptlrpcd set if request is changed and can be safety changed to use
1930 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1931 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1932 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1933 these macros and replace them with their difinitions instead.
1937 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1938 Details : Update which_patch & target file for SLES10 latest kernel.
1941 Description: Remove client patches from SLES 10 kernel.
1942 Details : This causes SLES 10 clients to behave as patchless clients
1943 even on a Lustre-patched (server) kernel.
1945 Severity : enhancement
1947 Description: self-adjustable client's lru lists
1948 Details : use adaptive algorithm for managing client cached locks lru
1949 lists according to current server load, other client's work
1950 pattern, memory activities, etc. Both, server and client
1951 side namespaces provide number of proc tunables for controlling
1954 Severity : enhancement
1956 Description: light-weight GSS support
1957 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1958 and reduce performance overhead.
1960 Severity : enhancement
1962 Description: Linux keyring support
1963 Details : Support using service of Linux keyring for Lustre GSS internal
1964 context refresh/cache mechanism.
1968 Description: Fix errors in lfs documentation
1969 Details : Fixes man pages
1973 Description: don't use GFP_* in generic Lustre code.
1974 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1978 Description: obdclass is limited by single OBD_ALLOC(idarray)
1979 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1983 Description: warnings with build patchless client with vanila 2.6.19 and up
1984 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1985 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1989 Frequency : Only for SLES
1990 Description: sanity_quota fail test_1
1991 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1992 file, which makes TSTID[2] inunique.
1996 Frequency : Only for SLES9
1997 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1998 Details : Update target/ChangeLog/which_patch .
2002 Frequency : Only for test_10 in sanity-quota.sh
2003 Description: a bug in quota test code
2004 Details : checking a test flag in a wrong place
2008 Frequency : Only for test_13 in sanity-quota.sh
2009 Description: a bug in quota test script
2010 Details : shouldn't check overfull allocated quota
2014 Description: jbd statistics
2015 Details : Port older jbd statistics patch for sles10
2017 Severity : enhancement
2019 Description: Add printing inode info into message about error in writepage.
2024 Description: reply_lock_interpret crash due to race with it and lock cancel.
2025 Details : Do not replay locks that are being cancelled. Do not reference
2026 locks by their address during replay, just by their handle.
2030 Frequency : When flocks are used.
2031 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2032 Details : Flock locks might destroy just granted lock if it could be merged
2033 with another existing flock, this is done in completion handler,
2034 so teach ldlm_cli_enquque_fini that this is a valid case for
2040 Description: Oops in read and write path when failing to allocate lock.
2041 Details : Check if lock allocation failed and return error back.
2045 Description: lstripe command fails for valid OST index
2046 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2047 instead of lov->desc.ld_active_tgt_count.
2051 Description: sanity.sh failed test 103
2052 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2057 Description: df doesn't work properly if diskfs blocksize != 4K
2058 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2062 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2063 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2065 Severity : enhancement
2067 Description: Adaptive timeouts
2068 Details : RPC timeouts adapt to changing server load and network
2069 conditions to reduce resend attempts and improve recovery time.
2073 Description: llapi_file_create() does not allow some changes
2074 Details : add llapi_file_open() that allows specifying the mode and
2075 open flags, and also returns an open file handle.
2079 Description: merge and cleanup kernel patches.
2080 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2084 Description: Add journal checksum support.(Kernel part)
2085 Details : The journal checksum feature adds two new flags i.e
2086 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2087 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2088 indicates that the commit block contains the checksum for
2089 the blocks described by the descriptor blocks. Now commit
2090 record can be sent to disk without waiting for descriptor
2091 blocks to be written to disk. This behavior is controlled
2092 using JBD2_FEATURE_ASYNC_COMMIT flag.
2096 Description: OSS needs mutliple precreate threads
2097 Details : Add ability to start more than one create thread per OSS.
2101 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2102 Details : execute lfs setstripe on client
2106 Description: mds_obd_create error creating tmp object
2107 Details : When the user sets quota on root, llog will be affected and can't
2108 create files and write files.
2111 Frequency : Always on ia64 patchless client, and possibly others.
2113 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2114 Details : This allows the patchless client to be loaded on architectures
2115 without this export.
2119 Description: RedHat Update kernel for RHEL5
2120 Details : Add the kernel config file for RHEL5.
2124 Description: RedHat Update kernel for RHEL5
2125 Details : Modify the kernel config file more closer RHEL5.
2129 Description: Build failure against Centos5 (RHEL5)
2130 Details : Define PAGE_SIZE when it isn't present.
2134 Description: test_7 on sanity-quota.sh can't be used on separate machines
2135 Details : add facet to handle it
2139 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2140 Details : with NFS, the anon dentry's parent was set to itself in
2141 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2142 corresponding dentry other than use rec->ur_name.
2144 Severity : enhancement
2146 Description: lfs setstripe enhancement
2147 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2151 Description: enable data checksumming by default
2152 Details : enable checksum by default, allow --disable-checksum
2153 configure option and "-o nochecksum" mount option. Checksums
2154 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2158 Description: lustre support for RHEL5
2159 Details : Add support for RHEL5.
2163 Description: Client eviction due to failover config
2164 Details : after a connection loss, the lustre client should attempt to
2165 reconnect to the last active server first before trying the
2166 other potential connections.
2170 Description: when mds and osts use different quota unit(32bit and 64bit),
2171 quota will be released repeatly.
2172 Details : void sending multiple quota reqs to mds, which will keep the
2173 status between the reqs.
2177 Description: osts not allocated evenly to files
2178 Details : change the condition to increase offset_idx
2181 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2183 Description: Data corruption for OSTs that are formatted larger than 2TB
2185 Details : When generating the bio request for lustre file writes the
2186 sector number would overflow a temporary variable before being
2187 used for the IO. The data reads correctly from Lustre (which
2188 will overflow in a similar manner) but other file data or
2189 filesystem metadata may be corrupted in some cases.
2193 Description: Tune the kernel for good SCSI performance.
2194 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2195 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2200 Description: rewrite ext2-derived code in obdclass/uuid.c
2201 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2202 from scratch preserving functionality.
2206 Description: "lfs find -obd UUID" prints directories
2207 Details : "lfs find -obd UUID" will return all directory names instead
2208 of just file names. It is incorrect because the directories
2209 do not reside on the OSTs.
2213 Description: use i_size_read and i_size_write in 2.6 port
2214 Details : replace inode->i_size access with i_size_read/write()
2217 Frequency : when removing large files
2219 Description: scheduling issue during removal of large Lustre files
2220 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2221 It causes scheduling issues when removing large files (17TB in the
2225 Frequency : only with liblustre clients on XT3
2227 Description: evictions taking too long
2228 Details : allow llrd to evict clients directly on OSTs
2231 Frequency : only on ppc
2233 Description: /proc/fs/lustre/devices broken on ppc
2234 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2235 We should make sure the type of 'index' is loff_t and then cast
2236 explicitly as needed below. Do not assign an explicitly cast
2240 Frequency : only for rhel5
2242 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2243 Details : Modify the target file & which_kernel.
2245 Severity : enhancement
2247 Description: omit set fsid for export NFS
2248 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2249 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2253 Frequency : Always, for kernels after 2.6.16
2254 Description: Fix warning idr_remove called for id=.. which is not allocated.
2255 Details : Last kernels save old s_dev before kill super and not allow
2256 to restore from callback - restore it before call kill_anon_super.
2260 Description: buffer overruns could theoretically occur
2261 Details : llapi_semantic_traverse() modifies the "path" argument by
2262 appending values to the end of the origin string, and a
2263 overrun may occur. Adding buffer overrun check in liblustreapi.
2267 Description: Fix error on 'ls .' at the top of the Lustre mount.
2268 Details : Don't revalidate dentry if it is a root dentry.
2272 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2273 Details : Modify vm-tunables-rhel4.patch.
2277 Description: Kernel config for 2.6.18-vanilla.
2278 Details : Modify targets/2.6-vanilla.target.in.
2279 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2280 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2281 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2282 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2286 Description: improve handling recoverable errors
2287 Details : if request processig with error which can be recoverable on server
2288 request should be resend, otherwise page released from cache and
2293 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2294 Details : Modify target file & which_patch.
2295 A flaw was found in the IA32 system call emulation provided
2296 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2297 value could be stored in the %RAX register, which could trigger an
2298 out-of-bounds system call table access. An untrusted local user
2299 could exploit this flaw to run code in the kernel
2300 (ie a root privilege escalation). (CVE-2007-4573).
2304 Description: O_DIRECT bypasses client statistics.
2305 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2306 read_ahead_stats were updated. copied the stats section from
2307 osc_send_oap_rpc() into async_internal().
2311 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2315 Description: change order of libsysio includes
2316 Details : '#include sysio.h' should always come before '#include xtio.h'
2320 Description: handle "serious error: objid * already exists" more gracefully
2321 Details : If LAST_ID value on disk is smaller than the objects existing in
2322 the O/0/d* directories, it indicates disk corruption and causes an
2323 LBUG(). If the object is 0-length, then we should use the existing
2324 object. This will help to avoid a full fsck in most cases.
2326 Severity : enhancement
2328 Description: adapt the lustre_config script to support the upgrade case
2329 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2330 server targets to 1.6 in parallel.
2334 Description: To avoid grant space > avaible space when the disk is almost
2335 full. Without this patch you might see the error "grant XXXX >
2336 available" or some LBUG about grant, when the disk is almost
2338 Details : In filter_check_grant, for non_grant cache write, we should
2339 check the left space by if (*left > ungranted + bytes), instead
2340 of (*left > ungranted), because only we are sure the left space
2341 is enough for another "bytes", then the ungrant space should be
2342 increase. In client, we should update cl_avail_grant only there
2343 is OBD_MD_FLGRANT in the reply.
2347 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2348 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2349 described in CVE-2007-4573. This problem could allow untrusted
2350 local users to gain root access.
2353 Frequency : when using O_DIRECT and quotas
2355 Description: Incorrect file ownership on O_DIRECT output files
2356 Details : block usage reported by 'lfs quota' does not take into account
2357 files that have been written with O_DIRECT.
2362 Description: touch file failed when fs is not full
2363 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2364 otherwise we can get ENOSP while fs is not full.
2368 Description: parallel lock callbacks
2369 Details : Instead of sending blocking and completion callbacks as separated
2370 requests, adding them to a set and sending in parallel.
2373 Frequency : only for Cray XT3
2374 Bugzilla : 12829/13455
2375 Description: Changing primary group doesn't change the group lustre assigns to
2377 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2378 overridden with the primary group provided by the group upcall,
2379 whereas the supplied fsgid can be trusted if it is in the list of
2380 supplementary groups returned by the group upcall.
2382 Severity : enhancement
2384 Description: Allow masking D_WARNING, D_ERROR messages from console
2385 Details : Console messages can now be disabled via lnet.printk.
2389 Description: User code with malformed file open parameter crashes client node
2390 Details : Before packing join_file req, all the related reference should be
2391 checked carefully in case some malformed flags cause fake
2392 join_file req on client.
2396 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2397 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2398 mode is equal to the granted mode. However, it should also check
2399 that LDLM_FL_CP_REQD is not set.
2403 Description: Heavy nfs access might result in deadlocks
2404 Details : After ELC code landed, it is now improper to enqueue any mds
2405 locks under och_sem, because enqueue might want to decide to
2406 cancel open locks for same inode we are holding och_sem for.
2410 Description: Client eviction while running blogbench
2411 Details : A lot of unlink operations with concurrent I/O can lead to a
2412 deadlock causing evictions. To address the problem, the number of
2413 oustanding OST_DESTROY requests is now throttled to
2414 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2415 callbacks are processed in priority.
2419 Description: enable ACLs on MDS by default
2420 Details : ACLs must be enabled on MDS by default.
2423 Frequency : PPC/PPC64 only
2425 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2426 Details : fix duplicated definitions between asm-ppc64/types.h and
2427 lustre_types.h on PPC.
2430 Frequency : PPC/PPC64 only
2432 Description: asm-ppc/segment.h does not exist
2433 Details : fix compile issue on PPC.
2437 Description: better handle error messages in extents code
2440 Frequency : RHEL4 only
2442 Description: mkfs is very slow on IA64/RHEL4
2443 Details : A performance regression has been discovered in the MPT Fusion
2444 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2445 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2446 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2448 Severity : enhancement
2450 Description: SNMP support enhancement
2451 Details : Adding total number of sampled request for an MDS node in snmp
2454 Severity : enhancement
2456 Description: Optimize ldlm waiting list processing for PR extent locks
2457 Details : When processing waiting list for read extent lock and meeting read
2458 lock that is same or wider to it that is not contended, skip
2459 processing rest of the list and immediatelly return current
2460 status of conflictness, since we are guaranteed there are no
2461 conflicting locks in the rest of the list.
2465 Description: Time out and refuse to reconnect
2466 Details : When the failover node is the primary node, it is possible
2467 to have two identical connections in imp_conn_list. We must
2468 compare not conn's pointers but NIDs, otherwise we can defeat
2469 connection throttling.
2473 Description: port llog fixes from b1_6 into HEAD
2474 Details : Port llog reference couting and some llog cleanups from b1_6
2475 (bug 10800) into HEAD, for protect from panic and access to already
2476 free llog structures.
2480 Description: Detect stride IO mode in read-ahead
2481 Details : When a client does stride read, read-ahead should detect that and
2482 read-ahead pages according to the detected stride pattern.
2486 Description: data checksumming impacts single node performance
2487 Details : add support for several checksum algorithm. Currently, only CRC32
2488 and Adler-32 are supported. The checksum type can be changed on
2489 the fly via /proc/fs/lustre/osc/*/checksum_type.
2493 Description: use adler32 for page checksums
2494 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2499 Description: build for x2 fails
2500 Details : fix compile issue on Cray systems.
2504 Description: Properly match for duplicate locks
2505 Details : Due to different lock order from skiplists code, we need to
2506 traverse entire list for now
2509 Frequency : only on PPC/SLES10
2511 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2512 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2513 be defined. Add a hack in mkfs_lustre.c to work around this compile
2518 Description: LASSERT on MDS when client holding flock lock dies
2519 Details : ldlm pool logic depends on number of granted locks equal to
2520 number of released locks which is not true for flock locks, so
2521 just exclude such locks from consideration.
2525 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2526 Details : Use fsfilt_sync() for both the whole filesystem sync and
2527 individual file sync to eliminate dangerous inode locking
2528 with I_LOCK that can lead to a deadlock.
2532 Description: performance in 1.6.3
2533 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2534 sure that 1MB requests are not fragmented by the block layer.
2536 Severity : enhancement
2538 Description: organize the server-side client stats on per-nid basis
2539 Details : Change the structure of stats under obdfilter and mds to
2547 The "uuid"s file would list the uuids of _active_ exports.
2548 And the clear entry is to clear all stats and stale nids.
2550 Severity : enhancement
2552 Description: eliminate client locks in face of contention
2553 Details : file contention detection and lockless i/o implementation
2554 for contended files.
2558 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2561 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2563 Description: MDS may lose file striping (and hence file data) in some cases
2564 Details : If there are additional extended attributes stored on the MDS,
2565 in particular ACLs, SELinux, or user attributes (if user_xattr
2566 is specified for the client mount options) then there is a risk
2567 of attribute loss. Additionally, the Lustre file striping
2568 needs to be larger than default (e.g. striped over all OSTs),
2569 and an additional attribute must be stored initially in the
2570 inode and then increase in size enough to be moved to the
2571 external attribute block (e.g. ACL growing in size) for file
2576 Description: skiplist implementation simplification
2577 Details : skiplists are used to group compatible locks on granted list
2578 that was implemented as tracking first and last lock of each
2579 lock group the patch changes that to using doubly linked lists
2583 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2584 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2585 because LASSERT on both the data supplied by a client, and the
2586 data on disk is dangerous and incorrect.
2588 Severity : enhancement
2590 Description: Slow truncate/writes to huge files at high offsets.
2591 Details : Directly associate cached pages to lock that protect those pages,
2592 this allows us to quickly find what pages to write and remove
2593 once lock callback is received.
2597 Description: more ldlm soft lockups
2598 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2599 starve other threads from the resource lock for a long time in
2600 case of long waiting queue, so change the debug level from
2601 D_OTHER to the less frequently used D_INFO.
2603 Severity : enhancement
2605 Description: add -gid, -group, -uid, -user options to lfs find
2609 Description: Hung threads in invalidate_inode_pages2_range
2610 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2611 once one is completed. As a result, some RPCs are stuck in the
2612 queue and are never sent.
2616 Description: filter threads hungs on waiting journal commit
2617 Details : Cleanup filter group llog code, then only filter group llog will
2618 be only created in the MDS/OST syncing process.
2622 Description: Procfs and llog threads access destoryed import sometimes.
2623 Details : Sync the import destoryed process with procfs and llog threads by
2624 the import refcount and semaphore.
2626 Severity : enhancement
2628 Description: openlock cache of b1_6 port to HEAD
2633 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2634 Details : If insertion of an extent fails, then discard the inode
2635 preallocation and free data blocks else it can lead to duplicate
2640 Description: don't always update ctime in ext3_xattr_set_handle()
2641 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2642 In some cases the ctime should not be updated, for example for
2643 2.0->1.8 compatibility it is necessary to delete an xattr and it
2644 should not update the ctime.
2648 Bugzilla : 15713/16362
2649 Description: Assertion in iopen_connect_dentry in 1.6.3
2650 Details : looking up an inode via iopen with the wrong generation number can
2651 populate the dcache with a disconneced dentry while the inode
2652 number is in the process of being reallocated. This causes an
2653 assertion failure in iopen since the inode's dentry list contains
2654 both a connected and disconnected dentry.
2658 Description: assertion failure in ldlm_handle2lock()
2659 Details : fix a race between class_handle_unhash() and class_handle2object()
2660 introduced in lustre 1.6.5 by bug 13622.
2665 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2666 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2671 Description: oops in sd_iostats_seq_show()
2672 Details : unloading/reloading the scsi low level driver triggers a kernel
2673 bug when trying to access the sd iostat file.
2678 Description: Kernel panics during QLogic driver reload
2679 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2680 patch, causing memory corruption.
2685 Description: journal_dev option does not work in b1_6
2686 Details : pass mount option during pre-mount.
2688 Severity : enhancement
2690 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2691 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2692 extent information of a file. It can be used to map logical blocks
2693 in a file to physical blocks in the block device.
2697 Description: LDLM soft lockups - improvement
2698 Details : It is be possible to send the lock handle along with each read
2699 or write request because the client is already doing a lock match
2700 itself so there isn't any reason the OST should have to re-do that
2704 Frequency : only on Cray X2
2706 Description: X2 build failures
2707 Details : fix build failures on Cray X2.
2711 Description: xid & resent requests
2712 Details : Initialize RPC XID from clock at startup (randomly if clock is
2715 Severity : enhancement
2717 Description: Add lustre_start utility to start or stop multiple Lustre servers
2722 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2723 Details : In case of memory pressure, list_del() can be called twice on
2724 req->rq_history_list, causing a kernel oops.
