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 4.18.0-372.9.1.el8 (RHEL8.6)
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 4.18.0-348.2.1.el8 (RHEL8.5)
19 4.4.120-92.70 (SLES12 SP2)
20 4.4.180-94.100 (SLES12 SP3)
21 4.4.0-131 (Ubuntu 16.04)
22 4.15.0-32 (Ubuntu 18.04)
23 5.4.0-48 (Ubuntu 20.04)
24 vanilla linux 5.4.0 (ZFS + ldiskfs)
25 vanilla linux 5.4.21 (ZFS + ldiskfs)
26 vanilla linux 5.4.136 (ZFS + ldiskfs)
27 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
28 * Client primary kernels built and tested during release cycle:
29 5.14.0-70.22.1.el9 (RHEL9.0)
30 4.18.0-372.9.1.el8 (RHEL8.6)
31 5.4.0-37 (Ubuntu 20.04)
32 5.14.21-150400.24.18 (SLES15 SP4)
33 * Other clients known to build on these kernels at some point (others may also work):
34 3.10.0-862.14.4.el7 (RHEL7.5)
35 3.10.0-957.27.2.el7 (RHEL7.6)
36 4.14.0-49.13.1.el7a (RHEL7.5)
37 4.14.0-115.2.2.el7a (RHEL7.6)
38 3.10.0-1062.18.1.el7 (RHEL7.7)
39 3.10.0-1127.19.1.el7 (RHEL7.8)
40 3.10.0-1160.76.1.el7 (RHEL7.9)
41 4.18.0-80.11.2.el8 (RHEL8.0)
42 4.18.0-147.8.1.el8 (RHEL8.1)
43 4.18.0-193.28.1.el8 (RHEL8.2)
44 4.18.0-240.22.1.el8 (RHEL8.3)
45 4.18.0-305.25.1.el8 (RHEL8.4)
46 4.18.0-348.2.1.el8 (RHEL8.5)
47 4.4.120-92.70 (SLES12 SP2)
48 4.4.180-94.100 (SLES12 SP3)
49 4.12.14-95.48 (SLES12 SP4)
50 4.12.14-122.91 (SLES12 SP5)
51 4.12.14-197.75 (SLES15 SP1)
52 5.3.18-24.96 (SLES15 SP2)
53 5.3.18-59.27 (SLES15 SP3)
54 4.4.0-131 (Ubuntu 16.04)
55 4.15.0-48 (Ubuntu 18.04)
56 5.8.0-53 (Ubuntu 20.04.2 HWE)
57 5.11.0-31 (Ubuntu 20.04.3 HWE)
58 5.11.0 (vanilla kernel.org)
59 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
60 * Recommended ZFS version: 2.1.2
61 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
62 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
63 stack overflow. For more information, please refer to bugzilla 17630.
64 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
65 the re-exporting nodes to be mounted with "32bitapi" mount option
67 --------------------------------------------------------------------------------
71 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
72 for currently supported client and server kernel versions.
73 * Server primary kernels built and tested during release cycle:
74 4.18.0-372.9.1.el8 (RHEL8.6)
75 * Other server kernels known to build and work at some point (others may also work):
76 3.10.0-862.14.4.el7 (RHEL7.5)
77 3.10.0-957.27.2.el7 (RHEL7.6)
78 3.10.0-1062.18.1.el7 (RHEL7.7)
79 3.10.0-1127.19.1.el7 (RHEL7.8)
80 3.10.0-1160.71.1.el7 (RHEL7.9)
81 4.18.0-80.11.2.el8 (RHEL8.0)
82 4.18.0-147.8.1.el8 (RHEL8.1)
83 4.18.0-193.28.1.el8 (RHEL8.2)
84 4.18.0-240.22.1.el8 (RHEL8.3)
85 4.18.0-305.25.1.el8 (RHEL8.4)
86 4.18.0-348.2.1.el8 (RHEL8.5)
87 4.4.120-92.70 (SLES12 SP2)
88 4.4.180-94.100 (SLES12 SP3)
89 4.4.0-131 (Ubuntu 16.04)
90 4.15.0-32 (Ubuntu 18.04)
91 5.4.0-48 (Ubuntu 20.04)
92 vanilla linux 5.4.0 (ZFS + ldiskfs)
93 vanilla linux 5.4.21 (ZFS + ldiskfs)
94 vanilla linux 5.4.136 (ZFS + ldiskfs)
95 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
96 * Client primary kernels built and tested during release cycle:
97 4.18.0-372.9.1.el8 (RHEL8.6)
98 5.4.0-37 (Ubuntu 20.04)
99 5.3.18-59.27 (SLES15 SP3)
100 * Other clients known to build on these kernels at some point (others may also work):
101 3.10.0-862.14.4.el7 (RHEL7.5)
102 3.10.0-957.27.2.el7 (RHEL7.6)
103 4.14.0-49.13.1.el7a (RHEL7.5)
104 4.14.0-115.2.2.el7a (RHEL7.6)
105 3.10.0-1062.18.1.el7 (RHEL7.7)
106 3.10.0-1127.19.1.el7 (RHEL7.8)
107 3.10.0-1160.71.1.el7 (RHEL7.9)
108 4.18.0-80.11.2.el8 (RHEL8.0)
109 4.18.0-147.8.1.el8 (RHEL8.1)
110 4.18.0-193.28.1.el8 (RHEL8.2)
111 4.18.0-240.22.1.el8 (RHEL8.3)
112 4.18.0-305.25.1.el8 (RHEL8.4)
113 4.18.0-348.2.1.el8 (RHEL8.5)
114 4.4.120-92.70 (SLES12 SP2)
115 4.4.180-94.100 (SLES12 SP3)
116 4.12.14-95.48 (SLES12 SP4)
117 4.12.14-122.91 (SLES12 SP5)
118 4.12.14-197.75 (SLES15 SP1)
119 5.3.18-24.96 (SLES15 SP2)
120 4.4.0-131 (Ubuntu 16.04)
121 4.15.0-48 (Ubuntu 18.04)
122 5.8.0-53 (Ubuntu 20.04.2 HWE)
123 5.11.0-31 (Ubuntu 20.04.3 HWE)
124 5.11.0 (vanilla kernel.org)
125 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
126 * Recommended ZFS version: 2.1.5
127 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
128 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
129 stack overflow. For more information, please refer to bugzilla 17630.
130 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
131 the re-exporting nodes to be mounted with "32bitapi" mount option
133 --------------------------------------------------------------------------------
137 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
138 for currently supported client and server kernel versions.
139 * Server primary kernels built and tested during release cycle:
140 4.18.0-372.9.1.el8 (RHEL8.6)
141 * Other server kernels known to build and work at some point (others may also work):
142 3.10.0-862.14.4.el7 (RHEL7.5)
143 3.10.0-957.27.2.el7 (RHEL7.6)
144 3.10.0-1062.18.1.el7 (RHEL7.7)
145 3.10.0-1127.19.1.el7 (RHEL7.8)
146 3.10.0-1160.49.1.el7 (RHEL7.9)
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 4.18.0-240.22.1.el8 (RHEL8.3)
151 4.18.0-305.25.1.el8 (RHEL8.4)
152 4.18.0-348.2.1.el8 (RHEL8.5)
153 4.4.120-92.70 (SLES12 SP2)
154 4.4.180-94.100 (SLES12 SP3)
155 4.4.0-131 (Ubuntu 16.04)
156 4.15.0-32 (Ubuntu 18.04)
157 5.4.0-48 (Ubuntu 20.04)
158 vanilla linux 5.4.0 (ZFS + ldiskfs)
159 vanilla linux 5.4.21 (ZFS + ldiskfs)
160 vanilla linux 5.4.136 (ZFS + ldiskfs)
161 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
162 * Client primary kernels built and tested during release cycle:
163 4.18.0-372.9.1.el8 (RHEL8.6)
164 5.4.0-37 (Ubuntu 20.04)
165 5.3.18-59.27 (SLES15 SP3)
166 * Other clients known to build on these kernels at some point (others may also work):
167 3.10.0-862.14.4.el7 (RHEL7.5)
168 3.10.0-957.27.2.el7 (RHEL7.6)
169 4.14.0-49.13.1.el7a (RHEL7.5)
170 4.14.0-115.2.2.el7a (RHEL7.6)
171 3.10.0-1062.18.1.el7 (RHEL7.7)
172 3.10.0-1127.19.1.el7 (RHEL7.8)
173 3.10.0-1160.49.1.el7 (RHEL7.9)
174 4.18.0-80.11.2.el8 (RHEL8.0)
175 4.18.0-147.8.1.el8 (RHEL8.1)
176 4.18.0-193.28.1.el8 (RHEL8.2)
177 4.18.0-240.22.1.el8 (RHEL8.3)
178 4.18.0-305.25.1.el8 (RHEL8.4)
179 4.18.0-348.2.1.el8 (RHEL8.5)
180 4.4.120-92.70 (SLES12 SP2)
181 4.4.180-94.100 (SLES12 SP3)
182 4.12.14-95.48 (SLES12 SP4)
183 4.12.14-122.91 (SLES12 SP5)
184 4.12.14-197.75 (SLES15 SP1)
185 5.3.18-24.96 (SLES15 SP2)
186 4.4.0-131 (Ubuntu 16.04)
187 4.15.0-48 (Ubuntu 18.04)
188 5.8.0-53 (Ubuntu 20.04.2 HWE)
189 5.11.0-31 (Ubuntu 20.04.3 HWE)
190 5.11.0 (vanilla kernel.org)
191 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
192 * Recommended ZFS version: 2.1.2
193 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
194 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
195 stack overflow. For more information, please refer to bugzilla 17630.
196 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
197 the re-exporting nodes to be mounted with "32bitapi" mount option
199 --------------------------------------------------------------------------------
203 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
204 for currently supported client and server kernel versions.
205 * Server primary kernels built and tested during release cycle:
206 3.10.0-1062.18.1.el7 (RHEL7.7)
207 3.10.0-1127.19.1.el7 (RHEL7.8)
208 3.10.0-1160.6.1.el7 (RHEL7.9)
209 * Other server kernels known to build and work at some point (others may also work):
210 3.10.0-862.14.4.el7 (RHEL7.5)
211 3.10.0-957.27.2.el7 (RHEL7.6)
212 4.18.0-80.11.2.el8 (RHEL8.0)
213 4.18.0-147.8.1.el8 (RHEL8.1)
214 4.18.0-193.28.1.el8 (RHEL8.2)
215 4.18.0-240.1.1.el8 (RHEL8.3)
216 4.4.120-92.70 (SLES12 SP2)
217 4.4.180-94.100 (SLES12 SP3)
218 4.4.0-131 (Ubuntu 16.04)
219 4.15.0-32 (Ubuntu 18.04)
220 5.4.0-48 (Ubuntu 20.04)
221 vanilla linux 5.4.0 (ZFS + ldiskfs)
222 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
223 * Client primary kernels built and tested during release cycle:
224 3.10.0-1062.18.1.el7 (RHEL7.7)
225 3.10.0-1127.19.1.el7 (RHEL7.8)
226 3.10.0-1160.6.1.el7 (RHEL7.9)
227 4.12.14-95.48 (SLES12 SP4)
228 4.12.14-122.20 (SLES12 SP5)
229 4.12.14-197.75 (SLES15 SP1)
230 5.3.18-24.24 (SLES15 SP2)
231 4.15.0-48 (Ubuntu 18.04)
232 * Other clients known to build on these kernels at some point (others may also work):
233 3.10.0-862.14.4.el7 (RHEL7.5)
234 3.10.0-957.27.2.el7 (RHEL7.6)
235 4.14.0-49.13.1.el7a (RHEL7.5)
236 4.14.0-115.2.2.el7a (RHEL7.6)
237 4.18.0-80.11.2.el8 (RHEL8.0)
238 4.18.0-147.8.1.el8 (RHEL8.1)
239 4.18.0-193.28.1.el8 (RHEL8.2)
240 4.18.0-240.1.1.el8 (RHEL8.3)
241 4.4.120-92.70 (SLES12 SP2)
242 4.4.180-94.100 (SLES12 SP3)
243 4.4.0-131 (Ubuntu 16.04)
244 5.4.0-37 (Ubuntu 20.04)
245 5.4.0 (vanilla kernel.org)
246 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
247 * Recommended ZFS version: 2.0.0
248 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
249 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
250 stack overflow. For more information, please refer to bugzilla 17630.
251 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
252 the re-exporting nodes to be mounted with "32bitapi" mount option
254 --------------------------------------------------------------------------------
258 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
259 for currently supported client and server kernel versions.
260 * Server primary kernels built and tested during release cycle:
261 3.10.0-1062.1.1.el7 (RHEL7.7)
262 * Other server kernels known to build and work at some point (others may also work):
263 3.10.0-862.14.4.el7 (RHEL7.5)
264 3.10.0-957.27.2.el7 (RHEL7.6)
265 4.18.0-80.7.1.el8 (RHEL8.0)
266 4.4.120-92.70 (SLES12 SP2)
267 4.4.180-94.100 (SLES12 SP3)
268 4.4.0-131 (Ubuntu 16.04)
269 4.15.0-32 (Ubuntu 18.04)
270 vanilla linux 4.6.7 (ZFS only)
271 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
272 * Client primary unpatched kernels built and tested during release cycle:
273 3.10.0-1062.1.1.el7 (RHEL7.7)
274 4.12.14-95.29 (SLES12 SP4)
275 4.15.0-48 (Ubuntu 18.04)
276 * Other clients known to build on unpatched kernels at some point (others may also work):
277 3.10.0-862.14.4.el7 (RHEL7.5)
278 3.10.0-957.27.2.el7 (RHEL7.6)
279 4.14.0-49.13.1.el7a (RHEL7.5)
280 4.14.0-115.2.2.el7a (RHEL7.6)
281 4.18.0-80.7.1.el8 (RHEL8.0)
282 4.4.120-92.70 (SLES12 SP2)
283 4.4.180-94.100 (SLES12 SP3)
284 4.4.0-131 (Ubuntu 16.04)
285 vanilla linux 4.15.0 (ZFS only)
286 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
287 * Recommended ZFS / SPL version: 0.7.13
288 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
289 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
290 stack overflow. For more information, please refer to bugzilla 17630.
291 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
292 the re-exporting nodes to be mounted with "32bitapi" mount option
293 * Note for Ubuntu 14.04.5:
294 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
295 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
297 --------------------------------------------------------------------------------
301 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
302 for currently supported client and server kernel versions.
303 * Server known to build on the following kernels (others may also work):
304 2.6.32-696.18.7.el6 (RHEL6.9)
305 3.10.0-862.14.4.el7 (RHEL7.5)
306 3.10.0-957.el7 (RHEL7.6)
307 4.4.120-92.70 (SLES12 SP2)
308 4.4.162-94.69 (SLES12 SP3)
309 4.4.0-131 (Ubuntu 16.04)
310 4.15.0-32 (Ubuntu 18.04)
311 vanilla linux 4.6.7 (ZFS only)
312 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
313 * Client known to build on unpatched kernels (others may also work):
314 2.6.32-696.18.7.el6 (RHEL6.9)
315 2.6.32-754.6.3.el6 (RHEL6.10)
316 3.10.0-862.14.4.el7 (RHEL7.5)
317 3.10.0-957.el7 (RHEL7.6)
318 4.14.0-49.13.1.el7a (RHEL7.5)
319 4.14.0-115.2.2.el7a (RHEL7.6)
320 4.4.120-92.70 (SLES12 SP2)
321 4.4.162-94.69 (SLES12 SP3)
322 4.4.0-131 (Ubuntu 16.04)
323 4.15.0-32 (Ubuntu 18.04)
324 vanilla linux 4.15.0 (ZFS only)
325 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
326 * Recommended ZFS / SPL version: 0.7.9
327 * Tested with ZFS / SPL version: 0.7.9
328 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
329 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
330 stack overflow. For more information, please refer to bugzilla 17630.
331 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
332 the re-exporting nodes to be mounted with "32bitapi" mount option
333 * Note for Ubuntu 14.04.5:
334 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
335 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
337 --------------------------------------------------------------------------------
339 2018-04-03 Intel Corporation
341 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
342 for currently supported client and server kernel versions.
343 * Server known to build on patched kernels:
344 2.6.32-431.29.2.el6 (RHEL6.5)
345 2.6.32-504.30.3.el6 (RHEL6.6)
346 2.6.32-573.26.1.el6 (RHEL6.7)
347 2.6.32-642.15.1.el6 (RHEL6.8)
348 2.6.32-696.18.7.el6 (RHEL6.9)
349 3.10.0-693.21.1.el7 (RHEL7.4)
350 3.0.101-0.47.71 (SLES11 SP3)
351 3.0.101-107 (SLES11 SP4)
352 3.12.74-60.64.40 (SLES12 SP1)
353 4.4.114-92.64 (SLES12 SP2)
354 4.4.114-94.11 (SLES12 SP3)
355 3.13.0-101 (Ubuntu 14.04, ZFS only)
356 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
357 4.4.0-116 (Ubuntu 16.04)
358 vanilla linux 4.6.7 (ZFS only)
359 * Client known to build on unpatched kernels:
360 2.6.32-431.29.2.el6 (RHEL6.5)
361 2.6.32-504.30.3.el6 (RHEL6.6)
362 2.6.32-573.26.1.el6 (RHEL6.7)
363 2.6.32-642.15.1.el6 (RHEL6.8)
364 2.6.32-696.18.7.el6 (RHEL6.9)
365 3.10.0-693.21.1.el7 (RHEL7.4)
366 3.0.101-0.47.71 (SLES11 SP3)
367 3.0.101-107 (SLES11 SP4)
368 3.12.74-60.64.40 (SLES12 SP1)
369 4.4.114-92.64 (SLES12 SP2)
370 4.4.114-94.11 (SLES12 SP3)
371 3.13.0-101 (Ubuntu 14.04)
372 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
373 4.4.0-116 (Ubuntu 16.04)
375 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
376 * Recommended ZFS / SPL version: 0.7.6
377 * Tested with ZFS / SPL version: 0.7.6
378 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
379 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
380 stack overflow. For more information, please refer to bugzilla 17630.
381 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
382 the re-exporting nodes to be mounted with "32bitapi" mount option
383 * Note for Ubuntu 14.04.5:
384 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
385 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
387 --------------------------------------------------------------------------------
389 2017-07-05 Intel Corporation
391 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
392 for currently supported client and server kernel versions.