2728 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2729 Details : fix stack overflow in the distributed lock manager by defering
2730 export eviction after a failed AST to the elt thread instead of
2731 handling it in the dlm interpret routine.
2735 Description: Convert some comments to new format.
2736 Details : Update documenting comments to match doxygen conventions.
2740 Description: Grammar fixes.
2741 Details : A couple of trivial spelling fixes.
2745 Description: OSD_COUNTERS-mandatory
2746 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2747 to simplify the code.
2751 Description: simplify cmm_device freeing logic.
2752 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2756 Description: Add lockdep support to dt_object_operations locking interface.
2757 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2758 parameter indicating lock ordering. Update mdd code to use new
2763 Description: Introduce failloc constants for lockless IO tests.
2764 Details : Add two new failloc constants to test lockless IO. Only one of
2765 them in implemented---another is checked in yet to be landed
2770 Description: Add lockdep support for inode mutex.
2771 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2772 in the situations where Lustre has to lock more than one inode
2777 Description: Add optional invariants checking support.
2778 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2779 configure switch. This macro is to be used for consistency and
2780 sanity checks that are too expensive to be left in `production'
2785 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2786 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2791 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2792 Details : Use spin_lock_nested() in (the only) situation where more than
2793 one ldlm_lock is locked simultaneously. Also, fix possible
2794 dead-lock in ldlm_lock_change_resource() by enforcing particular
2799 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2800 Details : Instead of specifying each ldlm_lock call-back through separate
2801 parameter, wrap them into struct ldlm_callback_suite.
2805 Description: Kill join_lru obd method and its callers.
2806 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2807 in memory---a requirement implemented through ->o_join_lru() obd
2808 method in HEAD. Remove this method and its users.
2812 Description: Add asynchronous ldlm ENQUEUE completion handler.
2813 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2814 case that stock ldlm_completion_ast() cannot handle as it waits
2815 until lock is granted. Introduce new ldlm_completion_ast_async()
2816 for this. Also comment ldlm_completion_ast().
2820 Description: ldlm_error <-> errno conversion.
2821 Details : Add functions to map (rather arbitrary) between LDLM error codes
2822 and standard errno values. CLIO needs this to prevent LDLM specific
2823 constants from escaping ldlm and osc.
2827 Description: Kill unused ldlm_handle2lock_ns() function.
2828 Details : Kill unused ldlm_handle2lock_ns() function.
2832 Description: Add lu_ref support to ldlm_lock
2833 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2834 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2835 and ldlm_resource. LDLM interface has to be changed, because code that
2836 releases a reference on a lock, has to "know" what reference this is.
2837 In the most frequent case
2839 lock = ldlm_handle2lock(handle);
2841 LDLM_LOCK_PUT(lock);
2843 no changes are required. When any other reference (received _not_ from
2844 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2845 instead of LDLM_LOCK_PUT().
2847 Arguably, changes are pervasive, and interface requires some discipline
2848 for proper use. On the other hand, it was very instrumental in finding
2849 a few leaked lock references.
2853 Description: Add ldlm_lock_addref_try().
2854 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2855 attempts to addref a lock that might be being canceled
2860 Description: Add ldlm_weigh_callback().
2861 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2862 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2863 hard-coded `number of pages' logic.
2867 Description: Add lockdep annotations to llog code.
2868 Details : Use appropriately tagged _nested() locking calls in the places
2869 where llog takes more than one ->lgh_lock lock.
2873 Description: Add loi_kms_set().
2874 Details : Wrap kms updates into a helper function.
2878 Description: Constify instances of struct lsm_operations.
2879 Details : Constify instances of struct lsm_operations.
2883 Description: lu_conf support.
2884 Details : On a server, a file system object is uniquely identified
2885 by a fid, which is sufficient to locate and load all object
2886 state (inode). On a client, on the other hand, more data are
2887 necessary instantiate an object. Change lu_object_find() and
2888 friends to take additional `lu_conf' argument describing object.
2889 Typically this includes layout information.
2893 Description: lu_context fixes.
2894 Details : Introduce new lu_context functions that are needed on the client
2895 side, where some system threads (ptlrpcd) are shared by multiple
2896 modules, and so cannot be stopped during module shutdown.
2900 Description: Add start and stop methods to lu_device_type_operations.
2901 Details : Introduce two new methods in lu_device_type_operations, that are
2902 invoked when first instance of a given type is created and last one
2903 is destroyed respectively. This is need by CLIO.
2907 Description: Add lu_ref support to struct lu_device.
2908 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2909 track leaked references.
2913 Description: Introduce lu_kmem_descr.
2914 Details : lu_kmem_descr and its companion interface allow to create
2915 and destroy a number of kmem caches at once.
2919 Description: Fix lu_object finalization race.
2920 Details : Fix a race between lu_object_find() finding an object and its
2921 concurrent finalization. This race is (most likely) not possible
2922 on the server, but might happen on the client.
2926 Description: Introduce lu_ref interface.
2927 Details : lu_ref is a debugging module allowing to track references to
2928 a given object. It is quite cpu expensive, and has to be
2929 explicitly enabled with --enable-lu_ref. See usage description
2934 Description: Factor lu_site procfs stats into a separate function.
2935 Details : Separate lu_site stats printing code into a separate function
2936 that can be reused on a client.
2940 Description: Constify instances of struct {lu,dt,md}_device_operations.
2941 Details : Constify instances of struct {lu,dt,md}_device_operations.
2945 Description: Introduce struct md_site and move meta-data specific parts of
2946 struct lu_site here.
2947 Details : Move md-specific fields out of struct lu_site into special struct
2948 md_site, so that lu_site can be used on a client.
2952 Description: Kill mdd_lov_destroy().
2953 Details : Remove unused mdd code.
2957 Description: Add st_block checking to multistat.c.
2958 Details : Add st_block checking to multistat.c.
2962 Description: Add lu_ref support to struct obd_device.
2963 Details : Add lu_ref tracking to obd_device.
2967 Description: Kill obd_set_fail_loc().
2968 Details : Remove unused code.
2972 Description: Add special type for ptlrpc_request interpret functions.
2973 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2974 there. Actual usage will be in CLIO.
2978 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2979 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2980 doesn't work with lockdep.
2984 Description: Add rwv.c test program.
2985 Details : New testing program exercising readv(2) and writev(2) (Qian).
2989 Description: Add sendfile.c test program.
2990 Details : New testing program exercising sendfile(2) (Jay).
2994 Description: Ratelimit a message that can be very frequent.
2995 Details : Ratelimit a memory allocation failure message that can
3000 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3001 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3005 Description: Liblustre build fixes.
3006 Details : Liblustre build fixes.
3010 Description: libcfs: add cfs_{need,cond}_resched() interface.
3011 Details : libcfs: add cfs_{need,cond}_resched() definition and
3012 implementations for Linux, NT, and liblustre.
3014 Severity : enhancement
3016 Description: More exported tunables for mballoc
3017 Details : Add support for tunable preallocation window and new tunables for
3018 large/small requests
3022 Description: Detect corruption of block bitmap and checking for preallocations
3023 Details : Checks validity of on-disk block bitmap. Also it does better
3024 checking of number of applied preallocations. When corruption is
3025 found, it turns filesystem readonly to prevent further corruptions.
3029 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3030 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3031 window to 0. If it does hit cache miss, and read-pattern is still
3032 stride-io mode, does not reset the stride window, but also does not
3033 increase the stride window length in this case.
3037 Frequency : only for big-endian servers
3038 Description: Check if system is big-endian while mounting fs with extents feature
3039 Details : Mounting a filesystem with extents feature will fail on big-endian
3040 systems since ext3-based ldiskfs is not supported on big-endian
3041 systems. This can be over-riden with "bigendian_extents" mount option.
3043 Severity : enhancement
3045 Description: The root squash functionality
3046 Details : A security feature, which is to prevent users from being able
3047 to mount lustre on their desktop, run as root, and delete
3048 all of the files in the filesystem. The goal is accomplished by
3049 remapping user id (UID) and group id (GID) of the root user to
3050 a UID and GID specified by the system administartor via Lustre
3051 configuration management server (MGS). The functionality also
3052 allows to specify sets of clients for which the remapping does
3057 Description: Excessive recovery window
3058 Details : With AT enabled, the recovery window can be excessively long (6000+
3059 seconds). To address this problem, we no longer use
3060 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3061 timeout no longer depends on the service time, it is set to
3062 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3063 time via pb_service_time.
3067 Description: Watchdog triggered on MDS failover
3068 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3069 the OSTs know that the MDS "UUID" can be reused for the same export
3070 from a different NID, so we do not need to wait for the export to be
3074 Frequency : rare, only if using MMP with Linux RAID
3076 Description: MMP doesn't work with Linux RAID
3077 Details : While using HA for Lustre servers with Linux RAID, it is possible
3078 that MMP will not detect multiple mounts. To make this work we
3079 need to unplug the device queue in RAID when the MMP block is being
3080 written. Also while reading the MMP block, we should read it from
3081 disk and not the cached one.
3083 Severity : enhancement
3085 Description: open file using fid
3086 Details : A file can be opened using just its fid, like
3087 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3090 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3093 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3094 Details : add proper locking for creation of jbd_4k slab cache
3098 Description: MMP check in ext3_remount() fails without displaying any error
3099 Details : When multiple mount protection fails during remount, proper error
3102 Severity : enhancement
3104 Description: Allow stripe size to be up to 4G-64k
3105 Details : Fix math logic to allow large stripe sizes.
3109 Description: add check for >8TB ldiskfs filesystems
3110 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3111 Don't allow >8TB ldiskfs filesystems to be mounted without
3112 force_over_8tb mount option
3117 Descriptoin: OSS complains "too many blocks" and crashes
3118 Details : In filter_commitrw_write():
3119 1. lquota_chkquota();
3120 2. obdo_from_inode();
3121 3. lquota_pending_commit();
3122 "2." has changed uid and gid in this case, then caused
3123 lquota_chkquota() and lquota_pending_commit() to use
3124 different uid and gid. That is the root of the bug.
3128 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3129 Details : Detect and print out the 'stripe count = -1' case properly.
3133 Description: Out or order replies might be lost on replay
3134 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3135 request with tid smaller than one currently being added, add it
3136 to the start, not end of the list.
3138 --------------------------------------------------------------------------------
3140 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3142 * Support for kernels:
3143 2.6.5-7.283 (SLES 9),
3144 2.6.9-55.EL (RHEL 4),
3145 2.6.16.46-0.14 (SLES 10),
3146 2.6.18.8 vanilla (kernel.org)
3147 * Client support for unpatched kernels:
3148 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3149 2.6.16 - 2.6.22 vanilla (kernel.org)
3150 * Due to recently discovered recovery problems, we do not recommend
3151 using patchless RHEL 4 clients with this or any earlier release.
3152 * Recommended e2fsprogs version: 1.39.cfs8
3153 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3154 * Starting with this release, the ldiskfs backing filesystem required
3155 by Lustre is now in its own package, lustre-ldiskfs. This package
3156 should be installed. It is versioned separately from Lustre and
3157 may be released separately in future.
3161 Description: block reactivating mgc import until all deactivates complete
3162 Details : Fix race when failing back MDT/MGS to itself (testing)
3165 Frequency : at statup only
3167 Description: mds_lov_synchronize race leads to various problems
3168 Details : simultaneous MDT->OST connections at startup can cause the
3169 sync to abort, leaving the OSC in a bad state.
3171 Severity : enhancement
3173 Description: add optional extra BUILD_VERSION info
3174 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3175 to override the lustre version.
3178 Frequency : 2.6.18 servers only
3180 Description: ll_kern_mount() doesn't release the module reference
3181 Details : The ldiskfs module reference count never drops down to 0
3182 because ll_kern_mount() doesn't release the module reference.
3187 Description: server LBUG when using old ost_num_threads parameter
3188 Details : Accept the old ost_num_threads parameter but warn that it
3189 is deprecated, and fix an off-by-one error that caused an LBUG.
3194 Description: Transient SCSI error results in persistent IO issue
3195 Details : iobuf->dr_error is not reinitialized to 0 between two
3199 Frequency : sometimes when underlying device returns I/O errors
3201 Description: OSTs not going read-only during write failures
3202 Details : OSTs are not remounted read-only when the journal commit threads
3203 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3204 instead of the ext3 wrappers.
3208 Description: poor connect scaling with increasing client count
3209 Details : Don't run filter_grant_sanity_check for more than 100 exports
3210 to improve scaling for large numbers of clients.
3213 Frequency : SLES10 only
3215 Description: sanity-quota.sh quotacheck failed: rc = -22
3216 Details : Quotas cannot be enabled on SLES10.
3219 Frequency : liblustre clients only
3221 Description: getdirentries does not give error when run on compute nodes
3222 Details : getdirentries does not fail when the size specified as an argument
3223 is too small to contain at least one entry
3225 Severity : enhancement
3227 Description: Add LNET router traceability for debug purposes
3228 Details : If a checksum failure occurs with a router as part of the
3229 IO path, the NID of the last router that forwarded the bulk data
3230 is printed so it can be identified.
3235 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3236 Details : Due to a race condition, liblustre clients were occasionally
3237 evicted incorrectly.
3239 Severity : enhancement
3241 Description: lfs setstripe use optional parameters instead of postional
3244 Severity : enhancement
3246 Description: Nanosecond timestamp support for ldiskfs
3247 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3248 resolution timestamps. There is not yet support for this at
3249 the Lustre filesystem level.
3252 Frequency : during server recovery
3254 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3255 Details : request with rq_no_resend flag not awake l_wait_event if they get
3259 Frequency : nfs export on patchless client
3261 Description: connectathon hang when test nfs export over patchless client
3262 Details : Disconnected dentry cannot be found with lookup, so we do not need
3263 to unhash it or make it invalid
3266 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3267 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3268 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3273 Description: open req refcounting wrong on reconnect
3274 Details : If reconnect happened between getting open reply from server and
3275 call to mdc_set_replay_data in ll_file_open, we will schedule
3276 replay for unreferenced request that we are about to free.
3277 Subsequent close will crash in variety of ways.
3278 Check that request is still eligible for replay in
3279 mdc_set_replay_data().
3284 Description: disable writes to filesystem when reading health_check file
3285 Details : the default for reading the health_check proc file has changed
3286 to NOT do a journal transaction and write to disk, because this
3287 can cause reads of the /proc file to hang and block HA state
3288 checking on a healthy but otherwise heavily loaded system. It
3289 is possible to return to the previous behaviour during configure
3290 with --enable-health-write.
3292 Severity : enhancement
3294 Description: 64-bit inode version
3295 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3296 to the inode. This will be required for version-based recovery.
3301 Description: MDS fails to start if a duplicate client export is detected
3302 Details : in some rare cases it was possible for a client to connect to
3303 an MDS multiple times. Upon recovery the MDS would detect this
3304 and fail during startup. Handle this more gracefully.
3306 Severity : enhancement
3308 Description: Add -o localflock option to simulate old noflock
3310 Details : This will achieve local-only flock/fcntl locks
3316 Description: log_commit_thread vs filter_destroy race leads to crash
3317 Details : Take import reference before releasing llog record semaphore
3322 Description: Wrong request locking in request set processing
3323 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3324 imp_delayed_list, in this place should be used imp_lock.
3327 Frequency : when reconnection
3329 Description: Grant Leak when osc reconnect to OST
3330 Details : When osc reconnect ost, OST(filter) should check whether it
3331 should grant more space to client by comparing fed_grant and
3332 cl_avail_grant, and return the granted space to client instead
3333 of "new granted" space, because client will call osc_init_grant
3334 to update the client grant space info.
3337 Frequency : when client reconnect to OST
3339 Description: Grant Leak when osc do resend and replay bulk write
3340 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3341 bulk write request, because the grant info while be sync between
3342 OSC and OST when reconnect, and we should ignore the grant info
3343 these of resend/replay write req.
3348 Description: Grant space more than available left space sometimes.
3349 Details : When then OST is about to be full, if two bulk writing from
3350 different clients came to OST. Accord the avaliable space of the
3351 OST, the first req should be permitted, and the second one
3352 should be denied by ENOSPC. But if the seconde arrived before
3353 the first one is committed. The OST might wrongly permit second
3354 writing, which will cause grant space > available space.
3357 Frequency : when client is evicted
3359 Description: Grant might be wrongly erased when osc is evicted by OST
3360 Details : when the import is evicted by server, it will fork another
3361 thread ptlrpc_invalidate_import_thread to invalidate the
3362 import, where the grant will be set to 0. While the original
3363 thread will update the grant it got when connecting. So if
3364 the former happened latter, the grant will be wrongly errased
3365 because of this race.
3370 Description: Checking Stale with correct fid
3371 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3372 to check whether it is stale, because sometimes, we want the
3373 enqueue happened anyway, and op_data.fid2 will not be initialized.
3375 Severity : enhancement
3377 Description: update patchless client
3378 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3381 Frequency : only with 2.4 kernel
3383 Description: random memory corruption
3384 Details : size of struct ll_inode_info is to big for union inode.u and this
3385 can be cause of random memory corruption.
3390 Description: Memory leak in recovery
3391 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3392 It should also check obd_fail before fsfilt_start, otherwise if
3393 fsfilt_start return -EROFS,(failover mds during mds recovery).
3394 then the req will return with repmsg->transno = 0 and rc = EROFS.
3395 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3396 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3397 be freed no matter whether fsfilt_commit success or not.
3400 Frequency : only with huge count clients
3402 Description: Prevents from taking the superblock lock in llap_from_page for
3404 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3405 for llap_from_page call in ll_removepage prevents from taking the
3406 superblock lock for a soon died page.
3411 Description: Not check open intent error before release open handle
3412 Details : in some rare cases, the open intent error is not checked before
3413 release open handle, which may cause
3414 ASSERTION(open_req->rq_transno != 0), because it tries to release
3415 the failed open handle.
3420 Description: Set cat log bitmap only after create log success.
3421 Details : in some rare cases, the cat log bitmap is set too early. and it
3422 should be set only after create log success.
3426 Description: Accessing a block bevice can re-enable I/O when Lustre is
3427 tearing down a device.
3428 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3433 Description: service threads may hog cpus when there are a lot of requests
3434 Details : Insert cond_resched to give other threads a chance to use some CPU
3439 Description: the cat log was not initialized in recovery
3440 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3441 unlink log on mds will not be initialized until mds post
3442 recovery. And also in mds post recovery, the unlink log will
3443 initialization will be done asynchronausly, so there will be race
3444 between add unlink log and unlink log initialization.
3448 Description: brw_stats were being printed incorrectly
3449 Details : brw_stats were being printed as log2 but all of them were not
3450 recorded as log2. Also remove some code duplication arising from
3451 filter_tally_{read,write}.
3455 Frequency : rare, only in recovery.
3456 Description: ASSERTION(req->rq_type != LI_POISON) failed
3457 Details : imp_lock should be held while iterating over imp_sending_list for
3458 prevent destroy request after get timeout in ptlrpc_queue_wait.
3462 Description: replay-single.sh test 52 fails
3463 Details : A lock's skiplist need to be cleanup when it being unlinked
3464 from its resource list.
3468 Description: Short directio read returns full requested size rather than
3470 Details : Direct I/O operations should return actual amount of bytes
3471 transferred rather than requested size.
3473 Severity : enhancement
3475 Description: metadata RPC reduction (e.g. for rm performance)
3476 Details : decrease the amount of synchronous RPC between clients and servers
3477 by canceling conflicing lock before the operation on the client
3478 and packing thier handles into the main operation RPC to server.
3480 Severity : enhancement
3482 Description: Async OSC create to avoid the blocking unnecessarily.
3483 Details : If an OST has no remaining object, system will block on the
3484 creation when it needs to create a new object on this OST. Now,
3485 ways use pre-created objects when available, instead of blocking on
3486 an empty osc while others are not empty. If we must block, we
3487 block for the shortest possible period of time.
3491 Description: improve handling recoverable errors
3492 Details : if request processig with error which can be recoverable on server
3493 request should be resend, otherwise page released from cache and
3496 Severity : enhancement
3498 Description: refine locking for avoid write wrong info into lov_objid file
3499 Details : fix possible races with add new target and write/update data in
3502 --------------------------------------------------------------------------------
3504 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3509 Frequency : on some architectures
3511 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3512 Details : Uninitialized flags sometimes cause configuration commands to
3516 Frequency : patchless clients only
3518 Description: missing __iget() symbol export
3519 Details : The __iget() symbol export is missing. To avoid the need for
3520 this on patchless clients the deathrow inode reaper is turned
3521 off, and we depend on the VM to clean up old inodes. This
3522 dependency was during via the fix for bug 12181.
3524 --------------------------------------------------------------------------------
3526 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3528 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3529 INTEROPERATE with older versions automatically. In many cases a
3530 special upgrade step is needed. Please read the
3531 user documentation before upgrading any part of a live system.
3532 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3533 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3534 * WARNING: Lustre configuration and startup changes are required with
3535 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3539 Severity : enhancement
3541 Description: MountConf
3542 Details : Lustre configuration is now managed via mkfs and mount
3543 commands instead of lmc and lconf. New obd types (MGS, MGC)
3544 are added for dynamic configuration management. See
3545 https://mail.clusterfs.com/wikis/lustre/MountConf for
3548 Severity : enhancement
3550 Description: dynamic OST addition
3551 Details : OSTs can now be added to a live filesystem
3553 Severity : enhancement
3555 Description: startup order invariance
3556 Details : MDTs and OSTs can be started in any order. Clients only
3557 require the MDT to complete startup.
3559 Severity : enhancement
3561 Description: parallel, asynchronous orphan cleanup
3562 Details : orphan cleanup is now performed in separate threads for each
3563 OST, allowing parallel non-blocking operation.
3565 Severity : enhancement
3567 Description: optimized stripe assignment
3568 Details : stripe assignments are now made based on ost space available,
3569 ost previous usage, and OSS previous usage, in order to try
3570 to optimize storage space and networking resources.
3572 Severity : enhancement
3574 Description: Permanently set tunables
3575 Details : All writable /proc/fs/lustre tunables can now be permanently
3576 set on a per-server basis, at mkfs time or on a live system.
3578 Severity : enhancement
3580 Description: Lustre message v2
3581 Details : Add lustre message format v2.
3583 Severity : enhancement
3585 Description: client OST exclusion list
3586 Details : Clients can be started with a list of OSTs that should be
3587 declared "inactive" for known non-responsive OSTs.
3591 Description: SPEC SFS validation failure on NFS v2 over lustre.
3592 Details : Changes the blocksize for regular files to be 2x RPC size,
3593 and not depend on stripe size.
3595 Severity : enhancement
3597 Description: Multiple MD RPCs in flight.
3598 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3599 To avoid overly-overloading MDS, introduce a limit on number of
3600 MDS RPCs in flight for a single client and add /proc controls
3601 to adjust this limit.
3603 Severity : enhancement
3605 Description: client read/write statistics
3606 Details : Add client read/write call usage stats for performance
3607 analysis of user processes.
3608 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3609 file access. extents_stats shows chunk size distribution.
3610 extents_stats_per_process show chunk size distribution per
3613 Severity : enhancement
3615 Description: mds statistics
3616 Details : Add detailed mds operations statistics in
3617 /proc/fs/lustre/mds/*/stats.
3621 Description: Failure of copying files with lustre special EAs.
3622 Details : Client side always return success for setxattr call for lustre
3623 special xattr (currently only "trusted.lov").
3628 Description: Refcount LNET uuids
3629 Details : The global LNET uuid list grew linearly with every startup;
3630 refcount repeated list entries instead of always adding to
3633 Severity : enhancement
3635 Description: Dynamic service threads
3636 Details : Within a small range, start extra service threads
3637 automatically when the request queue builds up.
3640 Frequency : mixed-endian client/server environments
3642 Description: mixed-endian crashes
3643 Details : The new msg_v2 system had some failures in mixed-endian
3646 Severity : enhancement
3648 Description: Easy OST removal
3649 Details : OSTs can be permanently deactivated with e.g. 'lctl
3650 conf_param lustre-OST0001.osc.active=0'
3652 Severity : enhancement
3654 Description: MGS proc entries
3655 Details : Added basic proc entries for the MGS showing what filesystems
3658 Severity : enhancement
3660 Description: provide MGS failover
3661 Details : Added config lock reacquisition after MGS server failover.
3663 Severity : enhancement
3665 Description: add Linux 2.4 support
3666 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3670 Description: a large application tries to do I/O to the same resource and dies
3671 in the middle of it.
3672 Details : Check the req->rq_arrival time after the call to
3673 ost_brw_lock_get(), but before we do anything about
3674 processing it & sending the BULK transfer request. This
3675 should help move old stale pending locks off the queue as
3676 quickly as obd_timeout.
3679 Frequency : when an incorrect nid is specified during startup
3681 Description: ptlrpc connect to non-existant node causes kernel crash
3682 Details : LNET can't be re-entered from an event callback, which
3683 happened when we expire a message after the export has been
3684 cleaned up. Instead, hand the zombie cleanup off to another
3687 Severity : enhancement
3689 Description: plain/inodebits lock performance improvement
3690 Details : Grouping plain/inodebits in granted list by their request modes
3691 and bits policy, thus improving the performance of search through
3695 Frequency : only if OST filesystem is corrupted
3697 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3698 Details : for a short time RPCs with bulk IO are in the replay list,
3699 but replay of bulk IOs is unimplemented. If the OST filesystem
3700 is corrupted due to disk cache incoherency and then replay is
3701 started it is possible to trip an assertion. Avoid putting
3702 committed RPCs into the replay list at all to avoid this issue.
3705 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3707 Description: System hang on startup
3708 Details : This bug allowed the liblustre (e.g. catamount) client to
3709 return to the app before handling all startup RPCs. This
3710 could leave the node unresponsive to lustre network traffic
3711 and manifested as a server ptllnd timeout.
3713 Severity : enhancement
3715 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3716 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3717 boolean to control whether to print peer debug info when a
3718 client's RPC times out.
3721 Frequency : only for kernels with patches from Lustre below 1.4.3
3723 Description: Remove old rdonly API
3724 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3727 Frequency : only for devices with external journals
3729 Description: Set external device read-only also
3730 Details : During a commanded failover stop, we set the disk device
3731 read-only while the server shuts down. We now also set any
3732 external journal device read-only at the same time.
3735 Frequency : when upgrading from 1.4 while trying to change parameters
3737 Description: The wrong (new) MDC name was used when setting parameters for
3738 upgraded MDT's. Also allows changing of OSC (and MDC)
3739 parameters if --writeconf is specified at tunefs upgrade time.
3742 Frequency : when setting specific ost indicies
3744 Description: QOS code breaks on skipped indicies
3745 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3746 created with --index need not be sequential.
3748 Severity : enhancement
3750 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3751 Details : The uninit_groups feature works in conjunction with the kernel
3752 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3753 up the pass1 processing of e2fsck. This is a read-only feature
3754 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3755 mount filesystems that have had this feature enabled.
3757 Severity : enhancement
3759 Description: Improve multi-block allocation algorithm to avoid fragmentation
3760 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3761 allocation locality and avoid filesystem fragmentation.
3763 ------------------------------------------------------------------------------
3765 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3767 * Support for kernels:
3768 2.4.21-47.0.1.EL (RHEL 3)
3769 2.6.5-7.283 (SLES 9)
3770 2.6.9-42.0.10.EL (RHEL 4)
3771 2.6.12.6 vanilla (kernel.org)
3772 2.6.16.27-0.9 (SLES 10)
3773 * Recommended e2fsprogs version: 1.39.cfs5
3775 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3779 Frequency : occasional, depends on client load and configuration
3780 Bugzilla : 12181, 12203
3781 Description: data loss for recently-modified files
3783 Details : In some cases it is possible that recently written or created
3784 files may not be written to disk in a timely manner (this should
3785 normally be within 30s unless client IO load is very high).
3786 The problem appears as zero-length files or files that are a
3787 multiple of 1MB in size after a client crash or client eviction
3788 that are missing data at the end of the file.
3790 This problem is more likely to be hit on clients where files are
3791 repeatedly created and unlinked in the same directory, clients
3792 have a large amount of RAM, have many CPUs, the filesystem has
3793 many OSTs, the clients are rebooted frequently, and/or the files
3794 are not accessed by other nodes after being written.
3796 The presence of the problem can be detected by looking at
3797 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3798 smaller than the second (nr_unused) then dirty files will not
3799 be flushed automatically to disk. "sync; sleep 10" should be
3800 run several times on the node before unmounting it to update
3801 Lustre (this is also safe to run on nodes without this problem).
3803 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3804 that can cause this same problem, so customers using that kernel
3805 also need to update the kernel in addition to Lustre. In order
3806 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3808 It is normal that files written just before a client crash (less
3809 than 30s) may not yet have been flushed to disk, even for local
3813 Frequency : frequent on thin XT3 nodes
3815 Description: UUID collision on thin XT3 Linux nodes
3816 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3817 randomly, since we relied on an insufficiently-seeded PRNG.
3822 Description: OSS hangs after "All ost request buffers busy"
3823 Details : A deadlock between quota and journal operations caused OSS
3824 hangs after printing "All ost request buffers busy."
3827 Frequency : always on liblustre builds
3829 Description: Cleanup compiler warnings on liblustre
3832 Frequency : always on liblustre builds on XT3
3834 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3838 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3839 > 32000 subdirectories
3840 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3841 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3842 subdirectory count crosses 32000. This will aid e2fsck to
3843 correctly handle more than 32000 subdirectories.
3846 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3848 Description: System hang on startup
3849 Details : This bug allowed the liblustre (e.g. catamount) client to
3850 return to the app before handling all startup RPCs. This
3851 could leave the node unresponsive to lustre network traffic
3852 and manifested as a server ptllnd timeout.
3854 Severity : enhancement
3856 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3857 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3858 boolean to control whether to print peer debug info when a
3859 client's RPC times out.
3864 Description: make O_SYNC working on 2.6 kernels
3865 Details : 2.6 kernels use different method for mark pages for write,
3866 so need add a code to lustre for O_SYNC work.
3871 Description: Failure to close file and release space on NFS
3872 Details : Put inode details into lock acquired in ll_intent_file_open.
3873 Use mdc_intent_lock in ll_intent_open to properly
3874 detect all kind of errors unhandled by mdc_enqueue.
3879 Description: proc file read during shutdown sometimes raced obd removal,
3881 Details : Add lock to prevent obd access after proc file removal.
3884 Frequency : Only for files larger than 4GB on 32-bit clients.
3886 Description: improperly doing page alignment of locks
3887 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3888 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3893 Description: RPCs being resent when they shouldn't be
3894 Details : Some RPCs that should not be resent are being resent. This
3895 can cause inconsistencies in the RPC state machine. Do not
3896 resend such requests.
3899 Frequency : rare, only with NFS export
3901 Description: Crash on NFS re-export node
3902 Details : under very unusual load conditions an assertion is hit in
3903 ll_intent_file_open()
3906 Frequency : only if OST filesystem is corrupted
3908 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3909 Details : for a short time RPCs with bulk IO are in the replay list,
3910 but replay of bulk IOs is unimplemented. If the OST filesystem
3911 is corrupted due to disk cache incoherency and then replay is
3912 started it is possible to trip an assertion. Avoid putting
3913 committed RPCs into the replay list at all to avoid this issue.
3918 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3919 Details : Large single O_DIRECT read and write calls can fail to allocate
3920 a sufficiently large buffer to process the request. In case of
3921 allocation failure the allocation is retried with a smaller
3922 buffer and broken into smaller requests.
3924 Severity : enhancement
3926 Description: Add -o localflock option to simulate old noflock behaviour.
3927 Details : This will achieve local-only flock/fcntl locks coherentness.
3932 Description: versioning check is incomplete
3933 Details : Checking the version difference of client vs. server, report
3934 error if the gap is too big.
3939 Description: add support PG_writeback bit
3940 Details : add support for PG_writeback bit for Lustre, for more carefull
3941 work with page cache in 2.6 kernel. This also fix some deadlocks
3942 and remove hack for work O_SYNC with 2.6 kernel.
3944 Severity : enhancement
3946 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3947 Details : The uninit_groups feature works in conjunction with the kernel
3948 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3949 up the pass1 processing of e2fsck. This is a read-only feature
3950 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3951 mount filesystems that have had this feature enabled.
3953 Severity : enhancement
3955 Description: Improve multi-block allocation algorithm to avoid fragmentation
3956 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3957 allocation locality and avoid filesystem fragmentation.
3959 ------------------------------------------------------------------------------
3961 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3963 * Support for kernels:
3964 2.6.9-42.0.3.EL (RHEL 4)
3965 2.6.5-7.276 (SLES 9)
3966 2.4.21-47.0.1.EL (RHEL 3)
3967 2.6.12.6 vanilla (kernel.org)
3968 2.6.16.21-0.8 (SLES10)
3969 * Recommended e2fsprogs version: 1.39.cfs2-0
3971 * The backwards-compatible /proc/sys/portals symlink has been removed
3972 in this release. Before upgrading, please ensure that you change
3973 any configuration scripts or /etc/sysctl.conf files that access
3974 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3975 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3976 in advance of the upgrade on any system running Lustre 1.4.6 or
3977 newer, since /proc/sys/lnet was added in that version.
3978 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3982 Frequency : only when quota is used
3984 Description: avoid scanning export list for quota master
3985 Details : Change the algorithms to avoid scanning export list in order
3986 to improve the efficiency.
3989 Frequency : MDS failover only, very rarely
3991 Description: "went back in time" messages on mds failover
3992 Details : The greatest transno may be lost when the current operation
3993 finishes with an error (transno==0) and the client's last_rcvd
3994 record is over-written. Save the greatest transno in the
3995 mds_last_transno for this case.
3998 Frequency : always for specific kernels and striping counts
4000 Description: client may get "Matching packet too big" without ACL support
4001 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4002 when trying to access files in certain configurations. The
4003 clients should in fact be denied when mounting because they do
4004 not understand ACLs.
4007 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4009 Description: many clients connecting with IO in progress causes connect timeouts
4010 Details : Avoid synchronous journal commits to avoid delays caused by many
4011 clients connecting/disconnecting when bulk IO is in progress.
4012 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4013 OST_IO_PORTAL to avoid delays behind potentially many pending
4017 Frequency : occasionally with multiple writers to a single file
4019 Description: shared writes to file may result in wrong size reported by stat()
4020 Details : Allow growing of kms when extent lock is cancelled
4023 Frequency : always with random mmap IO to multi-striped file
4025 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4026 Details : Only if the hole is at the end of OST object so that kms is too
4027 small. Fix is to increase kms accordingly in ll_nopage.
4030 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4032 Description: writes to a missing object would leak memory on the OST
4033 Details : If there is an inconsistency between the MDS and OST filesystems,
4034 such that the MDS references an object that doesn't exist, writes
4035 to that object will leak memory due to incorrect cleanup in the
4036 error handling path, eventually running out of memory on the OST.
4041 Description: Creating too long symlink causes lustre errors
4042 Details : Check symlink and name lengths before sending requests to MDS.
4045 Frequency : only if flock is enabled (not on by default)
4047 Description: posix locks not released on fd closure on 2.6.9+
4048 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4049 kernels, this caused such locks not to be released on fd close and
4050 then assertions on fs unmount about still used locks.
4053 Frequency : MDS failover only, very rarely
4055 Description: clients may get ASSERTION(granted_lock != NULL)
4056 Details : When request was taking a long time, and a client was resending
4057 a getattr by name lock request. The were multiple lock requests
4058 with the same client lock handle and
4059 mds_getattr_name->fixup_handle_for_resent_request found one of the
4060 lock handles but later failed with ASSERTION(granted_lock != NULL).
4065 Description: handle->h_buffer_credits > 0, assertion failure
4066 Details : h_buffer_credits is zero after truncate, causing assertion
4067 failure. This patch extends the transaction or creates a new
4071 Frequency : NFS re-export or patchless client
4072 Bugzilla : 11179, 10796
4073 Description: Crash on NFS re-export node (__d_move)
4074 Details : We do not want to hash the dentry if we don't have a lock.
4075 But if this dentry is later used in d_move, we'd hit uninitialised
4076 list head d_hash, so we just do this to init d_hash field but
4077 leave dentry unhashed.
4080 Frequency : NFS re-export or patchless client
4082 Description: NFS exports has problem with symbolic link
4083 Details : lustre client didn't properly install dentry when re-exported
4084 to NFS or running patchless client.
4087 Frequency : NFS re-export or patchless client
4089 Description: Various nfs/patchless fixes.
4090 Details : fixes reuse disconected alias for lookup process - this fixes
4091 warning "find_exported_dentry: npd != pd",
4092 fix permission error with open files at nfs.
4093 fix apply umask when do revalidate.
4096 Frequency : occasional
4098 Description: Crash on NFS re-export node
4099 Details : calling clear_page() on the wrong pointer triggered oops in
4100 generic_mapping_read().
4103 Frequency : rarely, using O_DIRECT IO
4105 Description: unaligned directio crashes client with LASSERT
4106 Details : check for unaligned buffers before trying any requests.
4109 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4111 Description: stale data returned from RAID cache
4112 Details : If only a small amount of IO is done to the RAID device before
4113 reading it again it is possible to get stale data from the RAID
4114 cache instead of reading it from disk.
4117 Frequency : always for sles10 kernel
4119 Description: sles10 support
4120 Details : ll_follow_link: compile fixes and using of nd_set_link
4121 under newer kernels.