393 * Server known to build on patched kernels:
394 2.6.32-431.29.2.el6 (RHEL6.5)
395 2.6.32-504.30.3.el6 (RHEL6.6)
396 2.6.32-573.26.1.el6 (RHEL6.7)
397 2.6.32-642.15.1.el6 (RHEL6.8)
398 2.6.32-696.3.1.el6 (RHEL6.9)
399 3.10.0-514.21.1.el7 (RHEL7.3)
400 3.0.101-0.47.71 (SLES11 SP3)
401 3.0.101-100 (SLES11 SP4)
402 3.12.74-60.64.40 (SLES12 SP1)
403 4.4.59-92.17 (SLES12 SP2)
404 vanilla linux 4.6.7 (ZFS only)
405 * Client known to build on unpatched kernels:
406 2.6.32-431.29.2.el6 (RHEL6.5)
407 2.6.32-504.30.3.el6 (RHEL6.6)
408 2.6.32-573.26.1.el6 (RHEL6.7)
409 2.6.32-642.15.1.el6 (RHEL6.8)
410 2.6.32-696.3.1.el6 (RHEL6.9)
411 3.10.0-514.21.1.el7 (RHEL7.3)
412 3.0.101-0.47.71 (SLES11 SP3)
413 3.0.101-100 (SLES11 SP4)
414 3.12.74-60.64.40 (SLES12 SP1)
415 4.4.59-92.17 (SLES12 SP2)
417 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
418 * Recommended ZFS / SPL version: 0.6.5.9
419 * Tested with ZFS / SPL version: 0.6.5.9
420 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
421 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
422 stack overflow. For more information, please refer to bugzilla 17630.
423 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
424 the re-exporting nodes to be mounted with "32bitapi" mount option
426 --------------------------------------------------------------------------------
428 12-07-2016 Intel Corporation
430 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
431 for currently supported client and server kernel versions.
432 * Server known to build on patched kernels:
433 2.6.32-431.29.2.el6 (RHEL6.5)
434 2.6.32-504.30.3.el6 (RHEL6.6)
435 2.6.32-573.26.1.el6 (RHEL6.7)
436 2.6.32-642.11.1.el6 (RHEL6.8)
437 3.10.0-514.el7 (RHEL7.3)
438 3.0.101-0.47.71 (SLES11 SP3)
439 3.0.101-80 (SLES11 SP4)
440 3.12.62-60.64.8 (SLES12 SP1)
441 vanilla linux 4.5.7 (ZFS only)
442 * Client known to build on unpatched kernels:
443 2.6.32-431.29.2.el6 (RHEL6.5)
444 2.6.32-504.30.3.el6 (RHEL6.6)
445 2.6.32-573.26.1.el6 (RHEL6.7)
446 2.6.32-642.11.1.el6 (RHEL6.8)
447 3.10.0-514.el7 (RHEL7.3)
448 3.0.101-0.47.71 (SLES11 SP3)
449 3.0.101-80 (SLES11 SP4)
450 3.12.62-60.64.8 (SLES12 SP1)
452 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
453 * Recommended ZFS / SPL version: 0.6.5.8
454 * Tested with ZFS / SPL version: 0.6.5.8
455 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
456 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
457 stack overflow. For more information, please refer to bugzilla 17630.
458 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
459 the re-exporting nodes to be mounted with "32bitapi" mount option
461 --------------------------------------------------------------------------------
463 02-29-2016 Intel Corporation
465 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
466 for currently supported client and server kernel versions.
467 * Server known to build on patched kernels:
468 2.6.32-431.29.2.el6 (RHEL6.5)
469 2.6.32-504.30.3.el6 (RHEL6.6)
470 2.6.32-573.12.1.el6 (RHEL6.7)
471 3.10.0-327.3.1.el7 (RHEL7.2)
472 3.0.101-0.47.71 (SLES11 SP3)
473 3.0.101-68 (SLES11 SP4)
474 vanilla linux 4.2.1 (ZFS only)
475 * Client known to build on unpatched kernels:
476 2.6.32-431.29.2.el6 (RHEL6.5)
477 2.6.32-504.30.3.el6 (RHEL6.6)
478 2.6.32-573.12.1.el6 (RHEL6.7)
479 3.10.0-327.3.1.el7 (RHEL7.2)
480 3.0.101-0.47.71 (SLES11 SP3)
481 3.0.101-68 (SLES11 SP4)
484 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
485 * Recommended ZFS / SPL version: 0.6.4.2
486 * Tested with ZFS / SPL version: 0.6.4.2
487 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
488 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
489 stack overflow. For more information, please refer to bugzilla 17630.
490 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
491 the re-exporting nodes to be mounted with "32bitapi" mount option
493 --------------------------------------------------------------------------------
495 03-10-2015 Intel Corporation
497 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
498 for currently supported client and server kernel versions.
499 * Server known to build on patched kernels:
500 2.6.32-431.29.2.el6 (RHEL6.5)
501 2.6.32-504.8.1.el6 (RHEL6.6)
502 3.0.101-0.46 (SLES11 SP3)
503 * Client known to build on unpatched kernels:
504 2.6.32-431.29.2.el6 (RHEL6.5)
505 2.6.32-504.8.1.el6 (RHEL6.6)
506 3.10.0-123.20.1.el7 (RHEL7)
507 3.0.101-0.46 (SLES11 SP3)
508 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
509 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
510 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
511 stack overflow. For more information, please refer to bugzilla 17630.
512 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
513 the re-exporting nodes to be mounted with "32bitapi" mount option
515 Severity : enhancement
517 Description: control OST-index in IDIF via ROCOMPAT flag.
518 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
519 last_rcvd file. For new formatted OST device, it will be auto set;
520 for the case of upgrading from old OST device, you can enable it
521 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
522 enabled, for new created OST-object, its IDIF-in-LMA will contain
523 the OST-index; for the existing OST-object, the OSD will convert
524 old format IDIF as new format IDIF with OST-index stored in the
525 LMA EA when accessing such OST-object or via OI scrub. Once such
526 flag is enabled, it cannot be reverted back, so the system cannot
527 be downgraded to the orignal incompatible version.
529 --------------------------------------------------------------------------------
531 07-30-2014 Intel Corporation
533 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
534 for currently supported client and server kernel versions.
535 * Server known to build on patched kernels:
536 2.6.32-431.20.3.el6 (RHEL6.5)
537 3.0.101-0.31 (SLES11 SP3)
538 * Client known to build on unpatched kernels:
539 2.6.32-431.20.3.el6 (RHEL6.5)
540 3.0.101-0.31 (SLES11 SP3)
543 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
544 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
545 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
546 stack overflow. For more information, please refer to bugzilla 17630.
547 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
548 the re-exporting nodes to be mounted with "32bitapi" mount option
550 --------------------------------------------------------------------------------
552 10-31-2013 Intel Corporation
554 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
555 * Server builds for kernels:
556 2.6.32-358.18.1.el6 (RHEL6.4)
557 * Client builds for unpatched kernels:
558 2.6.32-358.18.1.el6 (RHEL6.4)
559 3.0.92-0.8 (SLES11 SP3)
561 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
562 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
563 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
564 stack overflow. For more information, please refer to bugzilla 17630.
565 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
566 the re-exporting nodes to be mounted with "32bitapi" mount option
568 --------------------------------------------------------------------------------
570 05-30-2013 Intel Corporation
572 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
573 * Server builds for kernels:
574 2.6.32-358.6.2.el6 (RHEL6.4)
575 * Client builds for unpatched kernels:
576 2.6.32-358.6.2.el6 (RHEL6.4)
577 3.0.74-0.6.6 (SLES11 SP2)
579 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
580 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
581 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
582 stack overflow. For more information, please refer to bugzilla 17630.
583 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
584 the re-exporting nodes to be mounted with "32bitapi" mount option
586 --------------------------------------------------------------------------------
588 9-30-2012 Whamcloud, Inc.
590 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
591 * Server support for kernels:
592 2.6.32-279.5.1.el6 (RHEL6)
593 * Client support for unpatched kernels:
594 2.6.32-279.5.1.el6 (RHEL6)
596 2.6.18-238.19.1.el5 (RHEL5)
597 2.6.16.60-0.69.1 (SLES10)
598 * Recommended e2fsprogs version: 1.41.90.wc4
599 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
600 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
601 stack overflow. For more information, please refer to bugzilla 17630.
602 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
603 the re-exporting nodes to be mounted with "32bitapi" mount option
605 Severity : enhancement
607 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
609 Severity : enhancement
611 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
613 Severity : enhancement
615 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
617 Severity : enhancement
619 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
621 Severity : enhancement
623 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
625 --------------------------------------------------------------------------------
627 04-30-2012 Whamcloud, Inc.
629 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
630 * Server support for kernels:
631 2.6.32-220.4.2.el6 (RHEL6)
632 * Client support for unpatched kernels:
633 2.6.32-220.4.2.el6 (RHEL6)
635 2.6.18-238.19.1.el5 (RHEL5)
636 2.6.16.60-0.69.1 (SLES10)
637 * Recommended e2fsprogs version: 1.41.90.wc4
638 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
639 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
640 stack overflow. For more information, please refer to bugzilla 17630.
641 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
642 the re-exporting nodes to be mounted with "32bitapi" mount option
644 Severity : enhancement
646 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
648 Severity : enhancement
650 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
652 Severity : enhancement
654 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
656 Severity : enhancement
658 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
660 --------------------------------------------------------------------------------
662 09-30-2011 Whamcloud, Inc.
664 * Server support for kernels:
665 2.6.18-238.19.1.el5 (RHEL5)
666 2.6.32-131.6.1.el6 (RHEL6)
667 * Client support for unpatched kernels:
668 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
669 2.6.18-238.19.1.el5 (RHEL5)
670 2.6.32-131.6.1.el6 (RHEL6)
671 2.6.29.4-167.fc11 (FC11)
672 2.6.16.60-0.69.1 (SLES10)
674 * Recommended e2fsprogs version: 1.41.90.wc3
675 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
676 Lustre filesystem with 4K stack may cause a stack overflow. For more
677 information, please refer to bugzilla 17630.
678 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
679 mounted with "32bitapi" mount option
681 Severity : enhancement
682 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
684 Severity : enhancement
685 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
687 Severity : enhancement
688 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
691 Frequency : only with SLES10
693 Description: Use OFED "KMP" provided by Novell
694 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
695 Lustre is now built against this package. That means you need to
696 install the ofed-kmp package from Novell for the patchless client.
697 Note that the ofed-kmp that Novell ships may not exactly match the
698 kernel version but should still be compatible.
700 Severity : enhancement
702 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
703 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
705 --------------------------------------------------------------------------------
707 2010-07-15 Oracle, Inc.
709 * Server support for kernels:
710 2.6.18-164.11.1.el5 (RHEL 5)
711 2.6.18-164.11.1.0.1.el5 (OEL 5)
712 * Client support for unpatched kernels:
713 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
714 2.6.18-164.11.1.el5 (RHEL 5),
715 2.6.18-164.11.1.0.1.el5 (OEL 5)
716 2.6.16.60-0.42.8 (SLES 10),
718 2.6.29.4-167.fc11 (FC11)
719 * Recommended e2fsprogs version: 1.41.10-sun2
721 Severity : enhancement
723 Description: Distribute lfs_migrate script with lustre
724 Details : The lfs_migrate mentioned in the manual is now included in the
725 lustre tree in lustre/scripts and will distributed in lustre RPMs.
727 Severity : enhancement
729 Description: Failover nid list added to import proc file.
731 Severity : enhancement
733 Description: "som_preview" mount option required to use SOM
734 Details : The clients will need to use som_preview mount option
738 Description: Lustre does not do 1MB IOs to HW RAID
739 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
741 Severity : enhancement
743 Description: lfs quota output is not very convenient for awk/sed-parsing
744 Details : Some positions in lfs quota output table could be empty or
745 non-empty which made it hard to parse it with scripts, now
746 a dash is put instead of space where there is not supposed
751 Description: Update to OFED 1.5.1
752 Details : OFED 1.5.1 will be used in environments that don't use
753 vendor supplied OFED.
757 Description: Remove lnet. prefix from lctl params display
758 Details : Also renamed lnet.memused to lnet_memused to
763 Description: Use conf_param -d to remove permanant settings.
767 Description: Add list_param -R
768 Details : lctl list_param now supports recursive mode.
771 Bugzilla : 21829, 22312
772 Description: Fix llobdstat.
773 Details : 1. Now updating the write_bytes counter
774 2. Port script fixes from 1.8.
778 Description: decrease the usage of memory on clients.
779 Details : 1. On clients, recycle dentries and inodes unused.
780 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
785 Description: llobdstat fix and enhancement
786 Details : add a counter to set a limit to how many samples will be returned
787 fix a wildcard in the path to limit to obdfilter stats only
791 Description: lfs check only allowed for root
792 Details : Code cleanups and added a regresion test of non-root use of
797 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
798 Details : Fix hr->hr_index overload.
802 Description: Fix FIEMAP functionality.
806 Description: FC11 patchless client support added
810 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
811 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
815 Description: New acc-sm script and various test-framework enhancments.
819 Description: Previously landed patch was reverted.
823 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
827 Description: Add ldiskfs maxdirsize mount option.
831 Description: Fix mkfs.lustre support for 16TB LUNs.
835 Description: Print mmp_check_interval
836 Details : print mmp_check_interval and make it possible to abort
837 mount operation in case it takes too long.
841 Description: Add cascading_rw to lustre mpi tests.
845 Description: Store SOM data in EA.
846 Details : Also several SOM related cleanups.
850 Description: Add refcount to md_open_data
854 Description: Use refcount on lov_request_set to prevent memory corruption.
858 Description: Statahead: use parent dir's i_mutex for synchronization.
862 Description: loadgen improvements.
866 Description: Enable peer health detection in ksocklnd and ko2iblnd.
870 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
874 Description: Fix race on obd_devcie:md_stats
878 Description: Fix SLES 11 client builds.
879 Details : Resolved uninitalized pointer gcc warnings.
883 Description: Allow non-root access for "lfs check".
884 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
888 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
890 Severity : enhancement
892 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
893 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
897 Description: Sanity test_27p fails.
898 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
899 Sanity tests 27 were cleaned up and improved.
903 Description: Unify req format on client/servers
904 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
905 buffers and swabbers approach..
910 Description: drop join file stuff.
912 Severity : enhancement
914 Description: Add 2.6.27 and SLES11 (patchless client) support
918 Description: Fix in ptlrpc_expire_one_request() to print the signed time
921 Severity : enhancement
923 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
926 Severity : enhancement
928 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
930 Severity : enhancement
932 Description: correctly handle big reply message.
933 Details : send LNet event if reply is bigger than buffer and adjust this buffer
938 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
940 Severity : enhancement
942 Description: Add support for OEL5.
946 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
948 Severity : enhancement
950 Description: "weak-modules" support
951 Details : Implement "weak-modules" support which enables kernel modules
952 to be used with any kernel that implements the same kABI. In
953 order to achieve this modules are now installed in
954 /lib/modules/$(uname -r)/updates/kernel on all distributions.
958 Description: client could not reconnect to OST because of an active request.
959 Details : The client now retries to reconnect to the same server, if a
960 connect request failed with EBUSY or -EAGAIN. Test case 35b
961 conf-sanity.sh has been created.
965 Description: Don't allow make backward step on assignin osc next id.
966 Details : race between allocation next id and ll_sync thread can be cause
967 of set wrong osc next id and can be kill valid ost objects.
971 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
975 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
979 Description: Error handling in osc_statfs_interpret() has been improved.
980 Details : Check in osc_statfs_interpret() for EBADR.
984 Description: Do not send statfs() requests to OSTs disabled by administrator.
985 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
987 Severity : enhancement
989 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
991 Severity : enhancement
993 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
995 Severity : enhancement
997 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1001 Description: Improved cleanup procedure for conf-sanity test_46a.
1005 Description: Do not start bulk transfer, if client tried to reconnect.
1006 Details : Do not start bulk transfer, if client tried to reconnect,
1007 but failed due to active rpcs. Check for exp_abort_active_req
1008 has been implemented.
1012 Description: Prevent inconsistences between linux and lustre mount structures.
1013 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1014 Make the sleep interruptible.
1016 Severity : enhancement
1018 Description: provide server to client comms path
1019 Details : server to client communications path via new LDLM_SET_INFO rpc
1023 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1024 Details : ldiskfs_truncate calls grab_cache_page which may start page
1025 allocation under an open transaction. This may lead to
1026 calling prune_icache with consequent lustre reentrance.
1030 Description: send correctly lsm on open replay
1031 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1036 Description: enable client interface failover
1037 Details : When a child reconnects from another NID, properly update export
1038 nid hash position and ldlm reverse import.
1040 Severity : enhancement
1042 Description: Communicate OST degraded/readonly state via statfs to MDS
1043 Details : Flags in the statfs returned from OSTs indicate whether the
1044 OST is in a degraded RAID state, or if the filesystem has
1045 turned read-only after a filesystem error is detected.
1050 Description: don't panic if EPROTO was hit when reading symlink
1051 Details : correctly handling request reference in error cases.
1053 Severity : enhancement
1055 Description: Add LustreNetLink, a kernel-userspace communcation path.
1057 Severity : enhancement
1059 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1064 Description: access to llog context before init.
1065 Details : move handling CATALOGS file at osc layer and forbid access to llog
1066 context before init.
1069 Frequency : rare, only when client is writing to OST on local node
1071 Description: Avoid deadlock for local client writes
1072 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1073 memory freeing context. This allows OST threads to set the
1074 PF_MEMALLOC flag on task structures in order to allocate memory
1075 from reserved pools and complete IO.
1076 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1077 so that the OST threads generate memory pressure and allow
1078 inactive pages to be reclaimed.
1080 Severity : enhancement
1082 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1087 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1088 Details : move ldlm namespace creation in setup phase to avoid grab
1089 _lprocfs_lock with cli_sem held.
1093 Description: Temporarily disable grant shrink.
1094 Details : Disable the feature for debugging.
1097 Frequency : only when formatting test filesystems
1099 Description: Unable to run several mkfs.lustre on loop devices at the same time
1100 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1101 formatting. The solution is to proper handle the error.
1103 Severity : enhancement
1105 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1107 Severity : enhancement
1109 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1114 Description: resolve race between obd_disconnect and class_disconnect_exports
1115 Details : if obd_disconnect will be called to already disconnected export he
1116 forget release one reference and osc module can't unloaded.
1118 Severity : enhancement
1120 Description: Allow tuning service thread via /proc
1121 Details : For each service a new
1122 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1123 created that can be used to set min/max thread counts, and get the
1124 current number of running threads.
1129 Descriptoin: don't return error if have partially created objects for file.
1130 Details : lov_update_create_set uses set->set_success as index for created
1131 objects, so if some requests will be failed, they will have hole
1132 at end of array and qos_shrink_lsm for allocate correct lsm.
1134 Severity : enhancement
1136 Description: Update OFED support to 1.4.1
1141 Description: Reduce small size read RPC
1142 Details : Set read-ahead limit for every file and only do read-ahead when
1143 available read-ahead pages are bigger than 1M to avoid small size
1146 Severity : enhancement
1148 Description: move AT tunable parameters for more consistent usage
1149 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1151 Severity : enhancement
1153 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1154 Details : allow skip disconnected ost for send statfs request and hide error
1160 Description: fix racy locking of mballoc block bitmaps causing BUG
1161 Details : The locking of the mballoc buddy bitmap and the in-memory
1162 block bitmap was using two different spin locks in some
1163 cases. This made it possible to incorrectly access the
1164 mballoc bitmap while another process was modifying it,
1165 causing a sanity assertion to fail. While no on-disk corruption
1166 was reported, there was some risk of this happening.