4124 Frequency : depends on arch, kernel and compiler version, always on sles10
4127 Description: recursive or deep enough symlinks cause stack overflow
4128 Details : getting rid of large stack-allocated variable in
4132 Frequency : depends on hardware
4134 Description: lustre write performance loss in the SLES10 kernel
4135 Details : the performance loss is caused by using of write barriers in the
4136 ext3 code. The SLES10 kernel turns barrier support on by
4137 default. The fix is to undo that change for ldiskfs.
4139 ------------------------------------------------------------------------------
4141 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4143 * Support for kernels:
4144 2.6.9-42.0.3EL (RHEL 4)
4145 2.6.5-7.276 (SLES 9)
4146 2.4.21-47.0.1.EL (RHEL 3)
4147 2.6.12.6 vanilla (kernel.org)
4151 Frequency : quota enabled and large files being deleted
4153 Description: releasing more than 4GB of quota at once hangs OST
4154 Details : If a user deletes more than 4GB of files on a single OST it
4155 will cause the OST to spin in an infinite loop. Release
4156 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4161 Description: statfs data retrieved from /proc may be stale or zero
4162 Details : When reading per-device statfs data from /proc, in the
4163 {kbytes,files}_{total,free,avail} files, it may appear
4164 as zero or be out of date.
4167 Frequency : systems with MD RAID1 external journal devices
4169 Description: lconf's call to blkid is confused by RAID1 journal devices
4170 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4171 of returning all block devices that match the journal UUID.
4174 Frequency : always, for aggregate stripe size over 4GB
4176 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4177 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4178 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4181 Frequency : always if "lfs find" used on a local file/directory
4183 Description: "lfs find" segfaults if used on a local file/directory
4184 Details : The case where a directory component was not specified wasn't
4185 handled correctly. Handle this properly.
4188 Frequency : always on ppc64
4190 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4191 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4192 perform endianness conversion.
4195 Frequency : rarely (truncate to non-zero file size after write under load)
4196 Bugzilla : 10730, 10687
4197 Description: Files padded with zeros to next 4K multiple
4198 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4199 it is possible that files that are truncated to a non-zero size
4200 immediately after being written are filled with zero bytes beyond
4201 the truncated size. No file data is lost.
4203 Severity : enhancement
4205 Description: Allow recovery/failover for liblustre clients.
4206 Details : liblustre clients were unaware of failover configurations until
4209 Severity : enhancement
4211 Description: user file locks should fail when not mounting with flock option
4212 Details : Set up an error-returning stub in ll_file_operations.lock field
4213 to prevent incorrect behaviour when client is mounted without
4214 flock option. Also, set up properly f_op->flock field for
4218 Frequency : always on ia64
4220 Description: "lfs df" loops on printing out MDS statfs information
4221 Details : The obd_ioctl_data was not initialized and in some systems
4222 this caused a failure during the ioctl that did not return
4223 an error. Initialize the struct and return an error on failure.
4226 Frequency : SLES 9 only
4228 Description: Error of copying files with lustre special EAs as root
4229 Details : Client side always return success for setxattr call for lustre
4230 special xattr (currently only "trusted.lov").
4233 Frequency : rarely on clusters with both ia64+i386 clients
4235 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4236 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4237 clients are doing concurrent writes to the same file it is
4238 possible that the ia64 clients may LASSERT because the OST
4239 extent locks are not PAGE_SIZE aligned. Ensure that grown
4240 locks are always aligned on the request boundary.
4243 Frequency : specific use, occasional
4245 Description: Overwriting in use executable truncates on-disk binary image
4246 Details : If one node attempts to overwrite an executable in use by
4247 another node, we now correctly return ETXTBSY instead of
4248 truncating the file.
4250 Severity : enhancement
4252 Description: Async OSC create to avoid the blocking unnecessarily.
4253 Details : If an OST has no remaining object, system will block on the creating
4254 when need to create a new object on this OST. Now, ways use
4255 pre-created objects when available, instead of blocking on an
4256 empty osc while others are not empty. If we must block, we block
4257 for the shortest possible period of time.
4262 Description: chmod on Lustre root is propagated to other clients
4263 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4264 invalid by the follow_mount time.
4267 Frequency : liblustre clients only
4269 Description: Race in 'instant cancel' lock handling could lead to such locks
4270 never to be granted in case of SMP MDS
4271 Details : Do not destroy not yet granted but cbpending locks in
4275 Frequency : replay/resend of open
4277 Description: non null lock assetion failure in mds_intent_policy
4278 Details : Trying to replay/resend lockless open requests resulted in
4279 mds_open() returning 0 with no lock. Now it sets a flag if
4280 a lock is going to be returned.
4282 Severity : enhancement
4284 Description: Checksum enhancements
4285 Details : New checksum enhancements allow for resending RPCs that failed
4288 Severity : enhancement
4290 Description: Tunables on number of dirty pages in cacche
4291 Details : Allow to set limit on number of dirty pages cached.
4296 Description: client crash on unmount - lock still has references
4297 Details : In some error handling cases it was possible to leak a lock
4298 reference on a client while accessing a file. This was not
4299 harmful to the client during operation, but would cause the
4300 client to crash when the filesystem is unmounted.
4303 Frequency : specific case, rare
4305 Description: ETXTBSY on mds though file not in use by client
4306 Details : ETXTBSY is no longer incorrectly returned when attempting to
4307 chmod or chown a directory that the user previously tried to
4308 execute or a currently-executing binary.
4311 Frequency : extremely rare except on liblustre-based clients
4313 Description: Lustre space not freed when files are deleted
4314 Details : Clean up open-unlinked files after client eviction. Previously
4315 the unlink was skipped and the files remained as orphans.
4320 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4321 Details : In some cases it was possible to send a blocking callback to
4322 a client doing a glimpse, even though that client didn't get
4323 a lock granted. When the glimpse lock is cancelled on the OST
4324 the freed lock is left on the waiting list and corrupted the list.
4327 Frequency : all core dumps
4329 Description: Broke core dumps to lustre
4330 Details : Negative dentry may be unhashed if parent does not have UPDATE
4331 lock, but some callers, e.g. do_coredump, expect dentry to be
4332 hashed after successful create, hash it in ll_create_it.
4334 ------------------------------------------------------------------------------
4336 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4338 * Support for kernels:
4339 2.6.9-42.0.2.EL (RHEL 4)
4340 2.6.5-7.276 (SLES 9)
4341 2.4.21-40.EL (RHEL 3)
4342 2.6.12.6 vanilla (kernel.org)
4346 Frequency : always on RHEL 3
4348 Description: Number of open files grows over time
4349 Details : The number of open files grows over time, whether or not
4350 Lustre is started. This was due to a filp leak introduced
4351 by one of our kernel patches.
4353 ------------------------------------------------------------------------------
4355 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4357 * Support for kernels:
4358 2.6.9-42.EL (RHEL 4)
4359 2.6.5-7.276 (SLES 9)
4360 2.4.21-40.EL (RHEL 3)
4361 2.6.12.6 vanilla (kernel.org)
4366 Bugzilla : 5719, 9635, 9792, 9684
4367 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4368 Details : If a server is under heavy load and cannot reply to new
4369 connection requests before the client resends the (re)connect,
4370 the connection handling code can behave badly if two service
4371 threads are concurrently handing separate (re)connections from
4372 the same client. Add better locking to the connection handling
4373 code, and ensure that only a single connection will be processed
4374 for a given client UUID, even if the lock is dropped.
4376 Severity : enhancement
4378 Description: add TCP zero-copy support to kernel
4379 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4380 sends if the hardware supports scatter-gather and checksumming.
4381 This allows socklnd to do client-write and server-read more
4382 efficiently and reduce CPU utilization from skbuf copying.
4385 Frequency : only if NFS exporting from client
4387 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4388 Details : If setting attributes on a file created under NFS that had
4389 never been opened it would be possible to oops the client
4390 if the file had no objects.
4393 Frequency : always for liblustre
4395 Description: liblustre client does MDS+OSTs setattr RPC for each write
4396 Details : When doing a write from a liblustre client, the client
4397 incorrectly issued an RPC to the MDS and each OST the file was
4398 striped over in order to update the timestamps. When writing
4399 with small chunks and many clients this could overwhelm the MDS
4400 with RPCs. In all cases it would slow down the write because
4401 these RPCs are unnecessary.
4403 Severity : enhancement
4405 Description: allow number of MDS service threads to be changed at module load
4406 Details : It is now possible to change the number of MDS service threads
4407 running. Adding "options mds mds_num_threads={N}" to the MDS's
4408 /etc/modprobe.conf will set the number of threads for the next
4409 time Lustre is restarted (assuming the "mds" module is also
4410 reloaded at that time). The default number of threads will
4411 stay the same, 32 for most systems.
4416 Description: OST crash if filesystem is unformatted or corrupt
4417 Details : If an OST is started on a device that has never been formatted
4418 or if the filesystem is corrupt and cannot even mount then the
4419 error handling cleanup routines would dereference a NULL pointer.
4424 Description: NULL pointer deref in llap_from_page.
4425 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4426 incorrect) mapping if the page was truncated/reclaimed while it was
4427 searched for. Check for this condition and skip such pages when
4428 doing readahead. Introduce extra check to llap_from_page() to
4429 verify page->mapping->host is non-NULL (so page is not anonymous).
4432 Frequency : Sometimes when using sys_sendfile
4434 Description: "page not covered by a lock" warnings from ll_readpage
4435 Details : sendfile called ll_readpage without right page locks present.
4436 Now we introduced ll_file_sendfile that does necessary locking
4437 around call to generic_file_sendfile() much like we do in
4441 Frequency : with certain MDS communication failures at client mount time
4443 Description: NULL pointer deref after failed client mount
4444 Details : a client connection request may delayed by the network layer
4445 and not be sent until after the PTLRPC layer has timed out the
4446 request. If the client fails the mount immediately it will try
4447 to clean up before the network times out the request. Add a
4448 reference from the request import to the obd device and delay
4449 the cleanup until the network drops the request.
4452 Frequency : occasionally during client (re)connect
4454 Description: assertion failure during client (re)connect
4455 Details : processing a client connection request may be delayed by the
4456 client or server longer than the client connect timeout. This
4457 causes the client to resend the connection request. If the
4458 original connection request is replied in this interval, the
4459 client may trip an assertion failure in ptlrpc_connect_interpret()
4460 which thought it would be the only running connect process.
4463 Frequency : only with obd_echo servers and clients that are rebooted
4465 Description: kernel BUG accessing uninitialized data structure
4466 Details : When running an obd_echo server it did not start the ping_evictor
4467 thread, and when a client was evicted an uninitialized data
4468 structure was accessed. Start the ping_evictor in the RPC
4469 service startup instead of the OBD startup.
4471 Severity : enhancement
4472 Bugzilla : 10193 (patchless)
4473 Description: Remove dependency on various unexported kernel interfaces.
4474 Details : No longer need reparent_to_init, exit_mm, exit_files,
4475 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4478 Frequency : rare (only users of deprecated and unsupported LDAP config)
4480 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4481 Details : LDAP apparently contains 'lustreName' attributes instead of
4482 'name'. A simple remapping of the name is sufficient.
4485 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4487 Description: waiting_locks_callback trips kernel BUG if client is evicted
4488 Details : Running with the dump_on_timeout debug flag turned on makes
4489 it possible that the waiting_locks_callback() can try to dump
4490 the Lustre kernel debug logs from an interrupt handler. Defer
4491 this log dumping to the expired_lock_main() thread.
4493 Severity : enhancement
4495 Description: Support NFS exporting on 2.6 kernels.
4496 Details : Implement non-rawops metadata methods for NFS server to use without
4497 changing NFS server code.
4500 Frequency : very rare (synthetic metadata workload only)
4502 Description: two racing renames might cause an MDS thread to deadlock
4503 Details : Running the "racer" program may cause one MDS thread to rename
4504 a file from being the source of a rename to being the target of
4505 a rename at exactly the same time that another thread is doing
4506 so, and the second thread has already enqueued these locks after
4507 doing a lookup of the target and is trying to relock them in
4508 order. Ensure that we don't try to re-lock the same resource.
4511 Frequency : only very large systems with liblustre clients
4513 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4514 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4515 all OSTs in parallel. This allows the eviction of stale liblustre
4516 clients to proceed much faster than if they were done in series,
4517 and also offers similar improvements for other set_info RPCs.
4522 Description: excessive CPU usage during initial read phase on client
4523 Details : During the initial read phase on a client, it would agressively
4524 retry readahead on the file, consuming too much CPU and impacting
4525 performance (since 1.4.5.8). Improve the readahead algorithm
4526 to avoid this, and also improve some other common cases (read
4527 of small files in particular, where "small" is files smaller than
4528 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4533 Description: MDS crash when receiving packet with unknown intent.
4534 Details : Do not LBUG in unknown intent case, just return -EFAULT
4536 Severity : enhancement
4537 Bugzilla : 9293, 9385
4538 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4539 Details : Do not serialize getattr (non-intent version) and statfs.
4542 Frequency : occasional, when OST network is overloaded/intermittent
4544 Description: client evicted by OST after bulk IO timeout
4545 Details : If a client sends a bulk IO request (read or write) the OST
4546 may evict the client if it is unresposive to its data GET/PUT
4547 request. This is incorrect if the network is overloaded (takes
4548 too long to transfer the RPC data) or dropped the OST GET/PUT
4549 request. There is no need to evict the client at all, since
4550 the pinger and/or lock callbacks will handle this, and the
4551 client can restart the bulk request.
4554 Frequency : Always when mmapping file with no objects
4556 Description: client crashes when mmapping file with no objects
4557 Details : Check that we actually have objects in a file before doing any
4558 operations on objects in ll_vm_open, ll_vm_close and
4564 Description: Request leak when working with deleted CWD
4565 Details : Introduce advanced request refcount tracking for requests
4566 referenced from lustre intent.
4568 Severity : Enhancement
4570 Description: Cache open file handles on client.
4571 Details : MDS now will return special lock along with openhandle, if
4572 requested and client is allowed to hold openhandle, even if unused,
4573 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4574 closing files for every read/write openration.
4576 Severity : Enhancement
4578 Description: Cache open negative dentries on client when possible.
4579 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4580 negative dentries on lock revocation.
4585 Description: Remounting a client read-only wasn't possible with a zconf mount
4586 Details : It wasn't possible to remount a client read-only with llmount.
4588 Severity : enhancement
4589 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4590 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4591 MPICH is included to add Lustre-specific ADIO interfaces.
4592 This is based closely on the UFS ADIO layer and only differs
4593 in file creation, in order to allow the OST striping to be set.
4594 This is user-contributed code and not supported by CFS.
4599 Description: extended inode attributes (immutable, append-only) work improperly
4600 when 2.4 and 2.6 kernels are used on client/server or vice versa
4601 Details : Introduce kernel-independent values for these flags.
4603 Severity : enhancement
4606 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4607 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4608 granularity. Now, Fractional MB granularity is be supported,
4609 this is very useful for low memory system.
4611 Severity : enhancement
4613 Description: Getattr by fid
4614 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4615 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4616 have locks and avoids getattr rpc requests for it on every stat.
4619 Frequency : Always, for filesystems larger than 2TB
4621 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4622 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4623 filesystems larger than 2TB to be mounted with mballoc enabled.
4626 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4628 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4629 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4630 filesystems larger than 2TB is mounted then the kernel will
4631 silently corrupt the start of the filesystem. CONFIG_LBD is
4632 enabled for all CFS-supported kernels, but the possibility of
4633 this happening with a modified kernel config exists.
4635 Severity : enhancement
4637 Description: add client O_DIRECT support for 2.6 kernels
4638 Details : It is now possible to do O_DIRECT reads and writes to files
4639 in the Lustre client mountpoint on 2.6 kernel clients.
4641 Severity : enhancement
4643 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4644 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4645 parallel, not waiting for response from every OST before sending
4646 a rpc to the next OST.
4651 Description: setattr vs write race when updating file timestamps
4652 Details : Client processes that update a file timestamp into the past
4653 right after writing to the file (e.g. tar) it is possible that
4654 the updated file modification time can be reset to the current
4655 time due to a race between processing the setattr and write RPC.
4657 Severity : enhancement
4659 Description: Bring 'lfs find' closer in line with regular Linux find.
4660 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4661 -print0 options and obtains all the needed info through the lustre
4664 Severity : enhancement
4666 Description: support up to 1024 configured devices on one node
4667 Details : change obd_dev array from statically allocated to dynamically
4668 allocated structs as they are first used to reduce memory usage
4673 Description: Flush dirty partially truncated pages during truncate
4674 Details : Immediatelly flush partially truncated pages in filter_setattr,
4675 this way we completely avoid having any pages in page cache on OST
4676 and can retire ugly workarounds during writes to flush such pages.
4681 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4682 Details : protect inode from truncation within vfs_unlink() context
4683 just take a reference before calling vfs_unlink() and release it
4684 when parent's i_sem is free.
4689 Description: last_id value checked outside lock on OST caused LASSERT failure
4690 Details : If there were multiple MDS->OST object precreate requests in
4691 flight, it was possible that the OST's last object id was checked
4692 outside a lock and incorrectly tripped an assertion. Move checks
4693 inside locks, and discard old precreate requests.
4696 Frequency : always, if extents are used on OSTs
4698 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4699 Details : OSTs using the extents format would not zero the high 16 bits of
4700 the index physical block number. This is not a problem for any
4701 OST filesystems smaller than 16TB, and no kernels support ext3
4702 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4703 new/modified files) and can be fixed for existing filesystems
4704 with e2fsprogs-1.39-cfs1.
4709 Description: import connection selection may be incorrect if timer wraps
4710 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4711 import connections to be ignored if the 32-bit jiffies counter
4712 wraps. Use a 64-bit jiffies counter.
4715 Frequency : very large clusters immediately after boot
4717 Description: LNET request buffers exhausted under heavy short-term load
4718 Details : If a large number of client requests are generated on a service
4719 that has previously never seen so many requests it is possible
4720 that the request buffer growth cannot keep up with the spike in
4721 demand. Instead of dropping incoming requests, they are held in
4722 the LND until the RPC service can accept more requests.
4725 Frequency : Sometimes during replay
4727 Description: Assertion failure in ll_local_open after replay.
4728 Details : If replay happened on an open request reply before we were able
4729 to set replay handler, reply will become not swabbed tripping the
4730 assertion in ll_local_open. Now we set the handler right after
4731 recognising of open request
4734 Frequency : very rare
4736 Description: kernel reports "badness in vsnprintf"
4737 Details : Reading from the "recovery_status" /proc file in small chunks
4738 may cause a negative length in lprocfs_obd_rd_recovery_status()
4739 call to vsnprintf() (which is otherwise harmless). Exit early
4740 if there is no more space in the output buffer.
4742 Severity : enhancement
4744 Description: clear OBD RPC statistics by writing to them
4745 Details : It is now possible to clear the OBD RPC statistics by writing
4746 to the "stats" file.
4751 Description: Client mtime is not the same on different clients after utimes
4752 Details : In some cases, the client was using the utimes() syscall on
4753 a file cached on another node. The clients now validate the
4754 ctime from the MDS + OSTs to determine which one is right.
4759 Description: Inability to activate failout mode
4760 Details : lconf script incorrectly assumed that in pythong string's numeric
4761 value is used in comparisons.
4764 Frequency : always with multiple stripes per file
4766 Description: Inefficient object allocation for mutli-stripe files
4767 Details : When selecting which OSTs to stripe files over, for files with
4768 a stripe count that divides evenly into the number of OSTs,
4769 the MDS is always picking the same starting OST for each file.
4770 Return the OST selection heuristic to the original design.
4775 Description: mount failures may take full timeout to return an error
4776 Details : Under some heavy load conditions it is possible that a
4777 failed mount can wait for the full obd_timeout interval,
4778 possibly several minutes, before reporting an error.
4779 Instead return an error as soon as the status is known.
4781 Frequency : quota enabled and large files being deleted
4783 Description: releasing more than 4GB of quota at once hangs OST
4784 Details : If a user deletes more than 4GB of files on a single OST it
4785 will cause the OST to spin in an infinite loop. Release
4786 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4791 Description: statfs data retrieved from /proc may be stale or zero
4792 Details : When reading per-device statfs data from /proc, in the
4793 {kbytes,files}_{total,free,avail} files, it may appear
4794 as zero or be out of date.
4797 Frequency : systems with MD RAID1 external journal devices
4799 Description: lconf's call to blkid is confused by RAID1 journal devices
4800 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4801 of returning all block devices that match the journal UUID.
4804 Frequency : always, for aggregate stripe size over 4GB
4806 Description: assertion fails when trying to use 4GB stripe size
4807 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4808 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4811 Frequency : always on ppc64
4813 Description: the first write on an ext3 filesystem with mballoc got stuck
4814 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4815 perform endianness conversion.
4817 ------------------------------------------------------------------------------
4819 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4821 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4822 INTEROPERATE with older versions automatically. Please read the
4823 user documentation before upgrading any part of a live system.
4824 * WARNING: Lustre networking configuration changes are required with
4825 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4828 * Support for kernels:
4829 2.6.9-22.0.2.EL (RHEL 4)
4830 2.6.5-7.244 (SLES 9)
4831 2.6.12.6 vanilla (kernel.org)
4834 Severity : enhancement
4835 Bugzilla : 7981/8208
4836 Description: Introduced Lustre Networking (LNET)
4837 Details : LNET is new networking infrastructure for Lustre, it includes
4838 a reorganized network configuration mode (see the user
4839 documentation for full details) as well as support for routing
4840 between different network fabrics. Lustre Networking Devices
4841 (LNDs) for the supported network fabrics have also been
4842 created for this new infrastructure.
4844 Severity : enhancement
4845 Description: Introduced Access control lists
4846 Details : clients can set ACLs on files and directories in order to have
4847 more fine-grained permissions than the standard Unix UGO+RWX.
4848 The MDS must be started with the "-o acl" mount option.
4850 Severity : enhancement
4851 Description: Introduced filesystem quotas
4852 Details : Administrators may now establish per-user quotas on the
4855 Severity : enhancement
4857 Description: Configuration change for the XT3
4858 The PTLLND is now used to run Lustre over Portals on the XT3
4859 The configure option(s) --with-cray-portals are no longer used.
4860 Rather --with-portals=<path-to-portals-includes> is used to
4861 enable building on the XT3. In addition to enable XT3 specific
4862 features the option --enable-cray-xt3 must be used.
4867 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4868 Details : A race between allocating a new llcd and re-getting the llcd_lock
4869 allowed another thread to grab newly-allocated llcd.
4871 Severity : enhancement
4873 Description: 2.6 OST async journal commit and locking fix to improve performance
4874 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4875 2.6 kernels are now async as they already were in 2.4 kernels so
4876 that they can commit concurrently with the network bulk transfer.
4877 For block-allocated files the filter allocation semaphore is held
4878 to avoid filesystem fragmentation during allocation. BKL lock
4879 removed for 2.6 xattr operations where it is no longer needed.
4884 Description: lconf incorrectly determined whether two IP networks could talk
4885 Details : In some more complicated routing and multiple-network
4886 configurations, lconf will avoid trying to make a network
4887 connection to a disjoint part of the IP space. It was doing the
4888 math incorrectly for one set of cases.
4893 Description: Fix for potential infinite loop processing records in an llog.
4894 Details : If an llog record is corrupted/zeroed, it is possible to loop
4895 forever in llog_process(). Validate the llog record length
4896 and skip the remainder of the block on error.
4899 Frequency : occasional (liblustre only)
4901 Description: liblustre could not open files whose last component is a symlink
4902 Details : sysio_path_walk() would incorrectly pass the open intent to
4903 intermediate path components.
4906 Frequency : rare (liblustre only with non-standard tuning)
4907 Bugzilla : 7201 (7350)
4908 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4909 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4910 after it has been released, only do this for referenced locks
4912 Severity : enhancement
4914 Description: specifying an (invalid) directory default stripe_size of -1
4915 would reset the directory default striping
4916 Details : stripe_size -1 was used internally to signal directory stripe
4917 removal, now use "all default" to signal dir stripe removal
4918 as a directory striping of "all default" is not useful
4921 Frequency : common for large clusters running liblustre clients
4923 Description: doing an ls when liblustre clients are running is slow
4924 Details : sending a glimpse AST to a liblustre client waits for every AST
4925 to time out, as liblustre clients will not respond. Since they
4926 cannot cache data we refresh the OST lock LVB from disk instead.
4928 Severity : enhancement
4930 Description: doing an ls at the same time as file IO can be slow
4931 Details : enqueue and other "small" requests can be blocked behind many
4932 large IO requests. Create a new OST IO portal for non-IO
4933 requests so they can be processed faster.
4936 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4938 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4939 created file with mode 000
4940 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4941 allow an ftruncate() as a non-root user to a file with mode 000.
4942 Lustre now respects this flag to disable mode checks when
4943 truncating a file owned by the user
4946 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4949 Description: Revoking locks from clients that went dead or catatonic might take
4951 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4952 cancellation of such locks instant on servers without waiting for
4953 any reply from clients. Clients drops these locks when cancel
4954 notification from server is received without replying.
4957 Frequency : liblustre-only, when liblustre client dies or becomes busy
4959 Description: Doing ls on Linux clients can take a long time with active
4961 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4962 granting such locks to it in the first place if possible. Locks
4963 are taken by proxy on the OST during the read or write and
4964 dropped immediately afterward. Add connect flags handling, do
4965 not grant locks to liblustre clients for glimpse ASTs.
4967 Severity : enhancement
4969 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4970 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4971 sequential reads and behaves badly with applications doing random
4972 reads. Improve it by reading ahead at least read region, and
4973 avoiding excessive large RPC for small reads.
4975 Severity : enhancement
4977 Description: Creating more than 1000 files for a single job may cause a load
4978 imbalance on the OSTs if there are also a large number of OSTs.
4979 Details : qos_prep_create() uses an OST index reseed value that is an
4980 even multiple of the number of available OSTs so that if the
4981 reseed happens in the middle of the object allocation it will
4982 still utilize the OSTs as uniformly as possible.
4987 Description: OST or MDS may oops in ping_evictor_main()
4988 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4989 but doesn't restart at beginning of obd_exports_timed list
4992 Severity : enhancement
4994 Description: improve by-nid export eviction on the MDS and OST
4995 Details : allow multiple exports with the same NID to be evicted at one
4996 time without re-searching the exports list.
4999 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5001 Description: MDS may oops in groups_free()
5002 Details : in rare race conditions a newly allocated group_info struct is
5003 freed again, and this can be NULL. The 2.4 compatibility code
5004 for groups_free() checked for a NULL pointer, but 2.6 did not.
5007 Frequency : common for liblustre clients doing little filesystem IO
5008 Bugzilla : 9352, 7313
5009 Description: server may evict liblustre clients accessing contended locks
5010 Details : if a client is granted a lock or receives a completion AST
5011 with a blocking AST already set it would not reply to the AST
5012 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5015 Frequency : lfs setstripe, only systems with more than 160 OSTs
5017 Description: unable to set striping with a starting offset beyond OST 160
5018 Details : llapi_create_file() incorrectly limited the starting stripe
5019 index to the maximum single-file stripe count.
5022 Frequency : LDAP users only
5024 Description: lconf did not handle in-kernel recovery with LDAP properly
5025 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5027 Severity : enhancement
5029 Description: bind OST threads to NUMA nodes to improve performance
5030 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5031 node and do their allocations there to localize memory access
5033 Severity : enhancement
5035 Description: llmount can determine client NID directly from Myrinet (GM)
5036 Details : the client NID code from gmnalnid was moved directly into
5037 llmount, removing the need to use this or specifying the
5038 client NID explicitly when mounting GM clients with zeroconf
5041 Frequency : if client is started with down MDS
5043 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5044 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5045 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5046 ptlrpc_import_delayed_req() because we didn't check if the
5047 request was interrupted, and we also didn't break out of the
5048 event loop if there was no timeout
5053 Description: data loss during non-page-aligned writes to a single file from
5054 both multiple nodes and multiple threads on one node at same time
5055 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5056 inconsistency led to false short-reads, that were cached and later
5057 used by ->prepare_write() to fill in partially written page,
5058 leading to data loss.
5061 Frequency : always, if lconf --abort_recovery used
5063 Description: lconf --abort_recovery fails with 'Operation not supported'
5064 Details : lconf was attempting to abort recovery on the MDT device and not
5067 Severity : enhancement
5069 Description: remove cleanup logs
5070 Details : replace lconf-generated cleanup logs with lustre internal
5071 cleanup routines. Eliminates the need for client-cleanup and
5074 Severity : enhancement
5076 Description: add support for EAs (user and system) on lustre filesystems
5077 Details : it is now possible to store extended attributes in the Lustre
5078 client filesystem, and with the user_xattr mount option it
5079 is possible to allow users to store EAs on their files also
5081 Severity : enhancement
5083 Description: Add possibility (config option) to show minimal available OST free
5085 Details : When compiled with --enable-mindf configure option, statfs(2)
5086 (and so, df) will return least minimal free space available from
5087 all OSTs as amount of free space on FS, instead of summary of
5088 free spaces of all OSTs.
5090 Severity : enhancement
5092 Description: do not expand extent locks acquired on OST-side
5093 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5094 by server: they are not cached anyway.
5097 Frequency : when mmap is used/binaries executed from Lustre
5099 Description: Unmmap pages before throwing them away from read cache.
5100 Details : llap_shrink cache now attempts to unmap pages before discarding
5101 them (if unmapping failed - do not discard). SLES9 kernel has
5102 extra checks that trigger if this unmapping is not done first.
5107 Description: lconf didn't resolve symlinks before checking to see whether a
5108 given mountpoint was already in use
5111 Frequency : when migrating failover services
5112 Bugzilla : 6395, 9514
5113 Description: When migrating a subset of services from a node (e.g. failback
5114 from a failover service node) the remaining services would
5115 time out and evict clients.
5116 Details : lconf --force (implied by --failover) sets the global obd_timeout
5117 to 5 seconds in order to quickly disconnect, but this caused
5118 other RPCs to time out too quickly. Do not change the global
5119 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5121 Severity : enhancement
5122 Frequency : if MDS is started with down OST
5123 Bugzilla : 9439,5706
5124 Description: Allow startup/shutdown of an MDS without depending on the
5125 availability of the OSTs.
5126 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5127 Add appropriate locking and lov-osc refcounts for safe
5128 cleaning. Add osc abort_inflight calls in case the
5129 synchronize never started.
5132 Frequency : occasional (Cray XT3 only)
5134 Description: root not authorized to access files in CRAY_PORTALS environment
5135 Details : The client process capabilities were not honoured on the MDS in
5136 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5137 been accessed by an authorized user then root was able to access
5138 the file on the local client also. The root user capabilities
5139 are now allowed on the MDS, as this environment has secure UID.
5142 Frequency : occasional
5144 Description: ldiskfs "too long searching" message happens too often
5145 Details : A debugging message (otherwise harmless) prints too often on
5146 the OST console. This has been reduced to only happen when
5147 there are fragmentation problems on the filesystem.
5152 Description: Division by zero in statfs when all OSCs are inactive
5153 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5154 lov_statfs divides by value returned by lov_get_stripecnt().
5158 Bugzilla : 9489, 3273
5159 Description: First write from each client to each OST was only 4kB in size,
5160 to initialize client writeback cache, which caused sub-optimal
5161 RPCs and poor layout on disk for the first written file.
5162 Details : Clients now request an initial cache grant at (re)connect time
5163 and so that they can start streaming writes to the cache right
5164 away and always do full-sized RPCs if there is enough data.
5165 If the OST is rebooted the client also re-establishes its grant
5166 so that client cached writes will be honoured under the grant.
5171 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5172 Details : Now I/O RPCs go to different portal number and (presumably) fast
5173 lock requests (and glimses) and other RPCs get their own service
5174 threads pool that should be able to service those RPCs
5177 Severity : enhancement
5179 Description: Ability to exchange lustre version between client and servers and
5180 issue warnings at client side if client is too old. Also for
5181 liblustre clients there is ability to refuse connection of too old
5183 Details : New 'version' field is added to connect data structure that is
5184 filled with version info. That info is later checked by server and
5188 Frequency : rare, liblustre only.
5189 Bugzilla : 9296, 9581
5190 Description: Two simultaneous writes from liblustre at offset within same page
5191 might proceed at the same time overwriting eachother with stale
5193 Details : I/O lock withing llu_file_prwv was released too early, before data
5194 actually was hitting the wire. Extended lock-holding time until
5195 server acknowledges receiving data.
5198 Frequency : extremely rare. Never observed in practice.
5200 Description: avoid generating lustre_handle cookie of 0.
5201 Details : class_handle_hash() generates handle cookies by incrementing
5202 global counter, and can hit 0 occasionaly (this is unlikely, but
5203 not impossible, because initial value of cookie counter is
5204 selected randonly). Value of 0 is used as a sentinel meaning
5205 "unassigned handle" --- avoid it. Also coalesce two critical
5206 sections in this function into one.
5208 Severity : enhancement
5210 Description: allow liblustre clients to delegate truncate locking to OST
5211 Details : To avoid overhead of locking, liblustre client instructs OST to
5212 take extent lock in ost_punch() on client's behalf. New connection
5213 flag is added to handle backward compatibility.
5215 Severity : enhancement
5216 Bugzilla : 4928, 7341, 9758
5217 Description: allow number of OST service threads to be specified
5218 Details : a module parameter allows the number of OST service threads
5219 to be specified via "options ost ost_num_threads={N}" in the
5220 OSS's /etc/modules.conf or /etc/modprobe.conf.
5224 Bugzilla : 6146, 9635, 9895
5225 Description: servers crash with bad pointer in target_handle_connect()
5226 Details : In rare cases when a client is reconnecting it was possible that
5227 the connection request was the last reference for that export.
5228 We would temporarily drop the export reference and get a new
5229 one, but this may have been the last reference and the export
5230 was just destroyed. Get new reference before dropping old one.
5232 Severity : enhancement
5233 Frequency : if client is started with failover MDS
5235 Description: Allow multiple MDS hostnames in the mount command
5236 Details : Try to read the configuration from all specified MDS
5237 hostnames during a client mount in case the "primary"
5240 Severity : enhancement
5242 Description: Stop sending data to evicted clients as soon as possible.
5243 Details : Check if the client we are about to send or are sending data to
5244 was evicted already. (Check is done every second of waiting,
5245 for which l_wait_event interface was extended to allow checking
5246 of exit condition at specified intervals).
5249 Frequency : rare, normally only when NFS exporting is done from client
5251 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5253 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5256 Frequency : always, if srand() is called before liblustre initialization
5258 Description: Liblustre uses system PRNG disturbing its usage by user application
5259 Details : Introduce internal to lustre fast and high-quality PRNG for
5260 lustre usage and make liblustre and some other places in generic
5261 lustre code to use it.
5263 Severity : enhancement
5264 Bugzilla : 9477, 9557, 9870
5265 Description: Verify that the MDS configuration logs are updated when xml is
5266 Details : Check if the .xml configuration logs are newer than the config
5267 logs stored on the MDS and report an error if this is the case.
5268 Request --write-conf, or allow starting with --old_conf.
5270 Severity : enhancement
5272 Description: Handle symlinks in the path when checking if Lustre is mounted.
5273 Details : Resolve intermediate symlinks when checking if a client has
5274 mounted a filesystem to avoid duplicate client mounts.
5279 Description: lconf can hit an error exception but still return success.
5280 Details : The lconf command catches the Command error exception at the top
5281 level script context and will exit with the associated exit
5282 status, but doesn't ensure that this exit status is non-zero.
5287 Description: failure of ptlrpc thread startup can cause oops
5288 Details : Starting a ptlrpc service thread can fail if there are a large
5289 number of threads or the server memory is very fragmented.
5290 Handle this without oopsing.
5293 Frequency : always, only if liblustre and non-default acceptor port was used
5295 Description: liblustre cannot connect to servers with non-default acceptor port
5296 Details : tcpnal_set_default_params() was not called and was therefore
5297 ignoring the environment varaible TCPNAL_PORT, as well as other
5298 TCPNAL_ environment variables
5303 Description: two objects could be created on the same OST for a single file
5304 Details : If an OST is down, in some cases it was possible to create two
5305 objects on a single OST for a single file. No problems other
5306 than potential performance impact and spurious error messages.
5310 Bugzilla : 5681, 9562
5311 Description: Client may oops in ll_unhash_aliases
5312 Details : Client dcache may become inconsistent in race condition.
5313 In some cases "getcwd" can fail if the current directory is
5319 Description: Inode refcounting problems in NFS export code
5320 Details : link_raw functions used to call d_instantiate without obtaining
5321 extra inode reference first.
5325 Bugzilla : 9942, 9903
5326 Description: Referencing freed requests leading to crash, memleaks with NFS.
5327 Details : We used to require that call to ll_revalidate_it was always
5328 followed by ll_lookup_it. Also with revalidate_special() it is
5329 possible to call ll_revalidate_it() twice for the same dentry
5330 even if first occurence returned success. This fix changes semantic
5331 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5332 reference on a request referred from the intent.
5333 ll_intent_release() then releases such a request.
5336 Frequency : rare, normally benchmark loads only
5338 Description: unlinked inodes were kept in memory on the client
5339 Details : If a client is repeatedly creating and unlinking files it
5340 can accumulate a lot of stale inodes in the inode slab cache.
5341 If there is no other client load running this can cause the
5342 client node to run out of memory. Instead flush old inodes
5343 from client cache that have the same inode number as a new inode.