1169 Frequency : rare, on llog test 6
1171 Descriptoin: don't allow connect to already connected import
1172 Details : allowing connect to already connected import is hide connecting problem.
1177 Descriptoin: allow kill process which wait statahead result
1178 Details : for some reasons 'ls' can stick in waiting result from statahead,
1179 in this case need way for kill this process.
1181 Severity : enhancement
1183 Description: Add state history info file, enhance import info file
1184 Details : Track import connection state changes in a new osc/mdc proc file;
1185 add overview-type data to the osc/mdc import proc file.
1187 Severity : enhancement
1189 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1191 Severity : enhancement
1193 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1198 Descriptoin: OST grant too much space to client even there are not enough space.
1199 Details : Client will shrink its grant cache to OST if there are no write
1200 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1201 this grant cache if there are already not enough avaible space
1202 (left_space < total_clients * 32M).
1205 Frequency : start MDS on uncleanly shutdowned MDS device
1207 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1208 Details : stay in waiting mds<>ost recovery finished produce random bugs
1209 due race between two ll_sync thread for one lov target. send
1210 ACTIVATE event only if connect realy finished and import have
1214 Frequency : rare, connect and disconnect target at same time
1216 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1217 Details : don't call obd_disconnect under lov_lock. this long time
1218 operation and can block ptlrpcd which answer to connect request.
1223 Descriptoin: don't lose wakeup for imp_recovery_waitq
1224 Details : recover_import_no_retry or invalidate_import and import_close can
1225 both sleep on imp_recovery_waitq, but we was send only one wakeup
1229 Frequency : always with long access acl
1231 Descriptoin: mds can't pack reply with long acl.
1232 Details : mds don't control size of acl but they limited by reint/getattr
1235 Severity : enhancement
1237 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1239 Severity : enhancement
1241 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1244 Frequency : start MDS on uncleanly shutdowned MDS device
1246 Descriptoin: aborting recovery hang on MDS
1247 Details : don't throttle destroy RPCs for the MDT.
1250 Frequency : on remount
1252 Description: external journal device not working after the remount
1253 Details : clear dev_rdonly flag for external journal devices in
1259 Description: shutdown vs evict race
1260 Details : client_disconnect_export vs connect request race.
1261 if client will evicted at this time - we start invalidate
1262 thread without referece to import and import can be freed
1266 Frequency : rare, need acl's on inode.
1268 Description: client can't handle ost additional correctly
1269 Details : if ost was added after client connected to mds client can have
1270 hit lnet_try_match_md ... to big messages to wide striped files.
1271 in this case need teach client to handle config events about add
1272 lov target and update client max ea size at that event.
1274 Severity : enhancement
1276 Description: Changelogs
1277 Details : Changelogs are a lightweight mechanism to track filesystem
1278 metadata and namespace changes. The changelog is recorded
1279 permanently on the MDTs, and is periodically "consumed" / purged
1280 when records are no longer needed.
1282 Severity : enhancement
1284 Description: compact fld format with extents
1285 Details : Store range of seq rather than every seq in FLD. Seq
1286 controller update FLD rather than clients. In Case of CMD, mdt0
1287 has FLD, all other metadata server act as non persistent proxy
1288 for FLD queries and cache fld entries in fld cache.
1293 Description: don't skip ost target if they assigned to file
1294 Details : Drop slow OSCs if we can, but not for requested start idx.
1295 This means "if OSC is slow and it is not the requested
1296 start OST, then it can be skipped, otherwise skip it only
1297 if it is inactive/recovering/out-of-space.
1301 Description: more cleanup in mds_lov
1302 Details : not send LOV EA under replay, we can't know about they size at this
1303 time. Don't allow client connect to mds before any ost connected,
1304 for avoid problems with LOV EA size and returning EIO to client.
1306 Severity : enhancement
1308 Description: Interoperability at server side (Disk interoperability)
1310 Severity : enhancement
1312 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1314 Severity : enhancement
1316 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1318 Severity : enhancement
1320 Description: New client IO stack (CLIO).
1322 Severity : enhancement
1324 Description: Commit on sharing. Eliminate inter-client dependencies between
1325 uncommitted transactions by doing transaction commits.
1326 Thereby clients may recovery independently.
1329 Frequency : Create a symlink file with a very long name
1331 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1332 Details : If there is no extra space in the request for early cancels,
1333 ldlm_req_handles_avail() returns 0 instead of a negative value.
1335 Severity : enhancement
1337 Description: Add /proc entry for import status
1338 Details : The mdc, osc, and mgc import directories now have
1339 an import directory that contains useful import data for debugging
1340 connection problems.
1342 Severity : enhancement
1344 Description: Re-disable certain /proc logging
1345 Details : Enable and disable client's offset_stats, extents_stats and
1346 extents_stats_per_process stats logging on the fly.
1349 Frequency : Only on FC kernels 2.6.22+
1351 Description: oops in statahead
1352 Details : Do not drop reference count for the dentry from VFS when lookup,
1353 VFS will do that by itself.
1355 Severity : enhancement
1357 Description: Generic /proc file permissions
1358 Details : Set /Proc file permissions in a more generic way to enable non-
1359 root users operate on some /proc files.
1363 Description: Hitting mdc_commit_close() ASSERTION
1364 Details : Properly handle request reference release in
1365 ll_release_openhandle().
1369 Description: quota recovery deadlock during mds failover
1370 Details : This patch includes att18982, att18236, att18237 in bz14840.
1372 1. fix osts hang when mds does failover with quotaon
1373 2. prevent watchdog storm when osts threads wait for the
1378 Frequency : only patchless client
1379 Description: add workaround for race between add/remove dentry from hash
1381 Severity : enhancement
1383 Description: Allow OST glimpses to return PW locks
1387 Description: LBUG when llog conf file is full
1388 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1392 Description: Prevent import from entering FULL state when server in recovery
1396 Description: service mount cannot take device name with ":"
1397 Details : Only when device name contains ":/" will mount treat it as
1403 Description: replace ptlrpcd with the statahead thread to interpret the async
1404 statahead RPC callback
1408 Frequency : on recovery
1409 Description: I/O failures after umount during fail back
1410 Details : if client reconnected to restarted server we need join to recovery
1411 instead of find server handler is changed and process self
1412 eviction with cancel all locks.
1414 Severity : enhancement
1416 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1418 Severity : enhancement
1420 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1422 Severity : enhancement
1424 Description: Upcall on Lustre log has been dumped
1425 Details : Allow for a user mode script to be called once a Lustre log has
1426 been dumped. It passes the filename of the dumped log to the
1427 script, the location of the script can be specified via
1428 /proc/sys/lnet/debug_log_upcall.
1433 Description: avoid idr_remove called for id which is not allocated.
1434 Details : Move assigment s_dev for clustered nfs to end of initialization,
1435 to avoid problem with error handling.
1440 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1441 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1442 into CONN_USED_HASH and this prodice warning when put connection
1443 again in unused hash.
1446 Severity : enhancement
1448 Description: File striping can now be set to use an arbitrary pool of OSTs.
1450 Severity : enhancement
1452 Description: Export bytes_read/bytes_write count on OSC/OST.
1456 Description: Early reply size mismatch, MGC loses connection
1457 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1458 the connect flags are properly negotiated.
1463 Description: quotas are not honored with O_DIRECT
1464 Details : all writes with the flag O_DIRECT will use grants which leads to
1465 this problem. Now using OBD_BRW_SYNC to guard this.
1469 Description: add quota statistics
1470 Details : 1. sort out quota proc entries and proc code.
1471 2. add quota statistics
1473 Severity : enhancement
1475 Description: enable quota support for HEAD.
1479 Description: Properly propagate oinfo flags from lov to osc for statfs
1480 Details : restore missing copy oi_flags to lov requests.
1482 Severity : enhancement
1484 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1485 : l_getgroups(8), lst(8), routerstat(8)
1486 Details : included man pages for llobdstat(8), llstat(8),
1487 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1489 Severity : enhancement
1491 Description: configure's --enable-quota should check the
1492 : kernel .config for CONFIG_QUOTA
1493 Details : configure is terminated if --enable-quota is passed but
1494 : no quota support is in kernel
1498 Description: Remove portals compatibility
1499 Details : Remove portals compatibility, not interoperable with releases
1504 Description: Resolve device initialization race
1505 Details : Prevent proc handler from accessing devices added to the
1506 obd_devs array but yet be intialized.
1508 Severity : enhancement
1510 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1512 Severity : enhancement
1514 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1519 Description: Using wrong pointer in osc_brw_prep_request
1520 Details : Access to array[-1] can produce panic if kernel compiled with
1521 CONFIG_PAGE_ALLOC enabled
1525 Description: Client runs out of low memory
1526 Details : Consider only lowmem when counting initial number of llap pages
1530 Description: *optional* service tags registration
1531 Details : if the "service tags" package is installed on a Lustre node
1532 When the filesystem is mounted, a local-node service tag will
1533 be created. See http://inventory.sun.com/ for more information
1534 about the Service Tags asset management system.
1538 Description: Kernel BUG tries to release flock
1539 Details : Lustre does not destroy flock lock before last reference goes
1540 away. So always drop flock locks when client is evicted and
1541 perform unlock regardless of successfulness of speaking to MDS.
1545 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1549 Description: sanity test 65a fails if stripecount of -1 is set
1550 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1555 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1556 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1559 Severity : enhancement
1561 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1565 Description: do not process already freed flock
1566 Details : flock can possibly be freed by another thread before it reaches
1567 to ldlm_flock_completion_ast.
1571 Description: LBUG during stress test
1572 Details : Need properly lock accesses the flock deadlock detection list.
1576 Description: oops in page fault handler
1577 Details : kernel page fault handler can return two special 'pages' in error
1578 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1582 Description: timeout with invalidate import.
1583 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1584 should be handled by ptlrpcd. This produce long age waiting and
1585 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1587 Severity : enhancement
1589 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1593 Description: enable MGS and MDT services start separately
1594 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1595 a co-located MGS without starting the MGS, which is a complement
1596 to 'nosvc' mount option.
1600 Frequency : after recovery
1601 Description: precreate to many object's after del orphan.
1602 Details : del orphan st in oscc last_id == next_id and this triger growing
1603 count of precreated objects. Set flag LOW to skip increase count
1604 of precreated objects.
1608 Frequency : rare, on clear nid stats
1609 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1610 Details : when clean nid stats sometimes try destroy live entry,
1611 and this produce panic in free.
1615 Description: Stack overflow during MDS log replay
1616 ease stack pressure by using a thread dealing llog_process.
1620 Description: wait until IO finished before start new when do lock cancel.
1621 Details : VM protocol want old IO finished before start new, in this case
1622 need wait until PG_writeback is cleared until check dirty flag and
1623 call writepages in lock cancel callback.
1625 Severity : enhancement
1627 Description: using special macro for print time and cleanup in includes.
1631 Description: mds_mfd_close() ASSERTION(rc == 0)
1632 Details : In mds_mfd_close(), we need protect inode's writecount change
1633 within its orphan write semaphore to prevent possible races.
1637 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1641 Frequency : rare, on shutdown ost
1642 Description: don't hit live lock with umount ost.
1643 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1644 use shrink_dcache_sb instead.
1648 Description: don't panic with use echo client
1649 Details : echo client pass NULL as client nid pointer and this produce null
1650 pointer dereference.
1654 Description: fix build on ppc32
1655 Details : compile code with -m64 flag produce wrong object file for ppc32.
1659 Description: add message levels for liblustreapi
1663 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1664 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1665 will result in an -EAGAIN and in the exit call path we call
1666 alloc_rr() which will with increasing aggressiveness attempt to
1667 aquire precreated objects on the minimum number of required OSCs.
1671 Description: Use old size assignment to avoid deadlock
1672 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1673 the scheduling while holding a spinlock. We do not need locking
1674 for size in ll_update_inode() because size is only updated from
1675 the MDS for directories or files without objects, so there is no
1676 other place to do the update, and concurrent access to such inodes
1677 are protected by the inode lock.
1681 Description: resolve "_IOWR redefined" build error on SLES10
1685 Description: dump the memory debugging after all modules are unloaded to
1686 suppress false negative in conf_sanity test 39
1688 Severity : enhancement
1690 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1695 Description: fix SLES kernel versioning
1696 Details : the kernel version for our SLES 10 kernel did not include a "-"
1697 before the "smp" at the end. while this was not a problem in
1698 general, it did mean that software trying to use the kernel
1699 version to try to detect a vendor specific kernel would fail.
1700 this was most evident by the OFED build scripts.
1704 Description: Don't update lov_desc members until making sure they are valid
1705 Details : When updating lov_desc members via proc fs, need fix their
1706 validities before doing the real update.
1710 Description: don't put request into delay list while invalidate in flight.
1711 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1712 import in flight. this produce timeout for invalidate and sometimes
1713 can cause stale data.
1717 Frequency : on ppc only
1718 Description: not convert ost objects for directory because it's not exist.
1719 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1723 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1725 Severity : enhancement
1727 Description: Update to support 2.6.22.14 vanilla kernel.
1731 Frequency : rare, on recovery
1732 Description: read procfs can produce deadlock in some situation
1733 Details : Holding lprocfs lock which send rpc can produce block for destroy
1734 obd objects and this also block reconnect with -EALREADY. This
1735 isn't fix all lprocfs bugs - but make it rare.
1737 Severity : enhancement
1739 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1742 Frequency : frequent on X2 node
1744 Description: mdc_set_open_replay_data LBUG
1745 Details : Set replay data for requests that are eligible for replay.
1749 Description: lustre_mgs: operation 101 on unconnected MGS
1750 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1751 MGC, and late on MGC cannot successfully connect to MGS and a lot
1752 of the error messages complaining that MGS is not connected.
1756 Frequency : on network error
1757 Description: panic with double free request if network error
1758 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1759 but it's true only for synchronus enqueue, for async enqueue
1760 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1763 Severity : enhancement
1765 Description: client-side metadata stat-ahead during readdir(directory readahead)
1766 Details : perform client-side metadata stat-ahead when the client detects
1767 readdir and sequential stat of dir entries therein
1770 Frequency : on start mds
1772 Description: Implement get_info(last_id) in obdfilter.
1775 Frequency : occasional
1777 Description: Correctly check stale fid, not start epoch if ost not support SOM
1778 Details : open with flag O_CREATE need set old fid in op_fid3 because
1779 op_fid2 was overwritten with new generated fid, but mds can answer
1780 with one of these two fids and both is not stale. Setattr
1781 incorrectly started an epoch and assume will be called
1782 done_writing, but without SOM done_writing ever being called.
1785 Frequency : rare, depends on device drivers and load
1787 Description: MDS or OSS nodes crash due to stack overflow
1788 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1789 In some cases, in conjunction with device drivers that use a lot
1790 of stack the MDS (or possibly OSS) service threads could overflow
1791 the stack. One change which was identified to consume additional
1792 stack has been reworked to avoid the extra stack usage.
1795 Frequency : occasional
1797 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1798 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1799 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1800 oscc_has_objects and see OSC has no objests and call
1801 oscc_internal_create to resend create request.
1803 Severity : enhancement
1805 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1807 Severity : enhancement
1809 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1812 Frequency : very rare
1814 Description: Fix replay if there is an un-replied request and open
1815 Details : In some cases, older replay request will revert the
1816 mcd->mcd_last_xid on MDS which is used to record the client's
1817 latest sent request.
1819 Severity : enhancement
1821 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1823 Severity : enhancement
1825 Description: Add rhel5 support to HEAD.
1827 Serverity : enhancement
1829 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1834 Description: Don't allow skipping OSTs if index has been specified.
1835 Details : Don't allow skipping OSTs if index has been specified, make
1836 locking in internal create lots better.
1840 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1841 Details : access to bitfield in structure is always rounded to long
1842 and this produce problem with not atomic change any bit.
1846 Description: Lustre make rpms failed.
1847 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1848 builds Lustre rpms from tarball.
1851 Frequency : rare on shutdown ost
1853 Description: If llog cancel was not send before clean_exports phase, this can
1854 produce deadlock in llog code.
1855 Details : If llog thread has last reference to obd and call class_import_put
1856 this produce deadlock because llog_cleanup_commit_master wait when
1857 last llog_commit_thread exited, but this never success because was
1858 called from llog_commit_thread.
1862 Description: allow userland application know is lost one of stripes.
1863 Details : fill lvb_blocks with error code on ost and return it to
1864 application if error flag found.
1868 Description: NULL lov_tgts causing MDS oops
1869 Details : more safe checks for NULL lov_tgts for avoid oops.
1871 Severity : enhancement
1873 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1877 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1878 Details : If an OST is down the MDS will hang indefinitely in
1879 obd_statfs_rqset() waiting for the statfs data. While for
1880 MDS QOS usage of statfs, it should not stuck in waiting.
1882 Severity : enhancement
1884 Description: remote_acl support
1885 Details : Support ACL-based permission check for remote user.
1886 Support setfacl/getfacl for remote user with the utils
1887 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1888 the system "{s,g}etfacl" utils.
1890 Severity : enhancement
1892 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1894 Severity : enhancement
1896 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1900 Description: make mgs_setparam() handle fsname containing dash
1901 Details : fsname containing a dash does not work with lctl conf_param
1903 Severity : enhancement
1905 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1907 Severity : enhancement
1909 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1913 Frequency : rare, at shutdown
1914 Description: access already free / zero obd_namespace.
1915 Details : if client_disconnect_export was called without force flag set,
1916 and exist connect request in flight, this can produce access to
1917 NULL pointer (or already free pointer) when connect_interpret
1918 store ocd flags in obd_namespace.
1922 Frequency : only at startup
1923 Description: not alloc memory with spinlock held.
1924 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1925 if any spinlock held.
1927 Severity : enhancement
1929 Description: make lustre randomly fail allocating memory
1930 Details : Make lustre randomly failed allocating memory for testing purpose.
1932 Severity : enhancement
1934 Description: lost problems with lov objid file
1935 Details : Fixes some scability and access to not inited memory problems
1936 in work with lov objdid file.
1941 Description: lfs find does not continue on file error
1942 Details : Continue other files processing when a file/dir is absent.
1946 Description: Inconsistent usage of lustre_pack_reply()
1947 Details : Standardize the usage of lustre_pack_reply() such that it
1948 always generate a CERROR on failure.
1951 Frequency : occasional
1953 Description: MDS hang or stay in waiting lock
1954 Details : If client receive lock with CBPENDING flag ldlm need send lock
1955 cancel as separate rpc, to avoid situation when cancel request
1956 can't processed due all i/o threads stay in wait lock.
1960 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1964 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1968 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1972 Description: Update to RHEL4 latest kernel.
1974 Severity : enhancement
1976 Description: Build SLES10 patchless client fails
1977 Details : The configure was broken by run ./configure with
1978 --with-linux-obj=.... argument for patchless client. When the
1979 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1980 can't search header adequately. Use absolute path such as
1981 -I($LINUX)/include instead.
1985 Description: interrupt oig_wait produce painc on resend.
1986 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1987 private set, and this produce situation when rq_ptlrpcd_data not
1988 copyed to new allocated request and triggered LBUG on assert
1989 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1990 ptlrpcd set if request is changed and can be safety changed to use
1995 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1996 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1997 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1998 these macros and replace them with their difinitions instead.