5346 Frequency : SLES9 2.6.5 kernel and long filenames only
5347 Bugzilla : 9969, 10379
5348 Description: utime reports stale NFS file handle
5349 Details : SLES9 uses out-of-dentry names in some cases, which confused
5350 the lustre dentry revalidation. Change it to always use the
5354 Frequency : rare, unless heavy write-truncate concurrency is continuous
5355 Bugzilla : 4180, 6984, 7171, 9963, 9331
5356 Description: OST becomes very slow and/or deadlocked during object unlink
5357 Details : filter_destroy() was holding onto the parent directory lock
5358 while truncating+unlinking objects. For very large objects this
5359 may block other threads for a long time and slow overall OST
5360 responsiveness. It may also be possible to get a lock ordering
5361 deadlock in this case, or run out of journal credits because of
5362 the combined truncate+unlink. Solution is to do object truncate
5363 first in one transaction without parent lock, and then do the
5364 final unlink in a new transaction with the parent lock. This
5365 reduces the lock hold time dramatically.
5368 Frequency : rare, 2.4 kernels only
5370 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5371 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5372 before unmounting their filesystem in order to avoid deadlock.
5373 The kernel_locked() function in 2.4 kernels only checks whether
5374 the kernel lock is held, not whether it is this process that is
5375 holding it as 2.6 kernels do.
5380 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5381 Details : The client ptlrpc code may be trying to reconnect to a down
5382 server before a previous connection attempt has timed out.
5383 Increase the reconnect interval to be longer than the connection
5384 timeout interval to avoid sending duplicate connections to
5388 Frequency : echo_client brw_test command
5390 Description: fix echo_client to work with OST preallocated code
5391 Details : OST preallocation code (5137) didn't take echo_client IO path
5392 into account: echo_client calls filter methods outside of any
5393 OST thread and, hence, there is no per-thread preallocated
5394 pages and buffers to use. Solution: hijack pga pages for IO. As
5395 a byproduct, this avoids unnecessary data copying.
5399 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5400 Description: Client can oops in mdc_commit_close() after open replay
5401 Details : It was possible for the MDS to return an open request with no
5402 transaction number in mds_finish_transno() if the client was
5403 evicted, but without actually returning an error. Clients
5404 would later try to replay that open and may trip an assertion
5405 Simplify the client close codepath, and always return an error
5406 from the MDS in case the open is not successful.
5409 Frequency : rare, 2.6 OSTs only
5411 Description: OST may deadlock under high load on fragmented files
5412 Details : If there was a heavy load and highly-fragmented OST filesystems
5413 it was possible to have all the OST threads deadlock waiting on
5414 allocation of biovecs, because the biovecs were not released
5415 until the entire RPC IO was completed. Instead, release biovecs
5416 as soon as they are complete to ensure forward IO progress.
5418 Severity : enhancement
5420 Description: Support for specifying external journal device at mount
5421 Details : If an OST or MDS device is formatted with an external journal
5422 device, this device major/minor is stored in the ext3 superblock
5423 and may not be valid for failover. Allow detecting and
5424 specifying the external journal at mount time.
5429 Description: Mounting an MDS with pending unlinked files may cause oops
5430 Details : target_finish_recovery() calls mds_postrecov() which returned
5431 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5432 considers this an error and immediately begins cleaning up the
5433 lov, just after starting the mds_lov process
5435 Severity : enhancement
5437 Description: Implement 'lfs df' to report actual free space on per-OST basis
5438 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5439 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5440 usage of objects; '-h' to report in human readable format.
5442 ------------------------------------------------------------------------------
5444 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5451 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5452 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5453 at startup during orphan unlinking. Instead, skip buddy update
5454 before setup, as it will be regenerated after recovery is complete.
5459 Description: If an OST is inactive, its locks might reference stale inodes.
5460 Details : lov_change_cbdata() must iterate over all namespaces, even if
5461 they are inactive to clear inode references from the lock.
5463 Severity : enhancement
5464 Frequency : occasional, if non-standard max_dirty_mb used
5466 Description: Client will block write RPCs if not enough grant
5467 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5468 then the client will block writes while waiting for another RPC
5469 to complete instead of consuming its dirty limit. With change
5470 we get improved performance when max_dirty_mb is small.
5472 Severity : enhancement
5473 Bugzilla : 3389, 6253
5474 Description: Add support for supplementary groups on the MDS.
5475 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5476 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5477 lookups for user supplementary groups into a cache.
5481 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5482 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5483 for liblustre O_EXCL is enforced by MDS.
5487 Description: readdir never returns NULL in liblustre.
5488 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5489 dentry in struct dirent.
5493 Frequency : liblustre only, depends on application IO pattern
5494 Description: liblustre clients evicted if not contacting servers
5495 Details : Don't put liblustre clients into the ping_evictor list, so
5496 they will not be evicted by the pinger ever.
5498 Severity : enhancement
5500 Description: Add ability to evict clients by NID from MDS.
5501 Details : By echoing "nid:$NID" string into
5502 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5503 $NID would be instantly evicted from this MDS and from all active
5504 OSTs connected to it.
5508 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5509 Details : lookup_it_finish() used to query file size from OSTs that was not
5514 Description: service threads change working directory to that of init
5515 Details : Starting lustre service threads may pin the working directory
5516 of the parent thread, making that filesystem busy. Threads
5517 now change to the working directory of init to avoid this.
5521 Frequency : during shutdown only
5522 Description: shutdown with a failed MDS or OST can cause unmount to hang
5523 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5528 Frequency : 2.6 only
5529 Description: chmod/chown may include an extra supplementary group
5530 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5531 supplementary group and if none is specified this is used.
5534 Bugzilla : 5479 (6816)
5536 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5537 Details : If lookup is in progress on a file that is unlinked we might try
5538 to revalidate the inode and fail in revalidate after lookup is
5539 complete and ll_file_open() enqueues the open again but
5540 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5543 Frequency : always, if lconf --abort_recovery used
5545 Description: lconf --abort_recovery fails with 'Operation not supported'
5546 Details : lconf was attempting to abort recovery on the MDT device and not
5549 ------------------------------------------------------------------------------
5551 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5556 Frequency : rare (only unsupported configurations with a node running as an
5558 Bugzilla : 6514, 5137
5559 Description: Mounting a Lustre file system on a node running as an OST could
5561 Details : OSTs now preallocates memory needed to write out data at
5562 startup, instead of when needed, to avoid having to
5563 allocate memory in possibly low memory situations.
5564 Specifically, if the file system is mounted on on OST,
5565 memory pressure could force it to try to write out data,
5566 which it needed to allocate memory to do. Due to the low
5567 memory, it would be unable to do so and the node would
5568 become unresponsive.
5570 Severity : enhancement
5572 Description: Addition of lconf --service command line option
5573 Details : lconf now accepts a '--service <arg>' option, which is
5574 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5576 Severity : enhancement
5578 Description: Failover mode is now the default for OSTs.
5579 Details : By default, OSTs will now run in failover mode. To return to
5580 the old behaviour, add '--failout' to the lmc line for OSTs.
5582 Severity : enhancement
5584 Description: Health checks are now provided for MDS and OSTs
5585 Details : Additional detailed health check information on MSD and OSTs
5586 is now provided through the procfs health_check value.
5589 Frequency : occasional, depends on IO load
5591 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5592 after numerous create/delete cycles
5593 Details : The ext3 inode allocation policy would not allocate new inodes
5594 very well on the OSTs because there are no new directories
5595 being created. Instead we look for groups with free space if
5596 the parent directories are nearly full.
5601 Description: Network or server problems during mount may cause partially
5602 mounted clients instead of returning an error.
5603 Details : The config llog parsing code may overwrite the error return
5604 code during mount error handling, returning success instead
5610 Description: MDS can fail to allocate large reply buffers
5611 Details : After long uptimes the MDS can fail to allocate large reply
5612 buffers (e.g. zconf client mount config records) due to memory
5613 fragmentation or consumption by the buffer cache. Preallocate
5614 some large reply buffers so that these replies can be sent even
5615 under memory pressure.
5619 Frequency : rare (liblustre)
5620 Description: fsx running with liblustre complained that using truncate() to
5621 extend the file doesn't work. This patch corrects that issue.
5622 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5623 ATTR_SIZE and lsm use in llu_setattr_raw.
5627 Frequency : rare, only 2.6 kernels
5628 Description: Unusual file access patterns on the MDS may result in inode
5629 data being lost in very rare circumstances.
5630 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5631 code in the 2.6 kernel caused the inode and/or EA data not to
5632 be read from disk, causing single-file corruption.
5636 Frequency : rare, only 2.6 filesystems using extents
5637 Description: Heavy concurrent write and delete load may cause data corruption.
5638 Details : It was possible under high-load situations to have an extent
5639 metadata block in the block device cache from a just-unlinked
5640 file overwrite a newly-allocated data block. We now unmap any
5641 metadata buffers that alias just-allocated data blocks.
5645 Frequency : filesystems with default stripe_count larger than 77
5646 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5647 Details : lconf specifies an inode size of 4096 bytes when the default
5648 stripe_count is larger than 77. This conflicts with the default
5649 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5650 this case to avoid pinning too much memory for large EAs.
5652 ------------------------------------------------------------------------------
5654 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5659 Frequency : rare (extremely heavy IO load with hundreds of clients)
5661 Description: Client is evicted, gets IO error writing to file
5662 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5663 caused clients to be evicted for AST timeouts. The fixes in
5664 bug 5192 mean we no longer need to have such short AST timeouts
5665 so ldlm_timeout has been increased.
5668 Frequency : occasional during --force or --failover shutdown under load
5669 Bugzilla : 5949, 4834
5670 Description: Server oops/LBUG if stopped with --force or --failover under load
5671 Details : a collection of import/export refcount and cleanup ordering
5672 issues fixed for safer force cleanup
5675 Frequency : only filesystems larger than 120 OSTs
5676 Bugzilla : 5990, 6223
5677 Description: lfs getstripe would oops on a very large filesystem
5678 Details : lov_getconfig used kfree on vmalloc'd memory
5681 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5682 Bugzilla : 6242, 6243
5683 Description: reading from files that had been truncated to a non-zero size
5684 but never opened returned no data
5685 Details : ll_file_read() reads zeros from no-object files to EOF
5690 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5691 Details : There's little harm in aborting MDS/OSS recovery and letting it
5692 try again, so I removed the LASSERT and return an error instead.
5694 Severity : enhancement
5696 Description: New debugging infrastructure for tracking down data corruption
5697 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5698 (b) cover more of the client-side code path, and (c) try to narrow
5699 down where problems occurred
5703 Bugzilla : 3819, 4364, 4397, 6313
5704 Description: Racing close and eviction MDS could cause assertion in mds_close
5705 Details : It was possible to get multiple mfd references during close and
5706 client eviction, leading to one thread referencing a freed mfd.
5708 Severity: : enhancement
5709 Bugzilla : 3262, 6359
5710 Description: Attempts to reconnect to servers are now more aggressive.
5711 Details : This builds on the enhanced upcall-less recovery that was added
5712 in 1.4.2. When trying to reconnect to servers, clients will
5713 now try each server in the failover group every 10 seconds. By
5714 default, clients would previously try one server every 25 seconds.
5719 Description: After recovery, certain operations trigger a failed
5720 assertion on a client.
5721 Details : Failing over an mds, using lconf -d --failover, while a
5722 client was doing a readdir() call would cause the client to
5723 LBUG after recovery completed and the readdir() was resent.
5725 Severity : enhancement
5727 Description: Default groups are now added by lconf
5728 Details : You can now run lconf --group <servicename> without having to
5729 manually add groups with lmc.
5732 Frequency : occasional
5734 Description: Nodes with an elan id of 0 trigger a failed assertion
5737 Frequency : always when accessing e.g. tty/console device nodes
5739 Description: tty and some other devices nodes cannot be used on lustre
5740 Details : file's private_data field is used by device data and lustre
5741 values in there got lost. New field was added to struct file to
5742 store fs-specific private data.
5745 Frequency : when exporting Lustre via NFS
5747 Description: NFSD failed occasionally when looking up a path component
5748 Details : NFSD is looking up ".." which was broken in ext3 directories
5749 that had grown large enough to become hashed.
5752 Frequency : Clusters with multiple interfaces not on the same subnet
5754 Description: Nodes will repeatedly try to reconnect to an interface which it
5755 cannot reach and report an error to the log.
5756 Details : Extra peer list entries will be created by lconf with some peers
5757 unreachable. lconf now validates the peer before adding it.
5760 Frequency : Only if a default stripe is set on the filesystem root.
5762 Description: Setting a default stripe on the filesystem root prevented the
5763 filesystem from being remounted.
5764 Details : The client was sending extra request flags in the root getattr
5765 request and did not allocate a reply buffer for the dir EA.
5768 Frequency : occasional, higher if lots of files are accessed by one client
5769 Bugzilla : 6159, 6097
5770 Description: Client trips assertion regarding lsm mismatch/magic
5771 Details : While revalidating inodes the VFS looks up inodes with ifind()
5772 and in rare cases can find an inode that is being freed.
5773 The ll_test_inode() code will free the lsm during ifind()
5774 when it finds an existing inode and then the VFS later attaches
5775 this free lsm to a new inode.
5779 Bugzilla : 6422, 7030
5780 Description: MDS deadlock between mkdir and client eviction
5781 Details : Creating a new file via mkdir or mknod (starting a transaction
5782 and getting the ns lock) can deadlock with client eviction
5783 (gets ns lock and trying to finish a synchronous transaction).
5786 Frequency : occasional
5787 Description: While starting a server, the fsfilt_ext3 module could not be
5789 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5790 kernels. Previously, lconf would still use the ext3 name
5791 when trying to load modules. Now, it will correctly use
5792 ext3 on 2.4 and ldiskfs on 2.6.
5794 Severity : enhancement
5795 Description: The default stripe count has been changed to 1
5796 Details : The interpretation of the default stripe count (0, to lfs
5797 or lmc) has been changed to mean striping across a single
5798 OST, rather than all available. For general usage we have
5799 found a stripe count of 1 or 2 works best.
5801 Severity : enhancement
5802 Description: Add support for compiling against Cray portals.
5803 Details : Conditional compiling for some areas that are different
5807 Frequency : occasional
5808 Bugzilla : 6409, 6834
5809 Description: Creating files with an explicit stripe count may lead to
5810 a failed assertion on the MDS
5811 Details : If some OSTs are full or unavailable, creating files may
5812 trigger a failed assertion on the MDS. Now, Lustre will
5813 try to use other servers or return an error to the
5817 Frequency : occasional
5819 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5820 caused lock timeout during readahead (since 1.4.2).
5821 Details : Processes doing readahead might match a lock that hasn't been
5822 granted yet if there are overlapping and conflicting lock
5823 requests. The readahead process waits on ungranted lock
5824 (original lock is CBPENDING), while OST waits for that process
5825 to cancel CBPENDING read lock and eventually evicts client.
5827 Severity : enhancement
5829 Description: Initial enabling of flock support for clients
5830 Details : Implements fcntl advisory locking and file status functions.
5831 This feature is provided as an optional mount flag (default
5832 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5833 locking are implemented yet, and those that are are not guaranteed
5834 to be completely correct in production environments.
5835 mount -t lustre -o [flock|noflock] ...
5838 Frequency : occasional
5840 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5841 assertion in the ext3 JBD (journaling) layer.
5842 Details : The b_committed_data struct is protected by the big kernel lock
5843 in 2.4 kernels, serializing journal_commit_transaction() and
5844 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5845 access to this struct. In 2.6 kernels there is finer grained
5846 locking to improve SMP performance of the JBD layer.
5850 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5855 Description: merge and cleanup kernel patches.
5858 -----------------------------------------------------------------------------
5860 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5862 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5863 versions of Lustre. Please update all servers and clients to
5864 version 1.4.2 or later at the same time. You must also run
5865 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5866 to update the configuration logs.
5868 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5869 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5870 - swab llog records of type '0' so we get proper header size/idx (5861)
5871 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5872 - fix rename of one directory over another leaking an inode (5953)
5873 - avoid SetPageDirty on 2.6 (5981)
5874 - don't re-add just-being-destroyed locks to the waiting list (5653)
5875 - when creating new directories, inherit the parent's custom
5876 striping settings if present parent (3048)
5877 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5878 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5879 - don't hold client locks on temporary worklist from l_lru (5666)
5880 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5881 - automatically evict dead clients (5921)
5882 - Update file size properly in create+truncate+fstat case (6196)
5883 - Do not unhash mountpoint dentries, do not allow removal of
5885 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5886 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5887 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5888 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5889 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5890 - Correctly handle reads of files with no objects (6243)
5891 - lctl recover will also mark a device active if deactivate used (5933)
5893 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5894 - allow --write-conf on an MDS with different nettype than client (5619)
5895 - don't write config llogs to MDS for mounts not from that MDS (5617)
5896 - lconf should create multiple TCP connections from a client (5201)
5897 - init scripts are now turned off by default; run chkconfig --on
5898 lustre and chkconfig --on lustrefs to use them
5899 - upcalls are no longer needed for clients to recover to failover
5901 - add --abort-recovery option to lconf to abort recovery on device
5903 - add support for an arbitrary number of OSTs (3026)
5904 - Quota support protocol changes.
5905 - forward compatibility changes to wire structs (6007)
5906 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5907 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5908 - improved "lustre status" script
5909 - initialize blocksize for non-regular files (6062)
5910 - added --disable-server and --disable-client configure options (5782)
5911 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5912 - Vanilla 2.4.29 support
5913 - increase maximum number of obd devices to 520 (6242)
5914 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5915 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5916 SLES 9 2.6.5 kernels
5917 - limit stripes per file to 160 (the maximum EA size) (6093)
5919 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5922 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5923 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5924 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5925 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5926 - small ext3 extents cleanups and fixes (5733)
5927 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5928 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5929 - avoid needless client->OST connect, fix handle mismatch (5317)
5930 - fix DLM error path that led to out-of-sync client, long delays (5779)
5931 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5932 - fix several locking issues related to i_size (5492,5624,5654,5672)
5933 - don't move pending lock onto export if it is already evicted (5683)
5934 - fix kernel oops when creating .foo in unlinked directory (5548)
5935 - fix deadlock in obdfilter statistics vs. object create (5811)
5936 - use time_{before,after} to avoid timer jiffies wrap (5882)
5937 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5938 - Do not leak request if server was not able to process it (5154)
5939 - If mds_open unable to find parent dir, make that negative lookup(5154)
5940 - don't create new directories with extent-mapping (5909, 5936)
5942 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5943 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5944 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5945 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5946 - support for RHEL4 kernel on i686 (5773)
5947 - provide error messages when incompatible logs are encountered (5898)
5949 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5950 * version 1.4.0.10 (1.4.1 release candidate 1)
5952 - don't keep a lock reference when lock is not granted (4238)
5953 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5954 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5955 - drop import inflight refcount on signal_completed_replay error (5255)
5956 - unlock page after async write error during send (3677)
5957 - handle missing objects in filter_preprw_read properly (5265)
5958 - no transno return for symlink open, don't save no-trasno open (3440)
5959 - don't try to complete elan receive that already failed (4012)
5960 - free RPC server reply state on error (5406)
5961 - clean up thread from ptlrpc_start_thread() on error (5160)
5962 - readahead could read extra page into cache that wasn't ejected (5388)
5963 - prevent races in class_attach/setup/cleanup/detach (5260)
5964 - don't dereference de->d_inode after l_dput of de (5458)
5965 - use "int" for stripe value returned from lock_to_stripe (5544)
5966 - mballoc allocation and error-checking fixes in 2.6 (5504)
5967 - block device patches to fix I/O request sizes in 2.6 (5482)
5968 - look up hostnames for IB nals (5602)
5969 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5970 - don't start multiple acceptors for the same port (5277)
5971 - fix incorrect LASSERT in mds_getattr_name (5635)
5972 - export a proc file for general "ping" checking (5628)
5973 - fix "lfs check" to not block when the MDS is down (5628)
5975 - service request history (4965)
5976 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5977 - create an "evict_client" /proc entry on OSTs, like the MDS has
5978 - fix mount usage message, return errors per mount(8) (5168)
5979 - change grep [] to grep "[]" in tests so they work in more UMLs
5980 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5981 - remove ancient LOV_MAGIC_V0 EA support (5047)
5982 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5983 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5984 - allow readahead allocations to fail when low on memory (5383)
5985 - mmap locking landed again, after considerable improvement (2828)
5986 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5988 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5991 - send OST transaction number in read/write reply to free req (4966)
5992 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5993 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5994 - the watchdog thread now runs as interruptible (5246)
5995 - flock/lockf fixes (but it's still disabled, pending 5135)
5996 - don't use EXT3 constants in llite code (5094)
5997 - memory shortage at startup could cause assertion (5176)
5999 - reorganization of lov code
6000 - single portals codebase
6002 - add extents/mballoc support (5025)
6003 - direct I/O reads in the obdfilter (4048)
6004 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6006 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6009 - send OST transaction number in read/write reply to free req (4966)
6010 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6011 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6012 - the watchdog thread now runs as interruptible (5246)
6013 - handle missing objects in filter_preprw_read properly (5265)
6014 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6015 - drop import inflight refcount on signal_completed_replay error (5255)
6016 - unlock page after async write error during send (3677)
6017 - return original error code on reconstructed replies (3761)
6018 - no transno return for symlink open, don't save no-trasno open (3440)
6020 - add pid to ldlm debugging output (4922)
6021 - bump the watchdog timeouts -- we can't handle 30sec yet
6022 - extra debugging for orphan dentry/inode bug (5259)
6024 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6027 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6028 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6029 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6030 - fix NULL dereference in /proc/sys/portals/routes (4827)
6031 - allow failed mdc_close() operations to be interrupted (4561)
6032 - stop precreate on OST before MDS would time out on it (4778)
6033 - don't send partial-page writes before EOF from client (4410)
6034 - discard client grant for sub-page writes on large-page clients (4520)
6035 - don't free dentries not owned by NFS code, check generation (4806)
6036 - fix lsm leak if mds_create_objects() fails (4801)
6037 - limit debug_daemon file size, always print CERROR messages (4789)
6038 - use transno after validating reply (3892)
6039 - process timed out requests if import state changes (3754)
6040 - update mtime on OST during writes, return in glimpse (4829)
6041 - add mkfsoptions to LDAP (4679)
6042 - use ->max_readahead method instead of zapping global ra (5039)
6043 - don't interrupt __l_wait_event() during strace
6045 - add software watchdogs to catch hung threads quickly (4941)
6046 - make lustrefs init script start after nfs is mounted
6047 - fix CWARN/ERROR duplication (4930)
6048 - return async write errors to application if possible (2248)
6049 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6050 - print NAL number in %x format (4645)
6051 - update barely-supported suse-2.4.21-171 series (4842)
6052 - support for sles 9 %post scripts
6053 - support for building 2.6 kernel-source packages
6054 - support for sles km_* packages
6056 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6059 - ignore -ENOENT errors in osc_destroy (3639)
6060 - notify osc create thread that OSC is being cleaned up (4600)
6061 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6062 - reconstruct ost_handle() like mds_handle() (4657)
6063 - create a new thread to do import eviction to avoid deadlock (3969)
6064 - let lconf resolve symlinked-to devices (4629)
6065 - don't unlink "objects" from directory with default EA (4554)
6066 - hold socknal file ref over connect in case target is down (4394)
6067 - allow more than 32000 subdirectories in a single directory (3244)
6068 - fix blocks count for O_DIRECT writes (3751)
6069 - OST returns ENOSPC from object create when no space left (4539)
6070 - don't send truncate RPC if file size isn't changing (4410)
6071 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6072 - bind to privileged port in socknal and tcpnal (3689)
6074 - rate limit CERROR/CWARN console message to avoid overload (4519)
6075 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6076 - basic mmap support (3918)
6077 - kernel patch series update from b1_4 (4711)
6079 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6082 - avoid crash during MDS cleanup with OST shut down (2775)
6083 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6084 - don't use bad inodes on the MDS (3744)
6085 - dynamic object preallocation to improve recovery speed (4236)
6086 - don't hold spinlock over lock dumping or change debug flags (4401)
6087 - don't zero obd_dev when it is force cleaned (3651)
6088 - print grants to console if they go negative (4431)
6089 - "lctl deactivate" will stop automatic recovery attempts (3406)
6090 - look for existing locks in ldlm_handle_enqueue() (3764)
6091 - don't resolve lock handle twice in recovery avoiding race (4401)
6092 - revalidate should check working dir is a directory (4134)
6094 - don't always mark "slow" obdfilter messages as errors (4418)
6096 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6099 - don't close LustreDB during write_conf until it is done (3860)
6100 - fix typo in lconf for_each_profile (3821)
6101 - allow dumping logs from multiple threads at one time (3820)
6102 - don't allow multiple threads in OSC recovery (3812)
6103 - fix debug_size parameters (3864)
6104 - fix mds_postrecov to initialize import for llog ctxt (3121)
6105 - replace config semaphore with spinlock (3306)
6106 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6107 - don't allow enqueue to complete on a destroyed export (3822)
6108 - down write_lock before checking llog header bitmap (3825)
6109 - recover from lock replay timeout (3764)
6110 - up llog sem before sending rpc (3652)
6111 - reduce ns lock hold times when setting kms (3267)
6112 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6113 - fix NULL deref and obd_dev leak on setup error (3312)
6114 - replace some LBUG about llog ops with error handling (3841)
6115 - don't match INVALID dentries from d_lookup and spin (3784)
6116 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6117 - fix invalid assertion in ptlrpc_set_wait (3880)
6119 - add libwrap support for the TCP acceptor (3996)
6120 - add /proc/sys/portals/routes for non-root route listing (3994)
6121 - allow setting MDS UUID in .xml (2580)
6122 - print the stack of a process that LBUGs (4228)
6124 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6127 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6128 - make sure to unset replay flag from failed open requests (3440)
6129 - if default stripe count is 0, use OST count for inode size (3636)
6130 - update parent mtime/ctime on client for create/unlink (2611)
6131 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6132 - free recovery state on server during a forced cleanup (3571)
6133 - unregister_reply for resent reqs (3063)
6134 - loop back devices mounting and status check on 2.6 (3563)
6135 - fix resource-creation race that can provoke i_size == 0 (3513)
6136 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6137 - more debugging for page-accounting assertion (3746)
6138 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6139 - avoid deadlock after precreation failure (3758)
6140 - fix race and lock order deadlock in orphan handling (3450, 3750)
6141 - add validity checks when grabbing inodes from l_ast_data (3599)
6143 - add /proc/.../recovery_status to obdfilter (3428)
6144 - lightweight CDEBUG infrastructure, debug daemon (3668)
6145 - change default OSC RPC parameters to be better on small clusters
6146 - turn off OST read cache for files smaller than 32MB
6147 - install man pages and include them in rpms (3100)
6148 - add new init script for (un)mounting lustre filesystems (2593)
6149 - run chkconfig in %post for init scripts (3701)
6150 - drop scimac NAL (unmaintained)
6152 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6155 - clean kiobufs before and after use (3485)
6156 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6157 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6158 - change init script to fail more clearly if not run as root (1528)
6159 - allow clients to reconnect during replay (1742)
6160 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6161 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6162 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6163 - fix cond_resched (3554)
6164 - only evict unfinished clients after recovery (3515)
6165 - allow bulk resend, prevent data loss (3570)
6166 - dynamic ptlrpc request buffer allocation (2102)
6167 - don't allow unlinking open directory if it isn't empty (2904)
6168 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6169 - remove extraneous obd dereference causing LASSERT failure (3334)
6170 - don't use get_cycles() when creating temp. files on the mds (3156)
6171 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6172 - handle EEXIST for set-stripe, set proper directory name (3336)
6174 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6175 - fix ksocknal_fmb_callback() error messages (2918)
6177 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6180 - don't copy lvb into (possibly NULL) reply on error (2983)
6181 - don't deref dentry after dput, don't free lvb on error (2922)
6182 - use the kms to determine writeback rpc length (2947)
6183 - increment oti_logcookies when osc is inactive (2948)
6184 - update client's i_blocks count via lvb messages (2543)
6185 - handle intent open/close of special files properly (1557)
6186 - mount MDS with errors=remount-ro, like obdfilter (2009)
6187 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6188 - don't use cancelling-locks' kms values (2947)
6189 - use highest lock extent for kms, not last one (2925)
6190 - don't dereference ERR_PTR() dentry in error handling path (3107)
6191 - fix thread race in portals_debug_dumplog() (3122)
6192 - create lprocfs device entries at setup instead of at attach (1519)
6193 - common AST error handler, don't evict client on completion race (3145)
6194 - zero nameidata in detach_mnt in 2.6 (3118)
6195 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6196 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6197 - initialize RPC timeout timer earlier for 2.6 (3219)
6198 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6199 - print nal/nid for unknown nid (3258)
6200 - additional checks for oscc recovery before doing precreate (3284)
6201 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6202 - don't crash in mdc_close for bad permissions on open (3285)
6203 - zero i_rdev for non-device files (3147)
6204 - clear page->private before handing to FS, better assertion (3119)
6205 - tune the read pipeline (3236)
6206 - fix incorrect decref of invalidated dentry (2350)
6207 - provide read-ahead stats and refine rpc in flight stats (3328)
6208 - don't hold journal transaction open across create RPC (3313)
6209 - update atime on MDS at close time (3265)
6210 - close LDAP connection when recovering to avoid server load (3315)
6211 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6212 - don't leak open file on MDS after open resend (3325)
6213 - serialize filter_precreate and filter_destroy_precreated (3329)
6214 - loop device shouldn't call sync_dev() for nul device (3092)
6215 - clear page cache after eviction (2766)
6216 - resynchronize MDS->OST in background (2824)
6217 - refuse to mount the same filesystem twice on same mountpoint (3394)
6218 - allow llmount to create routes for mounting behind routers (3320)
6219 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6220 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6221 - fix rare problem with rename on htree directories (3417)
6223 - allow default OST striping configuration per directory (1414)
6224 - fix compilation for qswnal for 2.6 kernels (3125)
6225 - increase maximum number of MDS request buffers for large systems
6226 - change liblustreapi to be useful for external progs like lfsck (3098)
6227 - increase local configuration timeout for slow disks (3353)
6228 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6230 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6233 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6234 - don't overwrite extent policy data in reply if lock was blocked (2901)
6235 - drop filter export grants atomically with removal from device (2663)
6236 - del obd_self_export from work_list in class_disconnect_exports (2908)
6237 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6238 - swab reply message in mdc_close, other PPC fixes (2464)
6239 - fix destroying of named logs (2325)
6240 - overwrite old logs when running lconf --write_conf (2264)
6241 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6242 - fix race in target_handle_connect (2898)
6243 - mds_reint_create() should take same inode create lock (2926)
6244 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6245 - don't close files for self_export to avoid uninitialized obd (2936)
6246 - allow MDS with the same name as client node (2939)
6247 - hold dentry reference for closed log files for unlink (2325)
6248 - reserve space for all logs during transactions (2059)
6249 - don't evict page beyond end of stripe extent (2925)
6250 - don't oops on a deleted current working directory (2399)
6251 - handle hard links to targets without a parent properly (2517)
6252 - don't dereference NULL lock when racing during eviction (2867)
6253 - don't grow lock extents when lots of conflicting locks (2919)
6255 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6258 - account for cache space usage on clients to avoid data loss (974)
6259 - lfsck support in lustre kernel code (2349)
6260 - reduce journal credits needed for BRW writes (2370)
6261 - orphan handling to avoid losing space on client/server crashes
6262 - ptlrpcd can be blocked, stopping ALL progress (2477)
6263 - use lock value blocks to assist in proper KMS, faster stat (1021)
6264 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6265 - recovery for initial connections (2355)
6266 - fixes for mds_cleanup_orphans (1934)
6267 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6268 - block all file creations until orphan recovery completes (1901)
6269 - client remove rq_connection from request struct (2423)
6270 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6271 - recovery timer race (2670)
6272 - mdc_close recovey bug (2532)
6273 - ptlrpc cleanup bug (2710)
6274 - mds timeout on local locks (2588)
6275 - namespace lock held during RPCs (2431)
6276 - handle interrupted sync write properly (2503)
6277 - don't try to handle a message that hasn't been replied to (2699)
6278 - client assert failure during cleanup after abort recovery (2701)
6279 - leak mdc device after failed mount (2712)
6280 - ptlrpc_check_set allows timedout requests to complete (2714)
6281 - wait for inflight reqs when ptlrpcd finishes (2710)
6282 - make sure unregistered services are removed from the srv_list
6283 - reset bulk XID's when resending them (caught by 1138 test)
6284 - unregister_bulk after timeout
6285 - fix lconf error (2694)
6286 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6287 - readahead locks pages, leaves pending causing memory pressure (2673)
6288 - increase OST request buffers to 4096 on large machines (2729)
6289 - fix up permission of existing directories in simple_mkdir (2661)
6290 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6291 - don't assign transno to errored transactions (2742)
6292 - don't delete objects on OST if given a bogus objid from MDS (2751)
6293 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6294 - if rq_no_resend, then timeout request after recovery (2432)
6295 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6296 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6297 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6298 - MDS recovery shouldn't delete objects that it has given out (2730)
6299 - if enqueue arrives after completion, don't clobber LVB (2819)
6300 - don't unlock pages twice when trigger_group_io returns error (2814)
6301 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6302 - don't write pages to disk if there was an error (1450)
6303 - don't ping imports that have recovery disabled (2676)
6304 - take buffered bytes into account when balancing socknal conn (2817)
6305 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6306 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6307 - remove little-endian swabbing of llog records (1987)
6308 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6309 - retry reposting request buffers if they fail (1191)
6310 - grow extent at grant time to avoid granting a revoked lock (2809)
6311 - lock revoke doesn't evict page if covered by a second lock (2765)
6312 - disable VM readahead to avoid reading outside lock extents (2805)
6314 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6315 - updated kernel patches for hp-2.4.20 kernel (2681)
6317 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6320 - fix truncated write corruption (2366)
6321 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6323 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6324 - ensure len(uuid) < 37 in lmc (1171)
6325 - fix ia64 OOPS in llog_test (2255)
6326 - zero end of page at obdfilter for partial page writes (2648)
6327 - don't leave stale dentries around after renames (bug 2428)
6328 - fix timeouts when evicting a client with a single lock held (2642)
6329 - set deadline for the initial HELLO message to drain (2634)
6330 - print out dotted-quad IP addresses in the socknal (2302)
6332 - additional debugging for MDS client eviction problem (2443)
6333 - fix mkfsoptions support for osts (2603, 2604)
6335 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6338 - add series for the vanilla 2.6.0 kernel
6339 - add series for the vanilla 2.4.24 kernel
6340 - add series for a cray x86/64 UL kernel drop
6341 - fix xattr patches for the vanilla 2.4.19 series
6343 - generate true UUIDs in lmc (1171)
6344 - have portals stack dumping break in UML (2466)
6345 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6346 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6347 - wake callback waiting threads on client eviction (2460)
6348 - Add --ptldebug and --subsystem to lmc (1719)
6349 - update assertion to allow safe interrupt allocation
6350 - set rq_no_resend for cancel requests (2432)
6351 - recalculate ptlrpcd timeout after resend (2494)
6352 - call vfs_rmdir when removing pending directories (2368)
6353 - fix renaming a file to itself (2429)
6354 - lmc creates a default one-stripe lov (2454)
6355 - expand procfs space to handle large clusters (2326)
6356 - increase UML stack to avoid overflow
6357 - update lconf's list of debug and subsystem masks
6358 - fix lfs find --obd (2510)
6359 - /proc tunable for disabling filter read caching (2591)
6360 - stop rpm packages from altering slapd.conf (2301)
6361 - disable nagle in the socknal under 0conf (2578)
6362 - choose mds inode size based on stripe count (2572)
6363 - fix kernel-source rpm problems (2516)
6365 - add --disable-doc to avoid pdf generation (2421)
6366 - update documentation, tests, type-os, comments
6367 - avoid format warnings on ia64
6368 - remove the TOE NAL
6369 - tiny code cleanups by removing unused fields
6371 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6374 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6375 - remove the most blatant lies from BUILDING file (2371)
6376 - change default debug level to reasonable production setting
6377 - reduce client side cache size to reduce cache flush time
6378 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6379 - make TCP zerocopy and pinger support enabled by default (2476)