2002 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2003 Details : Update which_patch & target file for SLES10 latest kernel.
2006 Description: Remove client patches from SLES 10 kernel.
2007 Details : This causes SLES 10 clients to behave as patchless clients
2008 even on a Lustre-patched (server) kernel.
2010 Severity : enhancement
2012 Description: self-adjustable client's lru lists
2013 Details : use adaptive algorithm for managing client cached locks lru
2014 lists according to current server load, other client's work
2015 pattern, memory activities, etc. Both, server and client
2016 side namespaces provide number of proc tunables for controlling
2019 Severity : enhancement
2021 Description: light-weight GSS support
2022 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2023 and reduce performance overhead.
2025 Severity : enhancement
2027 Description: Linux keyring support
2028 Details : Support using service of Linux keyring for Lustre GSS internal
2029 context refresh/cache mechanism.
2033 Description: Fix errors in lfs documentation
2034 Details : Fixes man pages
2038 Description: don't use GFP_* in generic Lustre code.
2039 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2043 Description: obdclass is limited by single OBD_ALLOC(idarray)
2044 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2048 Description: warnings with build patchless client with vanila 2.6.19 and up
2049 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2050 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2054 Frequency : Only for SLES
2055 Description: sanity_quota fail test_1
2056 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2057 file, which makes TSTID[2] inunique.
2061 Frequency : Only for SLES9
2062 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2063 Details : Update target/ChangeLog/which_patch .
2067 Frequency : Only for test_10 in sanity-quota.sh
2068 Description: a bug in quota test code
2069 Details : checking a test flag in a wrong place
2073 Frequency : Only for test_13 in sanity-quota.sh
2074 Description: a bug in quota test script
2075 Details : shouldn't check overfull allocated quota
2079 Description: jbd statistics
2080 Details : Port older jbd statistics patch for sles10
2082 Severity : enhancement
2084 Description: Add printing inode info into message about error in writepage.
2089 Description: reply_lock_interpret crash due to race with it and lock cancel.
2090 Details : Do not replay locks that are being cancelled. Do not reference
2091 locks by their address during replay, just by their handle.
2095 Frequency : When flocks are used.
2096 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2097 Details : Flock locks might destroy just granted lock if it could be merged
2098 with another existing flock, this is done in completion handler,
2099 so teach ldlm_cli_enquque_fini that this is a valid case for
2105 Description: Oops in read and write path when failing to allocate lock.
2106 Details : Check if lock allocation failed and return error back.
2110 Description: lstripe command fails for valid OST index
2111 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2112 instead of lov->desc.ld_active_tgt_count.
2116 Description: sanity.sh failed test 103
2117 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2122 Description: df doesn't work properly if diskfs blocksize != 4K
2123 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2127 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2128 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2130 Severity : enhancement
2132 Description: Adaptive timeouts
2133 Details : RPC timeouts adapt to changing server load and network
2134 conditions to reduce resend attempts and improve recovery time.
2138 Description: llapi_file_create() does not allow some changes
2139 Details : add llapi_file_open() that allows specifying the mode and
2140 open flags, and also returns an open file handle.
2144 Description: merge and cleanup kernel patches.
2145 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2149 Description: Add journal checksum support.(Kernel part)
2150 Details : The journal checksum feature adds two new flags i.e
2151 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2152 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2153 indicates that the commit block contains the checksum for
2154 the blocks described by the descriptor blocks. Now commit
2155 record can be sent to disk without waiting for descriptor
2156 blocks to be written to disk. This behavior is controlled
2157 using JBD2_FEATURE_ASYNC_COMMIT flag.
2161 Description: OSS needs mutliple precreate threads
2162 Details : Add ability to start more than one create thread per OSS.
2166 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2167 Details : execute lfs setstripe on client
2171 Description: mds_obd_create error creating tmp object
2172 Details : When the user sets quota on root, llog will be affected and can't
2173 create files and write files.
2176 Frequency : Always on ia64 patchless client, and possibly others.
2178 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2179 Details : This allows the patchless client to be loaded on architectures
2180 without this export.
2184 Description: RedHat Update kernel for RHEL5
2185 Details : Add the kernel config file for RHEL5.
2189 Description: RedHat Update kernel for RHEL5
2190 Details : Modify the kernel config file more closer RHEL5.
2194 Description: Build failure against Centos5 (RHEL5)
2195 Details : Define PAGE_SIZE when it isn't present.
2199 Description: test_7 on sanity-quota.sh can't be used on separate machines
2200 Details : add facet to handle it
2204 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2205 Details : with NFS, the anon dentry's parent was set to itself in
2206 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2207 corresponding dentry other than use rec->ur_name.
2209 Severity : enhancement
2211 Description: lfs setstripe enhancement
2212 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2216 Description: enable data checksumming by default
2217 Details : enable checksum by default, allow --disable-checksum
2218 configure option and "-o nochecksum" mount option. Checksums
2219 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2223 Description: lustre support for RHEL5
2224 Details : Add support for RHEL5.
2228 Description: Client eviction due to failover config
2229 Details : after a connection loss, the lustre client should attempt to
2230 reconnect to the last active server first before trying the
2231 other potential connections.
2235 Description: when mds and osts use different quota unit(32bit and 64bit),
2236 quota will be released repeatly.
2237 Details : void sending multiple quota reqs to mds, which will keep the
2238 status between the reqs.
2242 Description: osts not allocated evenly to files
2243 Details : change the condition to increase offset_idx
2246 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2248 Description: Data corruption for OSTs that are formatted larger than 2TB
2250 Details : When generating the bio request for lustre file writes the
2251 sector number would overflow a temporary variable before being
2252 used for the IO. The data reads correctly from Lustre (which
2253 will overflow in a similar manner) but other file data or
2254 filesystem metadata may be corrupted in some cases.
2258 Description: Tune the kernel for good SCSI performance.
2259 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2260 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2265 Description: rewrite ext2-derived code in obdclass/uuid.c
2266 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2267 from scratch preserving functionality.
2271 Description: "lfs find -obd UUID" prints directories
2272 Details : "lfs find -obd UUID" will return all directory names instead
2273 of just file names. It is incorrect because the directories
2274 do not reside on the OSTs.
2278 Description: use i_size_read and i_size_write in 2.6 port
2279 Details : replace inode->i_size access with i_size_read/write()
2282 Frequency : when removing large files
2284 Description: scheduling issue during removal of large Lustre files
2285 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2286 It causes scheduling issues when removing large files (17TB in the
2290 Frequency : only with liblustre clients on XT3
2292 Description: evictions taking too long
2293 Details : allow llrd to evict clients directly on OSTs
2296 Frequency : only on ppc
2298 Description: /proc/fs/lustre/devices broken on ppc
2299 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2300 We should make sure the type of 'index' is loff_t and then cast
2301 explicitly as needed below. Do not assign an explicitly cast
2305 Frequency : only for rhel5
2307 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2308 Details : Modify the target file & which_kernel.
2310 Severity : enhancement
2312 Description: omit set fsid for export NFS
2313 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2314 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2318 Frequency : Always, for kernels after 2.6.16
2319 Description: Fix warning idr_remove called for id=.. which is not allocated.
2320 Details : Last kernels save old s_dev before kill super and not allow
2321 to restore from callback - restore it before call kill_anon_super.
2325 Description: buffer overruns could theoretically occur
2326 Details : llapi_semantic_traverse() modifies the "path" argument by
2327 appending values to the end of the origin string, and a
2328 overrun may occur. Adding buffer overrun check in liblustreapi.
2332 Description: Fix error on 'ls .' at the top of the Lustre mount.
2333 Details : Don't revalidate dentry if it is a root dentry.
2337 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2338 Details : Modify vm-tunables-rhel4.patch.
2342 Description: Kernel config for 2.6.18-vanilla.
2343 Details : Modify targets/2.6-vanilla.target.in.
2344 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2345 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2346 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2347 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2351 Description: improve handling recoverable errors
2352 Details : if request processig with error which can be recoverable on server
2353 request should be resend, otherwise page released from cache and
2358 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2359 Details : Modify target file & which_patch.
2360 A flaw was found in the IA32 system call emulation provided
2361 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2362 value could be stored in the %RAX register, which could trigger an
2363 out-of-bounds system call table access. An untrusted local user
2364 could exploit this flaw to run code in the kernel
2365 (ie a root privilege escalation). (CVE-2007-4573).
2369 Description: O_DIRECT bypasses client statistics.
2370 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2371 read_ahead_stats were updated. copied the stats section from
2372 osc_send_oap_rpc() into async_internal().
2376 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2380 Description: change order of libsysio includes
2381 Details : '#include sysio.h' should always come before '#include xtio.h'
2385 Description: handle "serious error: objid * already exists" more gracefully
2386 Details : If LAST_ID value on disk is smaller than the objects existing in
2387 the O/0/d* directories, it indicates disk corruption and causes an
2388 LBUG(). If the object is 0-length, then we should use the existing
2389 object. This will help to avoid a full fsck in most cases.
2391 Severity : enhancement
2393 Description: adapt the lustre_config script to support the upgrade case
2394 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2395 server targets to 1.6 in parallel.
2399 Description: To avoid grant space > avaible space when the disk is almost
2400 full. Without this patch you might see the error "grant XXXX >
2401 available" or some LBUG about grant, when the disk is almost
2403 Details : In filter_check_grant, for non_grant cache write, we should
2404 check the left space by if (*left > ungranted + bytes), instead
2405 of (*left > ungranted), because only we are sure the left space
2406 is enough for another "bytes", then the ungrant space should be
2407 increase. In client, we should update cl_avail_grant only there
2408 is OBD_MD_FLGRANT in the reply.
2412 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2413 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2414 described in CVE-2007-4573. This problem could allow untrusted
2415 local users to gain root access.
2418 Frequency : when using O_DIRECT and quotas
2420 Description: Incorrect file ownership on O_DIRECT output files
2421 Details : block usage reported by 'lfs quota' does not take into account
2422 files that have been written with O_DIRECT.
2427 Description: touch file failed when fs is not full
2428 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2429 otherwise we can get ENOSP while fs is not full.
2433 Description: parallel lock callbacks
2434 Details : Instead of sending blocking and completion callbacks as separated
2435 requests, adding them to a set and sending in parallel.
2438 Frequency : only for Cray XT3
2439 Bugzilla : 12829/13455
2440 Description: Changing primary group doesn't change the group lustre assigns to
2442 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2443 overridden with the primary group provided by the group upcall,
2444 whereas the supplied fsgid can be trusted if it is in the list of
2445 supplementary groups returned by the group upcall.
2447 Severity : enhancement
2449 Description: Allow masking D_WARNING, D_ERROR messages from console
2450 Details : Console messages can now be disabled via lnet.printk.
2454 Description: User code with malformed file open parameter crashes client node
2455 Details : Before packing join_file req, all the related reference should be
2456 checked carefully in case some malformed flags cause fake
2457 join_file req on client.
2461 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2462 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2463 mode is equal to the granted mode. However, it should also check
2464 that LDLM_FL_CP_REQD is not set.
2468 Description: Heavy nfs access might result in deadlocks
2469 Details : After ELC code landed, it is now improper to enqueue any mds
2470 locks under och_sem, because enqueue might want to decide to
2471 cancel open locks for same inode we are holding och_sem for.
2475 Description: Client eviction while running blogbench
2476 Details : A lot of unlink operations with concurrent I/O can lead to a
2477 deadlock causing evictions. To address the problem, the number of
2478 oustanding OST_DESTROY requests is now throttled to
2479 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2480 callbacks are processed in priority.
2484 Description: enable ACLs on MDS by default
2485 Details : ACLs must be enabled on MDS by default.
2488 Frequency : PPC/PPC64 only
2490 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2491 Details : fix duplicated definitions between asm-ppc64/types.h and
2492 lustre_types.h on PPC.
2495 Frequency : PPC/PPC64 only
2497 Description: asm-ppc/segment.h does not exist
2498 Details : fix compile issue on PPC.
2502 Description: better handle error messages in extents code
2505 Frequency : RHEL4 only
2507 Description: mkfs is very slow on IA64/RHEL4
2508 Details : A performance regression has been discovered in the MPT Fusion
2509 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2510 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2511 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2513 Severity : enhancement
2515 Description: SNMP support enhancement
2516 Details : Adding total number of sampled request for an MDS node in snmp
2519 Severity : enhancement
2521 Description: Optimize ldlm waiting list processing for PR extent locks
2522 Details : When processing waiting list for read extent lock and meeting read
2523 lock that is same or wider to it that is not contended, skip
2524 processing rest of the list and immediatelly return current
2525 status of conflictness, since we are guaranteed there are no
2526 conflicting locks in the rest of the list.
2530 Description: Time out and refuse to reconnect
2531 Details : When the failover node is the primary node, it is possible
2532 to have two identical connections in imp_conn_list. We must
2533 compare not conn's pointers but NIDs, otherwise we can defeat
2534 connection throttling.
2538 Description: port llog fixes from b1_6 into HEAD
2539 Details : Port llog reference couting and some llog cleanups from b1_6
2540 (bug 10800) into HEAD, for protect from panic and access to already
2541 free llog structures.
2545 Description: Detect stride IO mode in read-ahead
2546 Details : When a client does stride read, read-ahead should detect that and
2547 read-ahead pages according to the detected stride pattern.
2551 Description: data checksumming impacts single node performance
2552 Details : add support for several checksum algorithm. Currently, only CRC32
2553 and Adler-32 are supported. The checksum type can be changed on
2554 the fly via /proc/fs/lustre/osc/*/checksum_type.
2558 Description: use adler32 for page checksums
2559 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2564 Description: build for x2 fails
2565 Details : fix compile issue on Cray systems.
2569 Description: Properly match for duplicate locks
2570 Details : Due to different lock order from skiplists code, we need to
2571 traverse entire list for now
2574 Frequency : only on PPC/SLES10
2576 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2577 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2578 be defined. Add a hack in mkfs_lustre.c to work around this compile
2583 Description: LASSERT on MDS when client holding flock lock dies
2584 Details : ldlm pool logic depends on number of granted locks equal to
2585 number of released locks which is not true for flock locks, so
2586 just exclude such locks from consideration.
2590 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2591 Details : Use fsfilt_sync() for both the whole filesystem sync and
2592 individual file sync to eliminate dangerous inode locking
2593 with I_LOCK that can lead to a deadlock.
2597 Description: performance in 1.6.3
2598 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2599 sure that 1MB requests are not fragmented by the block layer.
2601 Severity : enhancement
2603 Description: organize the server-side client stats on per-nid basis
2604 Details : Change the structure of stats under obdfilter and mds to
2612 The "uuid"s file would list the uuids of _active_ exports.
2613 And the clear entry is to clear all stats and stale nids.
2615 Severity : enhancement
2617 Description: eliminate client locks in face of contention
2618 Details : file contention detection and lockless i/o implementation
2619 for contended files.
2623 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2626 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2628 Description: MDS may lose file striping (and hence file data) in some cases
2629 Details : If there are additional extended attributes stored on the MDS,
2630 in particular ACLs, SELinux, or user attributes (if user_xattr
2631 is specified for the client mount options) then there is a risk
2632 of attribute loss. Additionally, the Lustre file striping
2633 needs to be larger than default (e.g. striped over all OSTs),
2634 and an additional attribute must be stored initially in the
2635 inode and then increase in size enough to be moved to the
2636 external attribute block (e.g. ACL growing in size) for file
2641 Description: skiplist implementation simplification
2642 Details : skiplists are used to group compatible locks on granted list
2643 that was implemented as tracking first and last lock of each
2644 lock group the patch changes that to using doubly linked lists
2648 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2649 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2650 because LASSERT on both the data supplied by a client, and the
2651 data on disk is dangerous and incorrect.
2653 Severity : enhancement
2655 Description: Slow truncate/writes to huge files at high offsets.
2656 Details : Directly associate cached pages to lock that protect those pages,
2657 this allows us to quickly find what pages to write and remove
2658 once lock callback is received.
2662 Description: more ldlm soft lockups
2663 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2664 starve other threads from the resource lock for a long time in
2665 case of long waiting queue, so change the debug level from
2666 D_OTHER to the less frequently used D_INFO.
2668 Severity : enhancement
2670 Description: add -gid, -group, -uid, -user options to lfs find
2674 Description: Hung threads in invalidate_inode_pages2_range
2675 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2676 once one is completed. As a result, some RPCs are stuck in the
2677 queue and are never sent.
2681 Description: filter threads hungs on waiting journal commit
2682 Details : Cleanup filter group llog code, then only filter group llog will
2683 be only created in the MDS/OST syncing process.
2687 Description: Procfs and llog threads access destoryed import sometimes.
2688 Details : Sync the import destoryed process with procfs and llog threads by
2689 the import refcount and semaphore.
2691 Severity : enhancement
2693 Description: openlock cache of b1_6 port to HEAD
2698 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2699 Details : If insertion of an extent fails, then discard the inode
2700 preallocation and free data blocks else it can lead to duplicate
2705 Description: don't always update ctime in ext3_xattr_set_handle()
2706 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2707 In some cases the ctime should not be updated, for example for
2708 2.0->1.8 compatibility it is necessary to delete an xattr and it
2709 should not update the ctime.
2713 Bugzilla : 15713/16362
2714 Description: Assertion in iopen_connect_dentry in 1.6.3
2715 Details : looking up an inode via iopen with the wrong generation number can
2716 populate the dcache with a disconneced dentry while the inode
2717 number is in the process of being reallocated. This causes an
2718 assertion failure in iopen since the inode's dentry list contains
2719 both a connected and disconnected dentry.
2723 Description: assertion failure in ldlm_handle2lock()
2724 Details : fix a race between class_handle_unhash() and class_handle2object()
2725 introduced in lustre 1.6.5 by bug 13622.
2730 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2731 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2736 Description: oops in sd_iostats_seq_show()
2737 Details : unloading/reloading the scsi low level driver triggers a kernel
2738 bug when trying to access the sd iostat file.
2743 Description: Kernel panics during QLogic driver reload
2744 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2745 patch, causing memory corruption.
2750 Description: journal_dev option does not work in b1_6
2751 Details : pass mount option during pre-mount.
2753 Severity : enhancement
2755 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2756 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2757 extent information of a file. It can be used to map logical blocks
2758 in a file to physical blocks in the block device.
2762 Description: LDLM soft lockups - improvement
2763 Details : It is be possible to send the lock handle along with each read
2764 or write request because the client is already doing a lock match
2765 itself so there isn't any reason the OST should have to re-do that
2769 Frequency : only on Cray X2
2771 Description: X2 build failures
2772 Details : fix build failures on Cray X2.
2776 Description: xid & resent requests
2777 Details : Initialize RPC XID from clock at startup (randomly if clock is
2780 Severity : enhancement
2782 Description: Add lustre_start utility to start or stop multiple Lustre servers
2787 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2788 Details : In case of memory pressure, list_del() can be called twice on
2789 req->rq_history_list, causing a kernel oops.