6380 - sync writes completed after process exits caused crashes (2319)
6381 - maintain correct mount count on the MDS (2356)
6382 - backout 1557, because 2316 wasn't really fixed
6383 - better file I/O statistics gathering in /proc
6384 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6385 - another kernel patch to fix zero-copy TCP function export
6386 - don't take duplicate lock when processing re-sent getattr (2420)
6387 - lctl uses obd_self_export instead of creating new conn (2353)
6388 - MDS/OST recovery case which requires object creation asserted (2425)
6389 - move lfs from /usr/sbin to /usr/bin in packages
6390 - fix race between mds_client_add and mds_client_free (2417)
6391 - use kmalloc instead of slabs in portals (2430)
6392 - don't create duplicate records when a failover MDS is present (2442)
6393 - remove unnecessary mount age check (2332)
6394 - don't remove directory inodes from locks prematurely (2451)
6395 - don't break if MDS service name is the same as hostname (2103)
6396 - fix races in client write RPC generation when cache full (2482)
6398 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6401 - remove now-unused request->rq_obd (278)
6402 - if an allocation fails, print out how much memory we've used (1933)
6403 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6404 - add the "configurable stack size" patch to most series files (1256)
6405 - ability to write large log records, for 100+ OST configs (2306)
6406 - fix NULL deref when filter_prep fails (2314)
6407 - fix operator precedence error in filter_sync
6408 - dynamic allocation of socknal TX descriptors (2315)
6409 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6410 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6411 - ensure that CWARN messages in Portals always get to the syslog
6412 - __init/__exit are not for prototype decls (ldlm_init/exit)
6413 - x86-64 compile warning fixes
6414 - fix gateway LMC keyword conflict (2318)
6415 - fix MDS lock inversions in getattr/reint paths (1844)
6416 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6417 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6418 - DLM race condition prevented some lock evictions (2328)
6419 - ENOMEM detection and retry on socknal sends (2230)
6420 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6421 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6422 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6423 - better error messages when a client is rejected during recovery (1505)
6424 - avoid cancelling locks which were never granted, after failure (2330)
6425 - fix i_sem/journal inversion in mds_client_add (2333)
6426 - fix truncate/getattr lock cycle deadlock (2334)
6427 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6428 - fix two rare exit paths which could leak an l_lock() ref (2321)
6429 - fencepost error in MDS/OST orphan recovery (2226)
6430 - make log record alignment 8 bytes (1988)
6431 - lstripe now fails when requested offset > ost_count (2237)
6432 - ensure that all kernel series have a complete list.h (1607)
6433 - fix crashes in special-file operations (2316)
6434 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6436 - allow configurable automake binary, for testing new versions
6437 - small update to the lfs documentation
6439 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6441 * fix negative export reference count in fsfilt_sync (2312)
6443 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6444 * release candidate 0.9.1
6446 - orphans are moved into the PENDING directory for possible recovery
6447 - replayed opens now open by fid for orphan/rename safety (1042)
6448 - last close of an orphan inode generates a transno (683)
6449 - chdir() and mount() now pin the directory entry (1020)
6450 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6451 - discard very old requests without processing them (1502)
6452 - remove some common, well-understood CERRORs (1505)
6453 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6454 - clear "grant" flags in OST replies until OST grant code lands (1644)
6455 - fix read performance by not clobbering i_blksize on client (1598)
6456 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6457 - make LRU size a /proc tunable, clears locks when reduced (707)
6458 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6459 - server locks take references on exports now (1558)
6460 - build fixes for 2.4.20-rh trees (1663)
6461 - return an error from lov_create if all OSCs are inactive (1751)
6462 - fix import levels when a reconnect happens without a timeout (1597)
6463 - exit early from mds_open if we get a lookup error (1749)
6464 - partial page read at EOF wouldn't wait for disk before sending (1642)
6465 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6466 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6467 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6468 - write/truncate lock inversion (1639)
6469 - Don't auto-load obdclass, portals modules during cleanup (1495)
6470 - fix timestamps from jumping to "now" (1763)
6471 - extra journal assertions (1648)
6472 - add an extra multiunlink test (1771)
6473 - fix read_record/write_record API (1776)
6474 - fix leak of offset_extent, possible incorrect i_size later (1772)
6475 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6476 - Debugging for the kqswnal_get_idle_tx problems (1820)
6477 - Allow recovery to be attempted multiple times (1536)
6478 - Write out MDS last_rcvd file after it is first created (1600)
6479 - Fix tx_descriptor leak in failed transmit situations (1827)
6480 - ext3 journaling fixes for assertion failure after IO error (1871)
6481 - class_export_put() on freed export after completion AST error (1896)
6482 - Fix revalidate looping in VFS (1322)
6483 - Don't access a freed export during MDS_REINT timeout (1521)
6484 - Add open-unlink recovery support on the MDS (1673,1764)
6485 - Return an error if no MDS data was read from last_rcvd (1946)
6486 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6487 - Don't setup a disk device that doesn't match exported UUID (317)
6488 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6489 - avoid committing NULL handle in force close
6490 - local.sh is now a one-stripe LOV configuration
6491 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6492 - fix ext3 htree duplicate directory entry corruption (1516)
6493 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6494 - update to new LOV EA format (2097)
6495 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6497 2003-06-15 Phil Schwan <phil@clusterfs.com>
6500 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6501 - per-import/export recovery handling (958, 931, 959)
6502 - multiple last-rcvd slots, for serving multiple FSes (949)
6503 - connections are again shared between multiple imp/exports (963, 964)
6504 - "umount -f" would hang if any requests needed to be sent (393, 978)
6505 - avoid pinning large req buffer by copying for queued messages (989)
6506 - add "uuid" to "lctl device" command to help upcalls (991)
6507 - "open" RPCs with transnos would confuse recovery counters (1037)
6508 - do proper endian conversion of all wire messages (288, 340, 891)
6509 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6510 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6511 - fix ptlrpc_connection leak in target_handle_connect (1174)
6512 - fix import refcounting bug in OST and MDS cleanup (1134)
6513 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6514 - fix very unlikely obd_types race condition (501)
6515 - remove osc_open hack for echo_client (1187)
6516 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6517 - a failure in read_inode2 leads to deadlock (1139)
6518 - cancel ack-locks as soon as transaction is committed (1072)
6519 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6520 - make sure to commitrw after any preprw to avoid deadlock (1162)
6521 - failing to execute a file in a lustre FS would lock inode (1203)
6522 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6523 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6524 - obd_uuid2tgt would walk off the end of the list (1255)
6525 - on IA64 the osc would give portals incorrect bulk size (1258)
6526 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6527 - fix lock inversion caused by new llite matching code (1282)
6528 - limit the number of dirty pages on a client to 10MB (1286)
6529 - timed out locks were not being corrected cancelled (1289)
6530 - fix O_DIRECT above 4GB on IA-32 (1292)
6531 * major user-visible changes
6532 - fail out/fail over policy now controlled by the upcall (993)
6534 - add OBD_PING to check server availability and failure (954)
6535 - lustre messages are now sent in sending host order (288, 340, 891)
6536 - add eadatalen to MDS getattr reply (340)
6537 - OST read replies may contain second buffer, with per-page status (593)
6539 2003-03-11 Phil Schwan <phil@clusterfs.com>
6542 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6543 - failed open()s could cause deadlock; fixed (867, 869)
6544 - stop cancelling OST locks when files are closed (481)
6545 - overlapping XID spaces caused network corruption (851, 853)
6546 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6547 - setattr_raw added, to do single-RPC, server-side setattrs
6548 - lmc/lconf syntax change for OST UUIDs
6549 - fix crashy race condition between ptlrpc_free_req and osc_close
6550 - don't use request in mdc_enqueue if we hit a timeout (889)
6551 - don't set the inode i_size for regular files from the MDS (896)
6552 - handle out of order completion AST (842)
6553 - don't LBUG if a lock request times out after receiving AST (913)
6554 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6555 - if a bad lock AST arrives, send an error instead of dropping entirely
6556 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6557 - fix leak in bulk IO when only partially completed (899, 900, 926)
6558 - fix O_DIRECT for ia64 (55)
6559 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6560 - C-z after timeout could hang a process forever; fixed (977)
6562 - client-side I/O cache (678, 924, 929, 941, 970)
6564 - READPAGE and SETATTRs which don't take server-side locks get
6567 2003-02-11 Phil Schwan <phil@clusterfs.com>
6570 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6571 - Fix fsfilt_extN_readpage() to read a full page of directory
6572 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6573 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6574 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6575 - Most intent operations were removed, in favour of a new RPC mode
6576 that does a single RPC to the server and bypasses most of the VFS
6577 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6578 - Aggressively cancel local locks on DLM servers
6579 - mds_reint_unlink sends EA to the client if it's the last nlink.
6580 client uses that EA to unlink OST objects.
6581 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6582 - recursive symlinks were fixed (439)
6583 - fixed NULL deref in DEBUG_REQ
6584 - filter_update_lastobjid no longer calls sync, which annoyed extN
6585 - fixed multi-client small-writes to a single file problem (445)
6586 - fixed mtime updates during file writes (607)
6587 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6588 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6589 - fixed Linux OST size reporting problem (444, 656)
6590 - OST now updates object mtime with writes or setattr (607, 619)
6591 - client verifies file size before zeroing page past EOF (445)
6592 - OST now writes last allocated objid to disk with allocation (108)
6593 - LOV on echo now works (409)
6595 - mds_reint_unlink sends a new buffer, with the EA included. this
6596 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6597 set if a regular file was being unlinked, and it was the last link
6598 - use PtlGet from the target for bulk writes (315)
6599 - OST now updates object mtime with writes or setattr (607, 619)
6600 - LDLM now has a grant-time callback to revalidate locked items, if
6602 - Many MDS operations were reorganized to combat race conditions
6604 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6605 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6607 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6610 - Fully reactivate OST imports after reconnection (512, others)
6611 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6612 - More graceful error handling for truncating on dead OST (515)
6613 - Don't error out unless we're actually accessing dead stripes (474)
6614 - Fix garbage sizes when stripes are missing (410)
6615 - LRU counters were broken, causing constant lock purge (433, 432)
6616 - garbage on read from stripes with failed OSTs (441)
6617 - mark OSCs as active before reconnecting during recovery (438)
6618 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6619 - lfind did not preserve OST order in output (443)
6620 - symlinks cause hung clients, incorrect data (439)
6621 - stop dereferencing request after dropping refcount (457)
6622 - don't LASSERT(spin_is_locked) on non-SMP (455)
6623 - fixes for many rename() bugs
6624 - fstat didn't correctly synchronize attributes (399)
6625 - server must handle lock cancellation during blocking AST prep (487)
6626 - bulk descriptors were free()d too soon (511)
6627 - fix paths in lconf, which would load incorrect modules (451, 507)
6628 - fix confusing lconf 'host not found' error message (386)
6629 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6630 - fix race condition in mdc_blocking_ast() for inode access (526)
6631 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6632 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6633 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6634 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6635 - fix lstripe in several ways (kernel side) (527)
6636 - fix request leak in ldlm_cli_enqueue (262)
6637 - incorrect OSC was marked inactive after OST failure
6638 - call mds_fs_cleanup before unmounting filesystem (524)
6639 - fix races between taking ns_lock and ldlm_lock_change_resource
6640 - fix races updating LOV export open file list
6641 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6642 - fix recovery NULL deref in ldlm_cli_cancel_unused
6643 - fix some DLM races by using new hash table for lock handles (419)
6644 - permit the client to specify desired inodes, at replay
6645 - duplicate requests when we queue them for replay reintegration
6646 - fix last_rcvd offset calculation
6647 - sync after each recovered transaction, so we always make progress
6648 - never, not always, ERESTART requests without transnos
6649 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6651 - skip replay if the MDS says that the client is already connected
6652 - don't check for a recovery-enabled export to match lctl's UUID
6653 - don't INC_USE_COUNT for phantom exports
6654 - don't crash when cleaning up phantom exports (567)
6655 - don't double-finish or set replay data for errored mdc_open requests
6656 - abort requests when they time out, so we don't get old replies
6657 - send/receive replies for AST messages again
6658 - if the client says that it doesn't have the lock, cancel it on the
6660 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6661 mark it as destroyed, it will all work out when decref is called
6662 - fix module use counts (22, 581)
6664 - ASTs now expect a reply (server cancels lock on error reply)
6666 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6669 - fix many simultaneous client startup (392)
6670 - fix dentry->d_it clobbering
6671 - credentials weren't being shipped for readdir/getattr operations
6672 - remove invalid assertions triggered during some concurrent MD
6674 - proper Lustre versions added (336, 389)
6675 - fix memory leak for create error case (398)
6676 - fix LOV locking bug that would get cli/srv out of sync
6677 - fix echo client over LOV (409)
6678 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6679 - fix double-O_EXCL intent crash (424)
6680 - avoid sending multiple lock CANCELs (352)
6682 - MDS can do multi-client recovery (modulo bugs in new code)
6684 - many updates, edits, cleanups
6686 2002-11-18 Phil Schwan <phil@clusterfs.com>
6689 - fix null d_it dereference (346)
6690 - fix full OST/dbench hang (333)
6691 - fix permission problem with file removal (286)
6692 - fix removal of OSCs from LOV when they fail
6693 - fix NULL deref during bulk timeout (214)
6694 - fix problems related to multiple filesystems on one MDS (241)
6695 - fixed serious subtle metadata locking bugs
6696 - free locks on clients when inodes are removed due to memory
6698 - fix inode pointer in lock data (285)
6699 - partial support for multiple MDS on a single host (241)
6700 - data locks weren't cancelled at clear_inode time (290, 311)
6701 - intent locks could lead to unbounded lock growth (205)
6702 - added a maximum lock count, an LRU list, and a flusher
6703 - fix multiple rename (365)
6704 - properly abstracted the echo client
6705 - OSC locked 1 byte too many; fixed
6706 - rewrote brw callback code:
6707 - fixed recovery bugs related to LOVs (306)
6708 - fixed too-many-pages-in-one-write crash (191)
6709 - fixed (again) crash in sync_io_timeout (214)
6710 - probably fixed callback-related race (385)
6712 - Add capability to MDS protocol
6713 - LDLM cancellations and callbacks on different portals
6715 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6718 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6719 - defer open object destruction to close time (601981, 138)
6720 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6721 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6722 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6723 - don't call obd_setattr in ll_file_release for destroyed objects
6724 * protocol change to lustre_msg: move |version| and add |flags|
6725 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6726 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6727 * added replay of create, unlink, link and rename operations during
6728 MDS failover; recovery should be much more robust now
6729 * remove failed OSCs from LOVs (only lov_create uses this so far)
6730 * the lustre-HOWTO was brought (more) up to date (582544)
6732 2002-10-23 Phil Schwan <phil@clusterfs.com>
6735 - in-use dentries weren't being reused properly (617851)
6736 - prevent multiple LDLM setup (599178)
6737 - fix LOV size calculations for truncate (617853)
6738 - fix client handling of MDS intent errors (POSIX)
6739 - fix permission bug in lovstripe.c test (624321)
6740 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6741 - truncate past end of file could corrupt data
6742 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6743 - a race in recovery could return ETIMEDOUT to apps (623947)
6744 - building outside the source directory was fixed
6745 * the lustre-HOWTO was brought (more) up to date (582544)
6746 * major progress was made on recovery functionality
6748 2002-10-10 Phil Schwan <phil@clusterfs.com>
6751 - recovery deadlock fix
6752 - rm -rf causes LBUG fix (617817)
6753 - file open by multiple tasks fix (618962)
6754 - directory permissions bugs (602707 and 620007)
6755 - journal_stop fixed with locking (611313)
6756 - O_APPEND failures resolved (618273, perhaps 614459)
6757 - lconf PATH fix (619770)
6758 - IA64 build fix (621450)
6759 - RPC buffer sizes scale with amount of memory
6761 2002-10-01 Phil Schwan <phil@clusterfs.com>
6764 - locks would be cancelled without throwing away data pages,
6765 resulting in inconsistent data (605627)
6766 - inode attributes were not always being refreshed (605627, 612449)
6767 - lconf now continues to cleanup after lctl reports an error
6768 - MDS now enforces user permissions (602707)
6769 - lprocfs cleanup fixed, but not yet enabled (614157)
6770 - fixed infinite server hang, should a client not respond to an AST
6771 - avoid going into recovery if user calls readlink() with a buffer
6772 that's too small (613941)
6773 - AST RPCs no longer require replies (614867) -- this may be changed
6774 - don't crash server if client sends an IOV that's too big (611336)
6775 - fixed lock conversion deadlock (611892)
6776 - fixed the following of symlinks (614622)
6777 * recovery: the server can remove locks from a client that dies, other
6778 clients can make progress
6779 * more extN patch fixes
6780 * compile-time configurable ptlrpc buffer allocations
6782 - collaborative read cache document
6783 - Lustre Lite Performance CDR document-in-progress
6785 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6788 - fix typo in patch-2.4.18
6790 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6793 - clear ptlrpc request each time in handle_incoming_request()
6794 - unlink of files now destroys the object on the OST
6796 2002-09-19 Peter Braam <braam@clusterfs.com>
6798 * add hard link support
6799 * change obdfile creation method
6800 * kernel patch changed
6802 2002-09-19 Peter Braam <braam@clusterfs.com>
6805 - stack overflow bug in extN fixed
6807 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6809 * documentation updates
6810 - add man pages for config tools
6811 - update tests/README to describe testing with new config tools
6812 - finish metadata API descriptions
6813 * bug fixes and cleanups
6814 - statfs workaround for 16TB limit
6815 - LOV stripe allocation improved, can stripe on subset of OSTs
6816 - LOV file size/IO offset was wrong for files > 4GB in size
6817 - object EA data was being dropped, caused files to be unreadable
6818 - memory overflow with non-LOV OST caused memory corruption
6819 - fixed regression tests to work with new config tools, obdfilter
6820 - fixed bug when directory size became larger than 1 block
6821 - fixed bug (for single client case) when PWD was deleted
6822 - invalidate local directory pages when doing intent-based ops
6823 - avoid LDLM oops when lock callback contained bad data
6825 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6827 * documentation updates
6828 * bug fixes and cleanups
6829 - configuration tools
6832 - 64-bit compile warnings
6833 - 64-bit internal statfs data
6835 * test_brw on persistent OST devices
6837 * lprocfs (disabled)
6839 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6841 * documentation updates
6842 * bug fixes and cleanups
6843 * configuration tools
6845 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6848 * many small fixes to 0_5_4
6849 * io/network handling
6850 * thinkos in MDS operations
6852 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6855 * crucial basic fixes to 0.5.3
6856 * IOR, Iozone work over Elan
6859 2002-08-07 Phil Schwan <phil@clusterfs.com>
6860 * version 0_5_3, our first alpha
6861 * we use the new Portals iovs
6862 * documentation updates
6863 * bug fixes and cleanups
6864 * small changes in the DLM wire protocol
6866 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6867 * version 0_5_1 with some initial stability,
6868 * locking on MD and file I/O.
6869 * documentation updates
6870 * several bug fixes since 0.5.0
6871 * small changes in wire protocol
6873 2002-07-18 Phil Schwan <phil@clusterfs.com>
6875 * delivered as Lustre Light Alpha
6876 * fixed a crash after handling invalid MDS requests
6877 * fixed directory pages for architectures with non-4k pages sizes
6879 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6880 * release version v0_4_4
6881 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6883 * Fixes a number of interruption problems with OST operations.
6884 * Update documentation for portals header changes
6885 * Move all wire protocol structs/defines to lustre_idl.h
6886 * Fixes symlink length bug.
6887 * Add tcpdump to repository.
6889 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6890 * release version v0_4_3
6891 * Fixes statfs for inodes on extN.
6892 * Fixes bug in runtests which would delete /etc/hosts.
6893 * Use 64-bit object IDs wherever possible (not into VFS though)
6894 Remove ost_get_info, which is unused by lustre, and out of date.
6896 2002-07-03 Peter Braam <braam@clusterfs.com>
6897 * release version v0_4_2 Fixes a lookup error (type not passed)
6898 * move forward to head of Portals
6899 * move forward to latest Lustre kernel
6901 2002-06-25 Peter Braam <braam@clusterfs.com>
6902 * release version v0_4_1. Hopefully stable on single node use.