2793 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2794 Details : fix stack overflow in the distributed lock manager by defering
2795 export eviction after a failed AST to the elt thread instead of
2796 handling it in the dlm interpret routine.
2800 Description: Convert some comments to new format.
2801 Details : Update documenting comments to match doxygen conventions.
2805 Description: Grammar fixes.
2806 Details : A couple of trivial spelling fixes.
2810 Description: OSD_COUNTERS-mandatory
2811 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2812 to simplify the code.
2816 Description: simplify cmm_device freeing logic.
2817 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2821 Description: Add lockdep support to dt_object_operations locking interface.
2822 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2823 parameter indicating lock ordering. Update mdd code to use new
2828 Description: Introduce failloc constants for lockless IO tests.
2829 Details : Add two new failloc constants to test lockless IO. Only one of
2830 them in implemented---another is checked in yet to be landed
2835 Description: Add lockdep support for inode mutex.
2836 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2837 in the situations where Lustre has to lock more than one inode
2842 Description: Add optional invariants checking support.
2843 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2844 configure switch. This macro is to be used for consistency and
2845 sanity checks that are too expensive to be left in `production'
2850 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2851 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2856 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2857 Details : Use spin_lock_nested() in (the only) situation where more than
2858 one ldlm_lock is locked simultaneously. Also, fix possible
2859 dead-lock in ldlm_lock_change_resource() by enforcing particular
2864 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2865 Details : Instead of specifying each ldlm_lock call-back through separate
2866 parameter, wrap them into struct ldlm_callback_suite.
2870 Description: Kill join_lru obd method and its callers.
2871 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2872 in memory---a requirement implemented through ->o_join_lru() obd
2873 method in HEAD. Remove this method and its users.
2877 Description: Add asynchronous ldlm ENQUEUE completion handler.
2878 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2879 case that stock ldlm_completion_ast() cannot handle as it waits
2880 until lock is granted. Introduce new ldlm_completion_ast_async()
2881 for this. Also comment ldlm_completion_ast().
2885 Description: ldlm_error <-> errno conversion.
2886 Details : Add functions to map (rather arbitrary) between LDLM error codes
2887 and standard errno values. CLIO needs this to prevent LDLM specific
2888 constants from escaping ldlm and osc.
2892 Description: Kill unused ldlm_handle2lock_ns() function.
2893 Details : Kill unused ldlm_handle2lock_ns() function.
2897 Description: Add lu_ref support to ldlm_lock
2898 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2899 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2900 and ldlm_resource. LDLM interface has to be changed, because code that
2901 releases a reference on a lock, has to "know" what reference this is.
2902 In the most frequent case
2904 lock = ldlm_handle2lock(handle);
2906 LDLM_LOCK_PUT(lock);
2908 no changes are required. When any other reference (received _not_ from
2909 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2910 instead of LDLM_LOCK_PUT().
2912 Arguably, changes are pervasive, and interface requires some discipline
2913 for proper use. On the other hand, it was very instrumental in finding
2914 a few leaked lock references.
2918 Description: Add ldlm_lock_addref_try().
2919 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2920 attempts to addref a lock that might be being canceled
2925 Description: Add ldlm_weigh_callback().
2926 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2927 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2928 hard-coded `number of pages' logic.
2932 Description: Add lockdep annotations to llog code.
2933 Details : Use appropriately tagged _nested() locking calls in the places
2934 where llog takes more than one ->lgh_lock lock.
2938 Description: Add loi_kms_set().
2939 Details : Wrap kms updates into a helper function.
2943 Description: Constify instances of struct lsm_operations.
2944 Details : Constify instances of struct lsm_operations.
2948 Description: lu_conf support.
2949 Details : On a server, a file system object is uniquely identified
2950 by a fid, which is sufficient to locate and load all object
2951 state (inode). On a client, on the other hand, more data are
2952 necessary instantiate an object. Change lu_object_find() and
2953 friends to take additional `lu_conf' argument describing object.
2954 Typically this includes layout information.
2958 Description: lu_context fixes.
2959 Details : Introduce new lu_context functions that are needed on the client
2960 side, where some system threads (ptlrpcd) are shared by multiple
2961 modules, and so cannot be stopped during module shutdown.
2965 Description: Add start and stop methods to lu_device_type_operations.
2966 Details : Introduce two new methods in lu_device_type_operations, that are
2967 invoked when first instance of a given type is created and last one
2968 is destroyed respectively. This is need by CLIO.
2972 Description: Add lu_ref support to struct lu_device.
2973 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2974 track leaked references.
2978 Description: Introduce lu_kmem_descr.
2979 Details : lu_kmem_descr and its companion interface allow to create
2980 and destroy a number of kmem caches at once.
2984 Description: Fix lu_object finalization race.
2985 Details : Fix a race between lu_object_find() finding an object and its
2986 concurrent finalization. This race is (most likely) not possible
2987 on the server, but might happen on the client.
2991 Description: Introduce lu_ref interface.
2992 Details : lu_ref is a debugging module allowing to track references to
2993 a given object. It is quite cpu expensive, and has to be
2994 explicitly enabled with --enable-lu_ref. See usage description
2999 Description: Factor lu_site procfs stats into a separate function.
3000 Details : Separate lu_site stats printing code into a separate function
3001 that can be reused on a client.
3005 Description: Constify instances of struct {lu,dt,md}_device_operations.
3006 Details : Constify instances of struct {lu,dt,md}_device_operations.
3010 Description: Introduce struct md_site and move meta-data specific parts of
3011 struct lu_site here.
3012 Details : Move md-specific fields out of struct lu_site into special struct
3013 md_site, so that lu_site can be used on a client.
3017 Description: Kill mdd_lov_destroy().
3018 Details : Remove unused mdd code.
3022 Description: Add st_block checking to multistat.c.
3023 Details : Add st_block checking to multistat.c.
3027 Description: Add lu_ref support to struct obd_device.
3028 Details : Add lu_ref tracking to obd_device.
3032 Description: Kill obd_set_fail_loc().
3033 Details : Remove unused code.
3037 Description: Add special type for ptlrpc_request interpret functions.
3038 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3039 there. Actual usage will be in CLIO.
3043 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3044 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3045 doesn't work with lockdep.
3049 Description: Add rwv.c test program.
3050 Details : New testing program exercising readv(2) and writev(2) (Qian).
3054 Description: Add sendfile.c test program.
3055 Details : New testing program exercising sendfile(2) (Jay).
3059 Description: Ratelimit a message that can be very frequent.
3060 Details : Ratelimit a memory allocation failure message that can
3065 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3066 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3070 Description: Liblustre build fixes.
3071 Details : Liblustre build fixes.
3075 Description: libcfs: add cfs_{need,cond}_resched() interface.
3076 Details : libcfs: add cfs_{need,cond}_resched() definition and
3077 implementations for Linux, NT, and liblustre.
3079 Severity : enhancement
3081 Description: More exported tunables for mballoc
3082 Details : Add support for tunable preallocation window and new tunables for
3083 large/small requests
3087 Description: Detect corruption of block bitmap and checking for preallocations
3088 Details : Checks validity of on-disk block bitmap. Also it does better
3089 checking of number of applied preallocations. When corruption is
3090 found, it turns filesystem readonly to prevent further corruptions.
3094 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3095 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3096 window to 0. If it does hit cache miss, and read-pattern is still
3097 stride-io mode, does not reset the stride window, but also does not
3098 increase the stride window length in this case.
3102 Frequency : only for big-endian servers
3103 Description: Check if system is big-endian while mounting fs with extents feature
3104 Details : Mounting a filesystem with extents feature will fail on big-endian
3105 systems since ext3-based ldiskfs is not supported on big-endian
3106 systems. This can be over-riden with "bigendian_extents" mount option.
3108 Severity : enhancement
3110 Description: The root squash functionality
3111 Details : A security feature, which is to prevent users from being able
3112 to mount lustre on their desktop, run as root, and delete
3113 all of the files in the filesystem. The goal is accomplished by
3114 remapping user id (UID) and group id (GID) of the root user to
3115 a UID and GID specified by the system administartor via Lustre
3116 configuration management server (MGS). The functionality also
3117 allows to specify sets of clients for which the remapping does
3122 Description: Excessive recovery window
3123 Details : With AT enabled, the recovery window can be excessively long (6000+
3124 seconds). To address this problem, we no longer use
3125 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3126 timeout no longer depends on the service time, it is set to
3127 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3128 time via pb_service_time.
3132 Description: Watchdog triggered on MDS failover
3133 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3134 the OSTs know that the MDS "UUID" can be reused for the same export
3135 from a different NID, so we do not need to wait for the export to be
3139 Frequency : rare, only if using MMP with Linux RAID
3141 Description: MMP doesn't work with Linux RAID
3142 Details : While using HA for Lustre servers with Linux RAID, it is possible
3143 that MMP will not detect multiple mounts. To make this work we
3144 need to unplug the device queue in RAID when the MMP block is being
3145 written. Also while reading the MMP block, we should read it from
3146 disk and not the cached one.
3148 Severity : enhancement
3150 Description: open file using fid
3151 Details : A file can be opened using just its fid, like
3152 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3155 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3158 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3159 Details : add proper locking for creation of jbd_4k slab cache
3163 Description: MMP check in ext3_remount() fails without displaying any error
3164 Details : When multiple mount protection fails during remount, proper error
3167 Severity : enhancement
3169 Description: Allow stripe size to be up to 4G-64k
3170 Details : Fix math logic to allow large stripe sizes.
3174 Description: add check for >8TB ldiskfs filesystems
3175 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3176 Don't allow >8TB ldiskfs filesystems to be mounted without
3177 force_over_8tb mount option
3182 Descriptoin: OSS complains "too many blocks" and crashes
3183 Details : In filter_commitrw_write():
3184 1. lquota_chkquota();
3185 2. obdo_from_inode();
3186 3. lquota_pending_commit();
3187 "2." has changed uid and gid in this case, then caused
3188 lquota_chkquota() and lquota_pending_commit() to use
3189 different uid and gid. That is the root of the bug.
3193 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3194 Details : Detect and print out the 'stripe count = -1' case properly.
3198 Description: Out or order replies might be lost on replay
3199 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3200 request with tid smaller than one currently being added, add it
3201 to the start, not end of the list.
3203 --------------------------------------------------------------------------------
3205 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3207 * Support for kernels:
3208 2.6.5-7.283 (SLES 9),
3209 2.6.9-55.EL (RHEL 4),
3210 2.6.16.46-0.14 (SLES 10),
3211 2.6.18.8 vanilla (kernel.org)
3212 * Client support for unpatched kernels:
3213 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3214 2.6.16 - 2.6.22 vanilla (kernel.org)
3215 * Due to recently discovered recovery problems, we do not recommend
3216 using patchless RHEL 4 clients with this or any earlier release.
3217 * Recommended e2fsprogs version: 1.39.cfs8
3218 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3219 * Starting with this release, the ldiskfs backing filesystem required
3220 by Lustre is now in its own package, lustre-ldiskfs. This package
3221 should be installed. It is versioned separately from Lustre and
3222 may be released separately in future.
3226 Description: block reactivating mgc import until all deactivates complete
3227 Details : Fix race when failing back MDT/MGS to itself (testing)
3230 Frequency : at statup only
3232 Description: mds_lov_synchronize race leads to various problems
3233 Details : simultaneous MDT->OST connections at startup can cause the
3234 sync to abort, leaving the OSC in a bad state.
3236 Severity : enhancement
3238 Description: add optional extra BUILD_VERSION info
3239 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3240 to override the lustre version.
3243 Frequency : 2.6.18 servers only
3245 Description: ll_kern_mount() doesn't release the module reference
3246 Details : The ldiskfs module reference count never drops down to 0
3247 because ll_kern_mount() doesn't release the module reference.
3252 Description: server LBUG when using old ost_num_threads parameter
3253 Details : Accept the old ost_num_threads parameter but warn that it
3254 is deprecated, and fix an off-by-one error that caused an LBUG.
3259 Description: Transient SCSI error results in persistent IO issue
3260 Details : iobuf->dr_error is not reinitialized to 0 between two
3264 Frequency : sometimes when underlying device returns I/O errors
3266 Description: OSTs not going read-only during write failures
3267 Details : OSTs are not remounted read-only when the journal commit threads
3268 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3269 instead of the ext3 wrappers.
3273 Description: poor connect scaling with increasing client count
3274 Details : Don't run filter_grant_sanity_check for more than 100 exports
3275 to improve scaling for large numbers of clients.
3278 Frequency : SLES10 only
3280 Description: sanity-quota.sh quotacheck failed: rc = -22
3281 Details : Quotas cannot be enabled on SLES10.
3284 Frequency : liblustre clients only
3286 Description: getdirentries does not give error when run on compute nodes
3287 Details : getdirentries does not fail when the size specified as an argument
3288 is too small to contain at least one entry
3290 Severity : enhancement
3292 Description: Add LNET router traceability for debug purposes
3293 Details : If a checksum failure occurs with a router as part of the
3294 IO path, the NID of the last router that forwarded the bulk data
3295 is printed so it can be identified.
3300 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3301 Details : Due to a race condition, liblustre clients were occasionally
3302 evicted incorrectly.
3304 Severity : enhancement
3306 Description: lfs setstripe use optional parameters instead of postional
3309 Severity : enhancement
3311 Description: Nanosecond timestamp support for ldiskfs
3312 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3313 resolution timestamps. There is not yet support for this at
3314 the Lustre filesystem level.
3317 Frequency : during server recovery
3319 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3320 Details : request with rq_no_resend flag not awake l_wait_event if they get
3324 Frequency : nfs export on patchless client
3326 Description: connectathon hang when test nfs export over patchless client
3327 Details : Disconnected dentry cannot be found with lookup, so we do not need
3328 to unhash it or make it invalid
3331 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3332 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3333 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3338 Description: open req refcounting wrong on reconnect
3339 Details : If reconnect happened between getting open reply from server and
3340 call to mdc_set_replay_data in ll_file_open, we will schedule
3341 replay for unreferenced request that we are about to free.
3342 Subsequent close will crash in variety of ways.
3343 Check that request is still eligible for replay in
3344 mdc_set_replay_data().
3349 Description: disable writes to filesystem when reading health_check file
3350 Details : the default for reading the health_check proc file has changed
3351 to NOT do a journal transaction and write to disk, because this
3352 can cause reads of the /proc file to hang and block HA state
3353 checking on a healthy but otherwise heavily loaded system. It
3354 is possible to return to the previous behaviour during configure
3355 with --enable-health-write.
3357 Severity : enhancement
3359 Description: 64-bit inode version
3360 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3361 to the inode. This will be required for version-based recovery.
3366 Description: MDS fails to start if a duplicate client export is detected
3367 Details : in some rare cases it was possible for a client to connect to
3368 an MDS multiple times. Upon recovery the MDS would detect this
3369 and fail during startup. Handle this more gracefully.
3371 Severity : enhancement
3373 Description: Add -o localflock option to simulate old noflock
3375 Details : This will achieve local-only flock/fcntl locks
3381 Description: log_commit_thread vs filter_destroy race leads to crash
3382 Details : Take import reference before releasing llog record semaphore
3387 Description: Wrong request locking in request set processing
3388 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3389 imp_delayed_list, in this place should be used imp_lock.
3392 Frequency : when reconnection
3394 Description: Grant Leak when osc reconnect to OST
3395 Details : When osc reconnect ost, OST(filter) should check whether it
3396 should grant more space to client by comparing fed_grant and
3397 cl_avail_grant, and return the granted space to client instead
3398 of "new granted" space, because client will call osc_init_grant
3399 to update the client grant space info.
3402 Frequency : when client reconnect to OST
3404 Description: Grant Leak when osc do resend and replay bulk write
3405 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3406 bulk write request, because the grant info while be sync between
3407 OSC and OST when reconnect, and we should ignore the grant info
3408 these of resend/replay write req.
3413 Description: Grant space more than available left space sometimes.
3414 Details : When then OST is about to be full, if two bulk writing from
3415 different clients came to OST. Accord the avaliable space of the
3416 OST, the first req should be permitted, and the second one
3417 should be denied by ENOSPC. But if the seconde arrived before
3418 the first one is committed. The OST might wrongly permit second
3419 writing, which will cause grant space > available space.
3422 Frequency : when client is evicted
3424 Description: Grant might be wrongly erased when osc is evicted by OST
3425 Details : when the import is evicted by server, it will fork another
3426 thread ptlrpc_invalidate_import_thread to invalidate the
3427 import, where the grant will be set to 0. While the original
3428 thread will update the grant it got when connecting. So if
3429 the former happened latter, the grant will be wrongly errased
3430 because of this race.
3435 Description: Checking Stale with correct fid
3436 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3437 to check whether it is stale, because sometimes, we want the
3438 enqueue happened anyway, and op_data.fid2 will not be initialized.
3440 Severity : enhancement
3442 Description: update patchless client
3443 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3446 Frequency : only with 2.4 kernel
3448 Description: random memory corruption
3449 Details : size of struct ll_inode_info is to big for union inode.u and this
3450 can be cause of random memory corruption.
3455 Description: Memory leak in recovery
3456 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3457 It should also check obd_fail before fsfilt_start, otherwise if
3458 fsfilt_start return -EROFS,(failover mds during mds recovery).
3459 then the req will return with repmsg->transno = 0 and rc = EROFS.
3460 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3461 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3462 be freed no matter whether fsfilt_commit success or not.
3465 Frequency : only with huge count clients
3467 Description: Prevents from taking the superblock lock in llap_from_page for
3469 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3470 for llap_from_page call in ll_removepage prevents from taking the
3471 superblock lock for a soon died page.
3476 Description: Not check open intent error before release open handle
3477 Details : in some rare cases, the open intent error is not checked before
3478 release open handle, which may cause
3479 ASSERTION(open_req->rq_transno != 0), because it tries to release
3480 the failed open handle.
3485 Description: Set cat log bitmap only after create log success.
3486 Details : in some rare cases, the cat log bitmap is set too early. and it
3487 should be set only after create log success.
3491 Description: Accessing a block bevice can re-enable I/O when Lustre is
3492 tearing down a device.
3493 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3498 Description: service threads may hog cpus when there are a lot of requests
3499 Details : Insert cond_resched to give other threads a chance to use some CPU
3504 Description: the cat log was not initialized in recovery
3505 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3506 unlink log on mds will not be initialized until mds post
3507 recovery. And also in mds post recovery, the unlink log will
3508 initialization will be done asynchronausly, so there will be race
3509 between add unlink log and unlink log initialization.
3513 Description: brw_stats were being printed incorrectly
3514 Details : brw_stats were being printed as log2 but all of them were not
3515 recorded as log2. Also remove some code duplication arising from
3516 filter_tally_{read,write}.
3520 Frequency : rare, only in recovery.
3521 Description: ASSERTION(req->rq_type != LI_POISON) failed
3522 Details : imp_lock should be held while iterating over imp_sending_list for
3523 prevent destroy request after get timeout in ptlrpc_queue_wait.
3527 Description: replay-single.sh test 52 fails
3528 Details : A lock's skiplist need to be cleanup when it being unlinked
3529 from its resource list.
3533 Description: Short directio read returns full requested size rather than
3535 Details : Direct I/O operations should return actual amount of bytes
3536 transferred rather than requested size.
3538 Severity : enhancement
3540 Description: metadata RPC reduction (e.g. for rm performance)
3541 Details : decrease the amount of synchronous RPC between clients and servers
3542 by canceling conflicing lock before the operation on the client
3543 and packing thier handles into the main operation RPC to server.
3545 Severity : enhancement
3547 Description: Async OSC create to avoid the blocking unnecessarily.
3548 Details : If an OST has no remaining object, system will block on the
3549 creation when it needs to create a new object on this OST. Now,
3550 ways use pre-created objects when available, instead of blocking on
3551 an empty osc while others are not empty. If we must block, we
3552 block for the shortest possible period of time.
3556 Description: improve handling recoverable errors
3557 Details : if request processig with error which can be recoverable on server
3558 request should be resend, otherwise page released from cache and
3561 Severity : enhancement
3563 Description: refine locking for avoid write wrong info into lov_objid file
3564 Details : fix possible races with add new target and write/update data in
3567 --------------------------------------------------------------------------------
3569 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3574 Frequency : on some architectures
3576 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3577 Details : Uninitialized flags sometimes cause configuration commands to
3581 Frequency : patchless clients only
3583 Description: missing __iget() symbol export
3584 Details : The __iget() symbol export is missing. To avoid the need for
3585 this on patchless clients the deathrow inode reaper is turned
3586 off, and we depend on the VM to clean up old inodes. This
3587 dependency was during via the fix for bug 12181.
3589 --------------------------------------------------------------------------------
3591 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3593 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3594 INTEROPERATE with older versions automatically. In many cases a
3595 special upgrade step is needed. Please read the
3596 user documentation before upgrading any part of a live system.
3597 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3598 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3599 * WARNING: Lustre configuration and startup changes are required with
3600 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3604 Severity : enhancement
3606 Description: MountConf
3607 Details : Lustre configuration is now managed via mkfs and mount
3608 commands instead of lmc and lconf. New obd types (MGS, MGC)
3609 are added for dynamic configuration management. See
3610 https://mail.clusterfs.com/wikis/lustre/MountConf for
3613 Severity : enhancement
3615 Description: dynamic OST addition
3616 Details : OSTs can now be added to a live filesystem
3618 Severity : enhancement
3620 Description: startup order invariance
3621 Details : MDTs and OSTs can be started in any order. Clients only
3622 require the MDT to complete startup.
3624 Severity : enhancement
3626 Description: parallel, asynchronous orphan cleanup
3627 Details : orphan cleanup is now performed in separate threads for each
3628 OST, allowing parallel non-blocking operation.
3630 Severity : enhancement
3632 Description: optimized stripe assignment
3633 Details : stripe assignments are now made based on ost space available,
3634 ost previous usage, and OSS previous usage, in order to try
3635 to optimize storage space and networking resources.
3637 Severity : enhancement
3639 Description: Permanently set tunables
3640 Details : All writable /proc/fs/lustre tunables can now be permanently
3641 set on a per-server basis, at mkfs time or on a live system.
3643 Severity : enhancement
3645 Description: Lustre message v2
3646 Details : Add lustre message format v2.
3648 Severity : enhancement
3650 Description: client OST exclusion list
3651 Details : Clients can be started with a list of OSTs that should be
3652 declared "inactive" for known non-responsive OSTs.
3656 Description: SPEC SFS validation failure on NFS v2 over lustre.
3657 Details : Changes the blocksize for regular files to be 2x RPC size,
3658 and not depend on stripe size.
3660 Severity : enhancement
3662 Description: Multiple MD RPCs in flight.
3663 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3664 To avoid overly-overloading MDS, introduce a limit on number of
3665 MDS RPCs in flight for a single client and add /proc controls
3666 to adjust this limit.
3668 Severity : enhancement
3670 Description: client read/write statistics
3671 Details : Add client read/write call usage stats for performance
3672 analysis of user processes.
3673 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3674 file access. extents_stats shows chunk size distribution.
3675 extents_stats_per_process show chunk size distribution per
3678 Severity : enhancement
3680 Description: mds statistics
3681 Details : Add detailed mds operations statistics in
3682 /proc/fs/lustre/mds/*/stats.
3686 Description: Failure of copying files with lustre special EAs.
3687 Details : Client side always return success for setxattr call for lustre
3688 special xattr (currently only "trusted.lov").
3693 Description: Refcount LNET uuids
3694 Details : The global LNET uuid list grew linearly with every startup;
3695 refcount repeated list entries instead of always adding to
3698 Severity : enhancement
3700 Description: Dynamic service threads
3701 Details : Within a small range, start extra service threads
3702 automatically when the request queue builds up.
3705 Frequency : mixed-endian client/server environments
3707 Description: mixed-endian crashes
3708 Details : The new msg_v2 system had some failures in mixed-endian
3711 Severity : enhancement
3713 Description: Easy OST removal
3714 Details : OSTs can be permanently deactivated with e.g. 'lctl
3715 conf_param lustre-OST0001.osc.active=0'
3717 Severity : enhancement
3719 Description: MGS proc entries
3720 Details : Added basic proc entries for the MGS showing what filesystems
3723 Severity : enhancement
3725 Description: provide MGS failover
3726 Details : Added config lock reacquisition after MGS server failover.
3728 Severity : enhancement
3730 Description: add Linux 2.4 support
3731 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3735 Description: a large application tries to do I/O to the same resource and dies
3736 in the middle of it.
3737 Details : Check the req->rq_arrival time after the call to
3738 ost_brw_lock_get(), but before we do anything about
3739 processing it & sending the BULK transfer request. This
3740 should help move old stale pending locks off the queue as
3741 quickly as obd_timeout.
3744 Frequency : when an incorrect nid is specified during startup
3746 Description: ptlrpc connect to non-existant node causes kernel crash
3747 Details : LNET can't be re-entered from an event callback, which
3748 happened when we expire a message after the export has been
3749 cleaned up. Instead, hand the zombie cleanup off to another
3752 Severity : enhancement
3754 Description: plain/inodebits lock performance improvement
3755 Details : Grouping plain/inodebits in granted list by their request modes
3756 and bits policy, thus improving the performance of search through
3760 Frequency : only if OST filesystem is corrupted
3762 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3763 Details : for a short time RPCs with bulk IO are in the replay list,
3764 but replay of bulk IOs is unimplemented. If the OST filesystem
3765 is corrupted due to disk cache incoherency and then replay is
3766 started it is possible to trip an assertion. Avoid putting
3767 committed RPCs into the replay list at all to avoid this issue.
3770 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3772 Description: System hang on startup
3773 Details : This bug allowed the liblustre (e.g. catamount) client to
3774 return to the app before handling all startup RPCs. This
3775 could leave the node unresponsive to lustre network traffic
3776 and manifested as a server ptllnd timeout.
3778 Severity : enhancement
3780 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3781 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3782 boolean to control whether to print peer debug info when a
3783 client's RPC times out.
3786 Frequency : only for kernels with patches from Lustre below 1.4.3
3788 Description: Remove old rdonly API
3789 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3792 Frequency : only for devices with external journals
3794 Description: Set external device read-only also
3795 Details : During a commanded failover stop, we set the disk device
3796 read-only while the server shuts down. We now also set any
3797 external journal device read-only at the same time.
3800 Frequency : when upgrading from 1.4 while trying to change parameters
3802 Description: The wrong (new) MDC name was used when setting parameters for
3803 upgraded MDT's. Also allows changing of OSC (and MDC)
3804 parameters if --writeconf is specified at tunefs upgrade time.
3807 Frequency : when setting specific ost indicies
3809 Description: QOS code breaks on skipped indicies
3810 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3811 created with --index need not be sequential.
3813 Severity : enhancement
3815 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3816 Details : The uninit_groups feature works in conjunction with the kernel
3817 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3818 up the pass1 processing of e2fsck. This is a read-only feature
3819 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3820 mount filesystems that have had this feature enabled.
3822 Severity : enhancement
3824 Description: Improve multi-block allocation algorithm to avoid fragmentation
3825 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3826 allocation locality and avoid filesystem fragmentation.
3828 ------------------------------------------------------------------------------
3830 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3832 * Support for kernels:
3833 2.4.21-47.0.1.EL (RHEL 3)
3834 2.6.5-7.283 (SLES 9)
3835 2.6.9-42.0.10.EL (RHEL 4)
3836 2.6.12.6 vanilla (kernel.org)
3837 2.6.16.27-0.9 (SLES 10)
3838 * Recommended e2fsprogs version: 1.39.cfs5
3840 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3844 Frequency : occasional, depends on client load and configuration
3845 Bugzilla : 12181, 12203
3846 Description: data loss for recently-modified files
3848 Details : In some cases it is possible that recently written or created
3849 files may not be written to disk in a timely manner (this should
3850 normally be within 30s unless client IO load is very high).
3851 The problem appears as zero-length files or files that are a
3852 multiple of 1MB in size after a client crash or client eviction
3853 that are missing data at the end of the file.
3855 This problem is more likely to be hit on clients where files are
3856 repeatedly created and unlinked in the same directory, clients
3857 have a large amount of RAM, have many CPUs, the filesystem has
3858 many OSTs, the clients are rebooted frequently, and/or the files
3859 are not accessed by other nodes after being written.
3861 The presence of the problem can be detected by looking at
3862 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3863 smaller than the second (nr_unused) then dirty files will not
3864 be flushed automatically to disk. "sync; sleep 10" should be
3865 run several times on the node before unmounting it to update
3866 Lustre (this is also safe to run on nodes without this problem).
3868 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3869 that can cause this same problem, so customers using that kernel
3870 also need to update the kernel in addition to Lustre. In order
3871 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3873 It is normal that files written just before a client crash (less
3874 than 30s) may not yet have been flushed to disk, even for local
3878 Frequency : frequent on thin XT3 nodes
3880 Description: UUID collision on thin XT3 Linux nodes
3881 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3882 randomly, since we relied on an insufficiently-seeded PRNG.
3887 Description: OSS hangs after "All ost request buffers busy"
3888 Details : A deadlock between quota and journal operations caused OSS
3889 hangs after printing "All ost request buffers busy."
3892 Frequency : always on liblustre builds
3894 Description: Cleanup compiler warnings on liblustre
3897 Frequency : always on liblustre builds on XT3
3899 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3903 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3904 > 32000 subdirectories
3905 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3906 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3907 subdirectory count crosses 32000. This will aid e2fsck to
3908 correctly handle more than 32000 subdirectories.
3911 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3913 Description: System hang on startup
3914 Details : This bug allowed the liblustre (e.g. catamount) client to
3915 return to the app before handling all startup RPCs. This
3916 could leave the node unresponsive to lustre network traffic
3917 and manifested as a server ptllnd timeout.
3919 Severity : enhancement
3921 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3922 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3923 boolean to control whether to print peer debug info when a
3924 client's RPC times out.
3929 Description: make O_SYNC working on 2.6 kernels
3930 Details : 2.6 kernels use different method for mark pages for write,
3931 so need add a code to lustre for O_SYNC work.
3936 Description: Failure to close file and release space on NFS
3937 Details : Put inode details into lock acquired in ll_intent_file_open.
3938 Use mdc_intent_lock in ll_intent_open to properly
3939 detect all kind of errors unhandled by mdc_enqueue.
3944 Description: proc file read during shutdown sometimes raced obd removal,
3946 Details : Add lock to prevent obd access after proc file removal.
3949 Frequency : Only for files larger than 4GB on 32-bit clients.
3951 Description: improperly doing page alignment of locks
3952 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3953 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3958 Description: RPCs being resent when they shouldn't be
3959 Details : Some RPCs that should not be resent are being resent. This
3960 can cause inconsistencies in the RPC state machine. Do not
3961 resend such requests.
3964 Frequency : rare, only with NFS export
3966 Description: Crash on NFS re-export node
3967 Details : under very unusual load conditions an assertion is hit in
3968 ll_intent_file_open()
3971 Frequency : only if OST filesystem is corrupted
3973 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3974 Details : for a short time RPCs with bulk IO are in the replay list,
3975 but replay of bulk IOs is unimplemented. If the OST filesystem
3976 is corrupted due to disk cache incoherency and then replay is
3977 started it is possible to trip an assertion. Avoid putting
3978 committed RPCs into the replay list at all to avoid this issue.
3983 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3984 Details : Large single O_DIRECT read and write calls can fail to allocate
3985 a sufficiently large buffer to process the request. In case of
3986 allocation failure the allocation is retried with a smaller
3987 buffer and broken into smaller requests.
3989 Severity : enhancement
3991 Description: Add -o localflock option to simulate old noflock behaviour.
3992 Details : This will achieve local-only flock/fcntl locks coherentness.
3997 Description: versioning check is incomplete
3998 Details : Checking the version difference of client vs. server, report
3999 error if the gap is too big.
4004 Description: add support PG_writeback bit
4005 Details : add support for PG_writeback bit for Lustre, for more carefull
4006 work with page cache in 2.6 kernel. This also fix some deadlocks
4007 and remove hack for work O_SYNC with 2.6 kernel.
4009 Severity : enhancement
4011 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4012 Details : The uninit_groups feature works in conjunction with the kernel
4013 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4014 up the pass1 processing of e2fsck. This is a read-only feature
4015 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4016 mount filesystems that have had this feature enabled.
4018 Severity : enhancement
4020 Description: Improve multi-block allocation algorithm to avoid fragmentation
4021 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4022 allocation locality and avoid filesystem fragmentation.
4024 ------------------------------------------------------------------------------
4026 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4028 * Support for kernels:
4029 2.6.9-42.0.3.EL (RHEL 4)
4030 2.6.5-7.276 (SLES 9)
4031 2.4.21-47.0.1.EL (RHEL 3)
4032 2.6.12.6 vanilla (kernel.org)
4033 2.6.16.21-0.8 (SLES10)
4034 * Recommended e2fsprogs version: 1.39.cfs2-0
4036 * The backwards-compatible /proc/sys/portals symlink has been removed
4037 in this release. Before upgrading, please ensure that you change
4038 any configuration scripts or /etc/sysctl.conf files that access
4039 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4040 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4041 in advance of the upgrade on any system running Lustre 1.4.6 or
4042 newer, since /proc/sys/lnet was added in that version.
4043 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4047 Frequency : only when quota is used
4049 Description: avoid scanning export list for quota master
4050 Details : Change the algorithms to avoid scanning export list in order
4051 to improve the efficiency.
4054 Frequency : MDS failover only, very rarely
4056 Description: "went back in time" messages on mds failover
4057 Details : The greatest transno may be lost when the current operation
4058 finishes with an error (transno==0) and the client's last_rcvd
4059 record is over-written. Save the greatest transno in the
4060 mds_last_transno for this case.
4063 Frequency : always for specific kernels and striping counts
4065 Description: client may get "Matching packet too big" without ACL support
4066 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4067 when trying to access files in certain configurations. The
4068 clients should in fact be denied when mounting because they do
4069 not understand ACLs.
4072 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4074 Description: many clients connecting with IO in progress causes connect timeouts
4075 Details : Avoid synchronous journal commits to avoid delays caused by many
4076 clients connecting/disconnecting when bulk IO is in progress.
4077 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4078 OST_IO_PORTAL to avoid delays behind potentially many pending
4082 Frequency : occasionally with multiple writers to a single file
4084 Description: shared writes to file may result in wrong size reported by stat()
4085 Details : Allow growing of kms when extent lock is cancelled
4088 Frequency : always with random mmap IO to multi-striped file
4090 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4091 Details : Only if the hole is at the end of OST object so that kms is too
4092 small. Fix is to increase kms accordingly in ll_nopage.
4095 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4097 Description: writes to a missing object would leak memory on the OST
4098 Details : If there is an inconsistency between the MDS and OST filesystems,
4099 such that the MDS references an object that doesn't exist, writes
4100 to that object will leak memory due to incorrect cleanup in the
4101 error handling path, eventually running out of memory on the OST.
4106 Description: Creating too long symlink causes lustre errors
4107 Details : Check symlink and name lengths before sending requests to MDS.
4110 Frequency : only if flock is enabled (not on by default)
4112 Description: posix locks not released on fd closure on 2.6.9+
4113 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4114 kernels, this caused such locks not to be released on fd close and
4115 then assertions on fs unmount about still used locks.
4118 Frequency : MDS failover only, very rarely
4120 Description: clients may get ASSERTION(granted_lock != NULL)
4121 Details : When request was taking a long time, and a client was resending
4122 a getattr by name lock request. The were multiple lock requests
4123 with the same client lock handle and
4124 mds_getattr_name->fixup_handle_for_resent_request found one of the
4125 lock handles but later failed with ASSERTION(granted_lock != NULL).
4130 Description: handle->h_buffer_credits > 0, assertion failure
4131 Details : h_buffer_credits is zero after truncate, causing assertion
4132 failure. This patch extends the transaction or creates a new
4136 Frequency : NFS re-export or patchless client
4137 Bugzilla : 11179, 10796
4138 Description: Crash on NFS re-export node (__d_move)
4139 Details : We do not want to hash the dentry if we don't have a lock.
4140 But if this dentry is later used in d_move, we'd hit uninitialised
4141 list head d_hash, so we just do this to init d_hash field but
4142 leave dentry unhashed.
4145 Frequency : NFS re-export or patchless client
4147 Description: NFS exports has problem with symbolic link
4148 Details : lustre client didn't properly install dentry when re-exported
4149 to NFS or running patchless client.
4152 Frequency : NFS re-export or patchless client
4154 Description: Various nfs/patchless fixes.
4155 Details : fixes reuse disconected alias for lookup process - this fixes
4156 warning "find_exported_dentry: npd != pd",
4157 fix permission error with open files at nfs.
4158 fix apply umask when do revalidate.
4161 Frequency : occasional
4163 Description: Crash on NFS re-export node
4164 Details : calling clear_page() on the wrong pointer triggered oops in
4165 generic_mapping_read().
4168 Frequency : rarely, using O_DIRECT IO
4170 Description: unaligned directio crashes client with LASSERT
4171 Details : check for unaligned buffers before trying any requests.
4174 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4176 Description: stale data returned from RAID cache
4177 Details : If only a small amount of IO is done to the RAID device before
4178 reading it again it is possible to get stale data from the RAID
4179 cache instead of reading it from disk.
4182 Frequency : always for sles10 kernel
4184 Description: sles10 support
4185 Details : ll_follow_link: compile fixes and using of nd_set_link
4186 under newer kernels.
4189 Frequency : depends on arch, kernel and compiler version, always on sles10
4192 Description: recursive or deep enough symlinks cause stack overflow
4193 Details : getting rid of large stack-allocated variable in
4197 Frequency : depends on hardware
4199 Description: lustre write performance loss in the SLES10 kernel
4200 Details : the performance loss is caused by using of write barriers in the
4201 ext3 code. The SLES10 kernel turns barrier support on by
4202 default. The fix is to undo that change for ldiskfs.
4204 ------------------------------------------------------------------------------
4206 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4208 * Support for kernels:
4209 2.6.9-42.0.3EL (RHEL 4)
4210 2.6.5-7.276 (SLES 9)
4211 2.4.21-47.0.1.EL (RHEL 3)
4212 2.6.12.6 vanilla (kernel.org)
4216 Frequency : quota enabled and large files being deleted
4218 Description: releasing more than 4GB of quota at once hangs OST
4219 Details : If a user deletes more than 4GB of files on a single OST it
4220 will cause the OST to spin in an infinite loop. Release
4221 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4226 Description: statfs data retrieved from /proc may be stale or zero
4227 Details : When reading per-device statfs data from /proc, in the
4228 {kbytes,files}_{total,free,avail} files, it may appear
4229 as zero or be out of date.
4232 Frequency : systems with MD RAID1 external journal devices
4234 Description: lconf's call to blkid is confused by RAID1 journal devices
4235 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4236 of returning all block devices that match the journal UUID.
4239 Frequency : always, for aggregate stripe size over 4GB
4241 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4242 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4243 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4246 Frequency : always if "lfs find" used on a local file/directory
4248 Description: "lfs find" segfaults if used on a local file/directory
4249 Details : The case where a directory component was not specified wasn't
4250 handled correctly. Handle this properly.
4253 Frequency : always on ppc64
4255 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4256 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4257 perform endianness conversion.
4260 Frequency : rarely (truncate to non-zero file size after write under load)
4261 Bugzilla : 10730, 10687
4262 Description: Files padded with zeros to next 4K multiple
4263 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4264 it is possible that files that are truncated to a non-zero size
4265 immediately after being written are filled with zero bytes beyond
4266 the truncated size. No file data is lost.
4268 Severity : enhancement
4270 Description: Allow recovery/failover for liblustre clients.
4271 Details : liblustre clients were unaware of failover configurations until
4274 Severity : enhancement
4276 Description: user file locks should fail when not mounting with flock option
4277 Details : Set up an error-returning stub in ll_file_operations.lock field
4278 to prevent incorrect behaviour when client is mounted without
4279 flock option. Also, set up properly f_op->flock field for
4283 Frequency : always on ia64
4285 Description: "lfs df" loops on printing out MDS statfs information
4286 Details : The obd_ioctl_data was not initialized and in some systems
4287 this caused a failure during the ioctl that did not return
4288 an error. Initialize the struct and return an error on failure.
4291 Frequency : SLES 9 only
4293 Description: Error of copying files with lustre special EAs as root
4294 Details : Client side always return success for setxattr call for lustre
4295 special xattr (currently only "trusted.lov").
4298 Frequency : rarely on clusters with both ia64+i386 clients
4300 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4301 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4302 clients are doing concurrent writes to the same file it is
4303 possible that the ia64 clients may LASSERT because the OST
4304 extent locks are not PAGE_SIZE aligned. Ensure that grown
4305 locks are always aligned on the request boundary.
4308 Frequency : specific use, occasional
4310 Description: Overwriting in use executable truncates on-disk binary image
4311 Details : If one node attempts to overwrite an executable in use by
4312 another node, we now correctly return ETXTBSY instead of
4313 truncating the file.
4315 Severity : enhancement
4317 Description: Async OSC create to avoid the blocking unnecessarily.
4318 Details : If an OST has no remaining object, system will block on the creating
4319 when need to create a new object on this OST. Now, ways use
4320 pre-created objects when available, instead of blocking on an
4321 empty osc while others are not empty. If we must block, we block
4322 for the shortest possible period of time.
4327 Description: chmod on Lustre root is propagated to other clients
4328 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4329 invalid by the follow_mount time.
4332 Frequency : liblustre clients only
4334 Description: Race in 'instant cancel' lock handling could lead to such locks
4335 never to be granted in case of SMP MDS
4336 Details : Do not destroy not yet granted but cbpending locks in
4340 Frequency : replay/resend of open
4342 Description: non null lock assetion failure in mds_intent_policy
4343 Details : Trying to replay/resend lockless open requests resulted in
4344 mds_open() returning 0 with no lock. Now it sets a flag if
4345 a lock is going to be returned.
4347 Severity : enhancement
4349 Description: Checksum enhancements
4350 Details : New checksum enhancements allow for resending RPCs that failed
4353 Severity : enhancement
4355 Description: Tunables on number of dirty pages in cacche
4356 Details : Allow to set limit on number of dirty pages cached.
4361 Description: client crash on unmount - lock still has references
4362 Details : In some error handling cases it was possible to leak a lock
4363 reference on a client while accessing a file. This was not
4364 harmful to the client during operation, but would cause the
4365 client to crash when the filesystem is unmounted.
4368 Frequency : specific case, rare
4370 Description: ETXTBSY on mds though file not in use by client
4371 Details : ETXTBSY is no longer incorrectly returned when attempting to
4372 chmod or chown a directory that the user previously tried to
4373 execute or a currently-executing binary.
4376 Frequency : extremely rare except on liblustre-based clients
4378 Description: Lustre space not freed when files are deleted
4379 Details : Clean up open-unlinked files after client eviction. Previously
4380 the unlink was skipped and the files remained as orphans.
4385 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4386 Details : In some cases it was possible to send a blocking callback to
4387 a client doing a glimpse, even though that client didn't get
4388 a lock granted. When the glimpse lock is cancelled on the OST
4389 the freed lock is left on the waiting list and corrupted the list.
4392 Frequency : all core dumps
4394 Description: Broke core dumps to lustre
4395 Details : Negative dentry may be unhashed if parent does not have UPDATE
4396 lock, but some callers, e.g. do_coredump, expect dentry to be
4397 hashed after successful create, hash it in ll_create_it.
4399 ------------------------------------------------------------------------------
4401 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4403 * Support for kernels:
4404 2.6.9-42.0.2.EL (RHEL 4)
4405 2.6.5-7.276 (SLES 9)
4406 2.4.21-40.EL (RHEL 3)
4407 2.6.12.6 vanilla (kernel.org)
4411 Frequency : always on RHEL 3
4413 Description: Number of open files grows over time
4414 Details : The number of open files grows over time, whether or not
4415 Lustre is started. This was due to a filp leak introduced
4416 by one of our kernel patches.
4418 ------------------------------------------------------------------------------
4420 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4422 * Support for kernels:
4423 2.6.9-42.EL (RHEL 4)
4424 2.6.5-7.276 (SLES 9)
4425 2.4.21-40.EL (RHEL 3)
4426 2.6.12.6 vanilla (kernel.org)
4431 Bugzilla : 5719, 9635, 9792, 9684
4432 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4433 Details : If a server is under heavy load and cannot reply to new
4434 connection requests before the client resends the (re)connect,
4435 the connection handling code can behave badly if two service
4436 threads are concurrently handing separate (re)connections from
4437 the same client. Add better locking to the connection handling
4438 code, and ensure that only a single connection will be processed
4439 for a given client UUID, even if the lock is dropped.
4441 Severity : enhancement
4443 Description: add TCP zero-copy support to kernel
4444 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4445 sends if the hardware supports scatter-gather and checksumming.
4446 This allows socklnd to do client-write and server-read more
4447 efficiently and reduce CPU utilization from skbuf copying.
4450 Frequency : only if NFS exporting from client
4452 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4453 Details : If setting attributes on a file created under NFS that had
4454 never been opened it would be possible to oops the client
4455 if the file had no objects.
4458 Frequency : always for liblustre
4460 Description: liblustre client does MDS+OSTs setattr RPC for each write
4461 Details : When doing a write from a liblustre client, the client
4462 incorrectly issued an RPC to the MDS and each OST the file was
4463 striped over in order to update the timestamps. When writing
4464 with small chunks and many clients this could overwhelm the MDS
4465 with RPCs. In all cases it would slow down the write because
4466 these RPCs are unnecessary.
4468 Severity : enhancement
4470 Description: allow number of MDS service threads to be changed at module load
4471 Details : It is now possible to change the number of MDS service threads
4472 running. Adding "options mds mds_num_threads={N}" to the MDS's
4473 /etc/modprobe.conf will set the number of threads for the next
4474 time Lustre is restarted (assuming the "mds" module is also
4475 reloaded at that time). The default number of threads will
4476 stay the same, 32 for most systems.
4481 Description: OST crash if filesystem is unformatted or corrupt
4482 Details : If an OST is started on a device that has never been formatted
4483 or if the filesystem is corrupt and cannot even mount then the
4484 error handling cleanup routines would dereference a NULL pointer.
4489 Description: NULL pointer deref in llap_from_page.
4490 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4491 incorrect) mapping if the page was truncated/reclaimed while it was
4492 searched for. Check for this condition and skip such pages when
4493 doing readahead. Introduce extra check to llap_from_page() to
4494 verify page->mapping->host is non-NULL (so page is not anonymous).
4497 Frequency : Sometimes when using sys_sendfile
4499 Description: "page not covered by a lock" warnings from ll_readpage
4500 Details : sendfile called ll_readpage without right page locks present.
4501 Now we introduced ll_file_sendfile that does necessary locking
4502 around call to generic_file_sendfile() much like we do in
4506 Frequency : with certain MDS communication failures at client mount time
4508 Description: NULL pointer deref after failed client mount
4509 Details : a client connection request may delayed by the network layer
4510 and not be sent until after the PTLRPC layer has timed out the
4511 request. If the client fails the mount immediately it will try
4512 to clean up before the network times out the request. Add a
4513 reference from the request import to the obd device and delay
4514 the cleanup until the network drops the request.
4517 Frequency : occasionally during client (re)connect
4519 Description: assertion failure during client (re)connect
4520 Details : processing a client connection request may be delayed by the
4521 client or server longer than the client connect timeout. This
4522 causes the client to resend the connection request. If the
4523 original connection request is replied in this interval, the
4524 client may trip an assertion failure in ptlrpc_connect_interpret()
4525 which thought it would be the only running connect process.
4528 Frequency : only with obd_echo servers and clients that are rebooted
4530 Description: kernel BUG accessing uninitialized data structure
4531 Details : When running an obd_echo server it did not start the ping_evictor
4532 thread, and when a client was evicted an uninitialized data
4533 structure was accessed. Start the ping_evictor in the RPC
4534 service startup instead of the OBD startup.
4536 Severity : enhancement
4537 Bugzilla : 10193 (patchless)
4538 Description: Remove dependency on various unexported kernel interfaces.
4539 Details : No longer need reparent_to_init, exit_mm, exit_files,
4540 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4543 Frequency : rare (only users of deprecated and unsupported LDAP config)
4545 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4546 Details : LDAP apparently contains 'lustreName' attributes instead of
4547 'name'. A simple remapping of the name is sufficient.
4550 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4552 Description: waiting_locks_callback trips kernel BUG if client is evicted
4553 Details : Running with the dump_on_timeout debug flag turned on makes
4554 it possible that the waiting_locks_callback() can try to dump
4555 the Lustre kernel debug logs from an interrupt handler. Defer
4556 this log dumping to the expired_lock_main() thread.
4558 Severity : enhancement
4560 Description: Support NFS exporting on 2.6 kernels.
4561 Details : Implement non-rawops metadata methods for NFS server to use without
4562 changing NFS server code.
4565 Frequency : very rare (synthetic metadata workload only)
4567 Description: two racing renames might cause an MDS thread to deadlock
4568 Details : Running the "racer" program may cause one MDS thread to rename
4569 a file from being the source of a rename to being the target of
4570 a rename at exactly the same time that another thread is doing
4571 so, and the second thread has already enqueued these locks after
4572 doing a lookup of the target and is trying to relock them in
4573 order. Ensure that we don't try to re-lock the same resource.
4576 Frequency : only very large systems with liblustre clients
4578 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4579 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4580 all OSTs in parallel. This allows the eviction of stale liblustre
4581 clients to proceed much faster than if they were done in series,
4582 and also offers similar improvements for other set_info RPCs.
4587 Description: excessive CPU usage during initial read phase on client
4588 Details : During the initial read phase on a client, it would agressively
4589 retry readahead on the file, consuming too much CPU and impacting
4590 performance (since 1.4.5.8). Improve the readahead algorithm
4591 to avoid this, and also improve some other common cases (read
4592 of small files in particular, where "small" is files smaller than
4593 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4598 Description: MDS crash when receiving packet with unknown intent.
4599 Details : Do not LBUG in unknown intent case, just return -EFAULT
4601 Severity : enhancement
4602 Bugzilla : 9293, 9385
4603 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4604 Details : Do not serialize getattr (non-intent version) and statfs.
4607 Frequency : occasional, when OST network is overloaded/intermittent
4609 Description: client evicted by OST after bulk IO timeout
4610 Details : If a client sends a bulk IO request (read or write) the OST
4611 may evict the client if it is unresposive to its data GET/PUT
4612 request. This is incorrect if the network is overloaded (takes
4613 too long to transfer the RPC data) or dropped the OST GET/PUT
4614 request. There is no need to evict the client at all, since
4615 the pinger and/or lock callbacks will handle this, and the
4616 client can restart the bulk request.
4619 Frequency : Always when mmapping file with no objects
4621 Description: client crashes when mmapping file with no objects
4622 Details : Check that we actually have objects in a file before doing any
4623 operations on objects in ll_vm_open, ll_vm_close and
4629 Description: Request leak when working with deleted CWD
4630 Details : Introduce advanced request refcount tracking for requests
4631 referenced from lustre intent.
4633 Severity : Enhancement
4635 Description: Cache open file handles on client.
4636 Details : MDS now will return special lock along with openhandle, if
4637 requested and client is allowed to hold openhandle, even if unused,
4638 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4639 closing files for every read/write openration.
4641 Severity : Enhancement
4643 Description: Cache open negative dentries on client when possible.
4644 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4645 negative dentries on lock revocation.
4650 Description: Remounting a client read-only wasn't possible with a zconf mount
4651 Details : It wasn't possible to remount a client read-only with llmount.
4653 Severity : enhancement
4654 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4655 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4656 MPICH is included to add Lustre-specific ADIO interfaces.
4657 This is based closely on the UFS ADIO layer and only differs
4658 in file creation, in order to allow the OST striping to be set.
4659 This is user-contributed code and not supported by CFS.
4664 Description: extended inode attributes (immutable, append-only) work improperly
4665 when 2.4 and 2.6 kernels are used on client/server or vice versa
4666 Details : Introduce kernel-independent values for these flags.
4668 Severity : enhancement
4671 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4672 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4673 granularity. Now, Fractional MB granularity is be supported,
4674 this is very useful for low memory system.
4676 Severity : enhancement
4678 Description: Getattr by fid
4679 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4680 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4681 have locks and avoids getattr rpc requests for it on every stat.
4684 Frequency : Always, for filesystems larger than 2TB
4686 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4687 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4688 filesystems larger than 2TB to be mounted with mballoc enabled.
4691 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4693 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4694 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4695 filesystems larger than 2TB is mounted then the kernel will
4696 silently corrupt the start of the filesystem. CONFIG_LBD is
4697 enabled for all CFS-supported kernels, but the possibility of
4698 this happening with a modified kernel config exists.
4700 Severity : enhancement
4702 Description: add client O_DIRECT support for 2.6 kernels
4703 Details : It is now possible to do O_DIRECT reads and writes to files
4704 in the Lustre client mountpoint on 2.6 kernel clients.
4706 Severity : enhancement
4708 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4709 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4710 parallel, not waiting for response from every OST before sending
4711 a rpc to the next OST.
4716 Description: setattr vs write race when updating file timestamps
4717 Details : Client processes that update a file timestamp into the past
4718 right after writing to the file (e.g. tar) it is possible that
4719 the updated file modification time can be reset to the current
4720 time due to a race between processing the setattr and write RPC.
4722 Severity : enhancement
4724 Description: Bring 'lfs find' closer in line with regular Linux find.
4725 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4726 -print0 options and obtains all the needed info through the lustre
4729 Severity : enhancement
4731 Description: support up to 1024 configured devices on one node
4732 Details : change obd_dev array from statically allocated to dynamically
4733 allocated structs as they are first used to reduce memory usage
4738 Description: Flush dirty partially truncated pages during truncate
4739 Details : Immediatelly flush partially truncated pages in filter_setattr,
4740 this way we completely avoid having any pages in page cache on OST
4741 and can retire ugly workarounds during writes to flush such pages.
4746 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4747 Details : protect inode from truncation within vfs_unlink() context
4748 just take a reference before calling vfs_unlink() and release it
4749 when parent's i_sem is free.
4754 Description: last_id value checked outside lock on OST caused LASSERT failure
4755 Details : If there were multiple MDS->OST object precreate requests in
4756 flight, it was possible that the OST's last object id was checked
4757 outside a lock and incorrectly tripped an assertion. Move checks
4758 inside locks, and discard old precreate requests.
4761 Frequency : always, if extents are used on OSTs
4763 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4764 Details : OSTs using the extents format would not zero the high 16 bits of
4765 the index physical block number. This is not a problem for any
4766 OST filesystems smaller than 16TB, and no kernels support ext3
4767 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4768 new/modified files) and can be fixed for existing filesystems
4769 with e2fsprogs-1.39-cfs1.
4774 Description: import connection selection may be incorrect if timer wraps
4775 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4776 import connections to be ignored if the 32-bit jiffies counter
4777 wraps. Use a 64-bit jiffies counter.
4780 Frequency : very large clusters immediately after boot
4782 Description: LNET request buffers exhausted under heavy short-term load
4783 Details : If a large number of client requests are generated on a service
4784 that has previously never seen so many requests it is possible
4785 that the request buffer growth cannot keep up with the spike in
4786 demand. Instead of dropping incoming requests, they are held in
4787 the LND until the RPC service can accept more requests.
4790 Frequency : Sometimes during replay
4792 Description: Assertion failure in ll_local_open after replay.
4793 Details : If replay happened on an open request reply before we were able
4794 to set replay handler, reply will become not swabbed tripping the
4795 assertion in ll_local_open. Now we set the handler right after
4796 recognising of open request
4799 Frequency : very rare
4801 Description: kernel reports "badness in vsnprintf"
4802 Details : Reading from the "recovery_status" /proc file in small chunks
4803 may cause a negative length in lprocfs_obd_rd_recovery_status()
4804 call to vsnprintf() (which is otherwise harmless). Exit early
4805 if there is no more space in the output buffer.
4807 Severity : enhancement
4809 Description: clear OBD RPC statistics by writing to them
4810 Details : It is now possible to clear the OBD RPC statistics by writing
4811 to the "stats" file.
4816 Description: Client mtime is not the same on different clients after utimes
4817 Details : In some cases, the client was using the utimes() syscall on
4818 a file cached on another node. The clients now validate the
4819 ctime from the MDS + OSTs to determine which one is right.
4824 Description: Inability to activate failout mode
4825 Details : lconf script incorrectly assumed that in pythong string's numeric
4826 value is used in comparisons.
4829 Frequency : always with multiple stripes per file
4831 Description: Inefficient object allocation for mutli-stripe files
4832 Details : When selecting which OSTs to stripe files over, for files with
4833 a stripe count that divides evenly into the number of OSTs,
4834 the MDS is always picking the same starting OST for each file.
4835 Return the OST selection heuristic to the original design.
4840 Description: mount failures may take full timeout to return an error
4841 Details : Under some heavy load conditions it is possible that a
4842 failed mount can wait for the full obd_timeout interval,
4843 possibly several minutes, before reporting an error.
4844 Instead return an error as soon as the status is known.
4846 Frequency : quota enabled and large files being deleted
4848 Description: releasing more than 4GB of quota at once hangs OST
4849 Details : If a user deletes more than 4GB of files on a single OST it
4850 will cause the OST to spin in an infinite loop. Release
4851 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4856 Description: statfs data retrieved from /proc may be stale or zero
4857 Details : When reading per-device statfs data from /proc, in the
4858 {kbytes,files}_{total,free,avail} files, it may appear
4859 as zero or be out of date.
4862 Frequency : systems with MD RAID1 external journal devices
4864 Description: lconf's call to blkid is confused by RAID1 journal devices
4865 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4866 of returning all block devices that match the journal UUID.
4869 Frequency : always, for aggregate stripe size over 4GB
4871 Description: assertion fails when trying to use 4GB stripe size
4872 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4873 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4876 Frequency : always on ppc64
4878 Description: the first write on an ext3 filesystem with mballoc got stuck
4879 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4880 perform endianness conversion.
4882 ------------------------------------------------------------------------------
4884 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4886 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4887 INTEROPERATE with older versions automatically. Please read the
4888 user documentation before upgrading any part of a live system.
4889 * WARNING: Lustre networking configuration changes are required with
4890 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4893 * Support for kernels:
4894 2.6.9-22.0.2.EL (RHEL 4)
4895 2.6.5-7.244 (SLES 9)
4896 2.6.12.6 vanilla (kernel.org)
4899 Severity : enhancement
4900 Bugzilla : 7981/8208
4901 Description: Introduced Lustre Networking (LNET)
4902 Details : LNET is new networking infrastructure for Lustre, it includes
4903 a reorganized network configuration mode (see the user
4904 documentation for full details) as well as support for routing
4905 between different network fabrics. Lustre Networking Devices
4906 (LNDs) for the supported network fabrics have also been
4907 created for this new infrastructure.
4909 Severity : enhancement
4910 Description: Introduced Access control lists
4911 Details : clients can set ACLs on files and directories in order to have
4912 more fine-grained permissions than the standard Unix UGO+RWX.
4913 The MDS must be started with the "-o acl" mount option.
4915 Severity : enhancement
4916 Description: Introduced filesystem quotas
4917 Details : Administrators may now establish per-user quotas on the
4920 Severity : enhancement
4922 Description: Configuration change for the XT3
4923 The PTLLND is now used to run Lustre over Portals on the XT3
4924 The configure option(s) --with-cray-portals are no longer used.
4925 Rather --with-portals=<path-to-portals-includes> is used to
4926 enable building on the XT3. In addition to enable XT3 specific
4927 features the option --enable-cray-xt3 must be used.
4932 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4933 Details : A race between allocating a new llcd and re-getting the llcd_lock
4934 allowed another thread to grab newly-allocated llcd.
4936 Severity : enhancement
4938 Description: 2.6 OST async journal commit and locking fix to improve performance
4939 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4940 2.6 kernels are now async as they already were in 2.4 kernels so
4941 that they can commit concurrently with the network bulk transfer.
4942 For block-allocated files the filter allocation semaphore is held
4943 to avoid filesystem fragmentation during allocation. BKL lock
4944 removed for 2.6 xattr operations where it is no longer needed.
4949 Description: lconf incorrectly determined whether two IP networks could talk
4950 Details : In some more complicated routing and multiple-network
4951 configurations, lconf will avoid trying to make a network
4952 connection to a disjoint part of the IP space. It was doing the
4953 math incorrectly for one set of cases.
4958 Description: Fix for potential infinite loop processing records in an llog.
4959 Details : If an llog record is corrupted/zeroed, it is possible to loop
4960 forever in llog_process(). Validate the llog record length
4961 and skip the remainder of the block on error.
4964 Frequency : occasional (liblustre only)
4966 Description: liblustre could not open files whose last component is a symlink
4967 Details : sysio_path_walk() would incorrectly pass the open intent to
4968 intermediate path components.
4971 Frequency : rare (liblustre only with non-standard tuning)
4972 Bugzilla : 7201 (7350)
4973 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4974 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4975 after it has been released, only do this for referenced locks
4977 Severity : enhancement
4979 Description: specifying an (invalid) directory default stripe_size of -1
4980 would reset the directory default striping
4981 Details : stripe_size -1 was used internally to signal directory stripe
4982 removal, now use "all default" to signal dir stripe removal
4983 as a directory striping of "all default" is not useful
4986 Frequency : common for large clusters running liblustre clients
4988 Description: doing an ls when liblustre clients are running is slow
4989 Details : sending a glimpse AST to a liblustre client waits for every AST
4990 to time out, as liblustre clients will not respond. Since they
4991 cannot cache data we refresh the OST lock LVB from disk instead.
4993 Severity : enhancement
4995 Description: doing an ls at the same time as file IO can be slow
4996 Details : enqueue and other "small" requests can be blocked behind many
4997 large IO requests. Create a new OST IO portal for non-IO
4998 requests so they can be processed faster.
5001 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5003 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5004 created file with mode 000
5005 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5006 allow an ftruncate() as a non-root user to a file with mode 000.
5007 Lustre now respects this flag to disable mode checks when
5008 truncating a file owned by the user
5011 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5014 Description: Revoking locks from clients that went dead or catatonic might take
5016 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5017 cancellation of such locks instant on servers without waiting for
5018 any reply from clients. Clients drops these locks when cancel
5019 notification from server is received without replying.
5022 Frequency : liblustre-only, when liblustre client dies or becomes busy
5024 Description: Doing ls on Linux clients can take a long time with active
5026 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5027 granting such locks to it in the first place if possible. Locks
5028 are taken by proxy on the OST during the read or write and
5029 dropped immediately afterward. Add connect flags handling, do
5030 not grant locks to liblustre clients for glimpse ASTs.
5032 Severity : enhancement
5034 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5035 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5036 sequential reads and behaves badly with applications doing random
5037 reads. Improve it by reading ahead at least read region, and
5038 avoiding excessive large RPC for small reads.
5040 Severity : enhancement
5042 Description: Creating more than 1000 files for a single job may cause a load
5043 imbalance on the OSTs if there are also a large number of OSTs.
5044 Details : qos_prep_create() uses an OST index reseed value that is an
5045 even multiple of the number of available OSTs so that if the
5046 reseed happens in the middle of the object allocation it will
5047 still utilize the OSTs as uniformly as possible.
5052 Description: OST or MDS may oops in ping_evictor_main()
5053 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5054 but doesn't restart at beginning of obd_exports_timed list
5057 Severity : enhancement
5059 Description: improve by-nid export eviction on the MDS and OST
5060 Details : allow multiple exports with the same NID to be evicted at one
5061 time without re-searching the exports list.
5064 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5066 Description: MDS may oops in groups_free()
5067 Details : in rare race conditions a newly allocated group_info struct is
5068 freed again, and this can be NULL. The 2.4 compatibility code
5069 for groups_free() checked for a NULL pointer, but 2.6 did not.
5072 Frequency : common for liblustre clients doing little filesystem IO
5073 Bugzilla : 9352, 7313
5074 Description: server may evict liblustre clients accessing contended locks
5075 Details : if a client is granted a lock or receives a completion AST
5076 with a blocking AST already set it would not reply to the AST
5077 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5080 Frequency : lfs setstripe, only systems with more than 160 OSTs
5082 Description: unable to set striping with a starting offset beyond OST 160
5083 Details : llapi_create_file() incorrectly limited the starting stripe
5084 index to the maximum single-file stripe count.
5087 Frequency : LDAP users only
5089 Description: lconf did not handle in-kernel recovery with LDAP properly
5090 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5092 Severity : enhancement
5094 Description: bind OST threads to NUMA nodes to improve performance
5095 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5096 node and do their allocations there to localize memory access
5098 Severity : enhancement
5100 Description: llmount can determine client NID directly from Myrinet (GM)
5101 Details : the client NID code from gmnalnid was moved directly into
5102 llmount, removing the need to use this or specifying the
5103 client NID explicitly when mounting GM clients with zeroconf
5106 Frequency : if client is started with down MDS
5108 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5109 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5110 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5111 ptlrpc_import_delayed_req() because we didn't check if the
5112 request was interrupted, and we also didn't break out of the
5113 event loop if there was no timeout
5118 Description: data loss during non-page-aligned writes to a single file from
5119 both multiple nodes and multiple threads on one node at same time
5120 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5121 inconsistency led to false short-reads, that were cached and later
5122 used by ->prepare_write() to fill in partially written page,
5123 leading to data loss.
5126 Frequency : always, if lconf --abort_recovery used
5128 Description: lconf --abort_recovery fails with 'Operation not supported'
5129 Details : lconf was attempting to abort recovery on the MDT device and not
5132 Severity : enhancement
5134 Description: remove cleanup logs
5135 Details : replace lconf-generated cleanup logs with lustre internal
5136 cleanup routines. Eliminates the need for client-cleanup and
5139 Severity : enhancement
5141 Description: add support for EAs (user and system) on lustre filesystems
5142 Details : it is now possible to store extended attributes in the Lustre
5143 client filesystem, and with the user_xattr mount option it
5144 is possible to allow users to store EAs on their files also
5146 Severity : enhancement
5148 Description: Add possibility (config option) to show minimal available OST free
5150 Details : When compiled with --enable-mindf configure option, statfs(2)
5151 (and so, df) will return least minimal free space available from
5152 all OSTs as amount of free space on FS, instead of summary of
5153 free spaces of all OSTs.
5155 Severity : enhancement
5157 Description: do not expand extent locks acquired on OST-side
5158 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5159 by server: they are not cached anyway.
5162 Frequency : when mmap is used/binaries executed from Lustre
5164 Description: Unmmap pages before throwing them away from read cache.
5165 Details : llap_shrink cache now attempts to unmap pages before discarding
5166 them (if unmapping failed - do not discard). SLES9 kernel has
5167 extra checks that trigger if this unmapping is not done first.
5172 Description: lconf didn't resolve symlinks before checking to see whether a
5173 given mountpoint was already in use
5176 Frequency : when migrating failover services
5177 Bugzilla : 6395, 9514
5178 Description: When migrating a subset of services from a node (e.g. failback
5179 from a failover service node) the remaining services would
5180 time out and evict clients.
5181 Details : lconf --force (implied by --failover) sets the global obd_timeout
5182 to 5 seconds in order to quickly disconnect, but this caused
5183 other RPCs to time out too quickly. Do not change the global
5184 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5186 Severity : enhancement
5187 Frequency : if MDS is started with down OST
5188 Bugzilla : 9439,5706
5189 Description: Allow startup/shutdown of an MDS without depending on the
5190 availability of the OSTs.
5191 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5192 Add appropriate locking and lov-osc refcounts for safe
5193 cleaning. Add osc abort_inflight calls in case the
5194 synchronize never started.
5197 Frequency : occasional (Cray XT3 only)
5199 Description: root not authorized to access files in CRAY_PORTALS environment
5200 Details : The client process capabilities were not honoured on the MDS in
5201 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5202 been accessed by an authorized user then root was able to access
5203 the file on the local client also. The root user capabilities
5204 are now allowed on the MDS, as this environment has secure UID.
5207 Frequency : occasional
5209 Description: ldiskfs "too long searching" message happens too often
5210 Details : A debugging message (otherwise harmless) prints too often on
5211 the OST console. This has been reduced to only happen when
5212 there are fragmentation problems on the filesystem.
5217 Description: Division by zero in statfs when all OSCs are inactive
5218 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5219 lov_statfs divides by value returned by lov_get_stripecnt().
5223 Bugzilla : 9489, 3273
5224 Description: First write from each client to each OST was only 4kB in size,
5225 to initialize client writeback cache, which caused sub-optimal
5226 RPCs and poor layout on disk for the first written file.
5227 Details : Clients now request an initial cache grant at (re)connect time
5228 and so that they can start streaming writes to the cache right
5229 away and always do full-sized RPCs if there is enough data.
5230 If the OST is rebooted the client also re-establishes its grant
5231 so that client cached writes will be honoured under the grant.
5236 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5237 Details : Now I/O RPCs go to different portal number and (presumably) fast
5238 lock requests (and glimses) and other RPCs get their own service
5239 threads pool that should be able to service those RPCs
5242 Severity : enhancement
5244 Description: Ability to exchange lustre version between client and servers and
5245 issue warnings at client side if client is too old. Also for
5246 liblustre clients there is ability to refuse connection of too old
5248 Details : New 'version' field is added to connect data structure that is
5249 filled with version info. That info is later checked by server and
5253 Frequency : rare, liblustre only.
5254 Bugzilla : 9296, 9581
5255 Description: Two simultaneous writes from liblustre at offset within same page
5256 might proceed at the same time overwriting eachother with stale
5258 Details : I/O lock withing llu_file_prwv was released too early, before data
5259 actually was hitting the wire. Extended lock-holding time until
5260 server acknowledges receiving data.
5263 Frequency : extremely rare. Never observed in practice.
5265 Description: avoid generating lustre_handle cookie of 0.
5266 Details : class_handle_hash() generates handle cookies by incrementing
5267 global counter, and can hit 0 occasionaly (this is unlikely, but
5268 not impossible, because initial value of cookie counter is
5269 selected randonly). Value of 0 is used as a sentinel meaning
5270 "unassigned handle" --- avoid it. Also coalesce two critical
5271 sections in this function into one.
5273 Severity : enhancement
5275 Description: allow liblustre clients to delegate truncate locking to OST
5276 Details : To avoid overhead of locking, liblustre client instructs OST to
5277 take extent lock in ost_punch() on client's behalf. New connection
5278 flag is added to handle backward compatibility.
5280 Severity : enhancement
5281 Bugzilla : 4928, 7341, 9758
5282 Description: allow number of OST service threads to be specified
5283 Details : a module parameter allows the number of OST service threads
5284 to be specified via "options ost ost_num_threads={N}" in the
5285 OSS's /etc/modules.conf or /etc/modprobe.conf.
5289 Bugzilla : 6146, 9635, 9895
5290 Description: servers crash with bad pointer in target_handle_connect()
5291 Details : In rare cases when a client is reconnecting it was possible that
5292 the connection request was the last reference for that export.
5293 We would temporarily drop the export reference and get a new
5294 one, but this may have been the last reference and the export
5295 was just destroyed. Get new reference before dropping old one.
5297 Severity : enhancement
5298 Frequency : if client is started with failover MDS
5300 Description: Allow multiple MDS hostnames in the mount command
5301 Details : Try to read the configuration from all specified MDS
5302 hostnames during a client mount in case the "primary"
5305 Severity : enhancement
5307 Description: Stop sending data to evicted clients as soon as possible.
5308 Details : Check if the client we are about to send or are sending data to
5309 was evicted already. (Check is done every second of waiting,
5310 for which l_wait_event interface was extended to allow checking
5311 of exit condition at specified intervals).
5314 Frequency : rare, normally only when NFS exporting is done from client
5316 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5318 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5321 Frequency : always, if srand() is called before liblustre initialization
5323 Description: Liblustre uses system PRNG disturbing its usage by user application
5324 Details : Introduce internal to lustre fast and high-quality PRNG for
5325 lustre usage and make liblustre and some other places in generic
5326 lustre code to use it.
5328 Severity : enhancement
5329 Bugzilla : 9477, 9557, 9870
5330 Description: Verify that the MDS configuration logs are updated when xml is
5331 Details : Check if the .xml configuration logs are newer than the config
5332 logs stored on the MDS and report an error if this is the case.
5333 Request --write-conf, or allow starting with --old_conf.
5335 Severity : enhancement
5337 Description: Handle symlinks in the path when checking if Lustre is mounted.
5338 Details : Resolve intermediate symlinks when checking if a client has
5339 mounted a filesystem to avoid duplicate client mounts.
5344 Description: lconf can hit an error exception but still return success.
5345 Details : The lconf command catches the Command error exception at the top
5346 level script context and will exit with the associated exit
5347 status, but doesn't ensure that this exit status is non-zero.
5352 Description: failure of ptlrpc thread startup can cause oops
5353 Details : Starting a ptlrpc service thread can fail if there are a large
5354 number of threads or the server memory is very fragmented.
5355 Handle this without oopsing.
5358 Frequency : always, only if liblustre and non-default acceptor port was used
5360 Description: liblustre cannot connect to servers with non-default acceptor port
5361 Details : tcpnal_set_default_params() was not called and was therefore
5362 ignoring the environment varaible TCPNAL_PORT, as well as other
5363 TCPNAL_ environment variables
5368 Description: two objects could be created on the same OST for a single file
5369 Details : If an OST is down, in some cases it was possible to create two
5370 objects on a single OST for a single file. No problems other
5371 than potential performance impact and spurious error messages.
5375 Bugzilla : 5681, 9562
5376 Description: Client may oops in ll_unhash_aliases
5377 Details : Client dcache may become inconsistent in race condition.
5378 In some cases "getcwd" can fail if the current directory is
5384 Description: Inode refcounting problems in NFS export code
5385 Details : link_raw functions used to call d_instantiate without obtaining
5386 extra inode reference first.
5390 Bugzilla : 9942, 9903
5391 Description: Referencing freed requests leading to crash, memleaks with NFS.
5392 Details : We used to require that call to ll_revalidate_it was always
5393 followed by ll_lookup_it. Also with revalidate_special() it is
5394 possible to call ll_revalidate_it() twice for the same dentry
5395 even if first occurence returned success. This fix changes semantic
5396 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5397 reference on a request referred from the intent.
5398 ll_intent_release() then releases such a request.
5401 Frequency : rare, normally benchmark loads only
5403 Description: unlinked inodes were kept in memory on the client
5404 Details : If a client is repeatedly creating and unlinking files it
5405 can accumulate a lot of stale inodes in the inode slab cache.
5406 If there is no other client load running this can cause the
5407 client node to run out of memory. Instead flush old inodes
5408 from client cache that have the same inode number as a new inode.
5411 Frequency : SLES9 2.6.5 kernel and long filenames only
5412 Bugzilla : 9969, 10379
5413 Description: utime reports stale NFS file handle
5414 Details : SLES9 uses out-of-dentry names in some cases, which confused
5415 the lustre dentry revalidation. Change it to always use the
5419 Frequency : rare, unless heavy write-truncate concurrency is continuous
5420 Bugzilla : 4180, 6984, 7171, 9963, 9331
5421 Description: OST becomes very slow and/or deadlocked during object unlink
5422 Details : filter_destroy() was holding onto the parent directory lock
5423 while truncating+unlinking objects. For very large objects this
5424 may block other threads for a long time and slow overall OST
5425 responsiveness. It may also be possible to get a lock ordering
5426 deadlock in this case, or run out of journal credits because of
5427 the combined truncate+unlink. Solution is to do object truncate
5428 first in one transaction without parent lock, and then do the
5429 final unlink in a new transaction with the parent lock. This
5430 reduces the lock hold time dramatically.
5433 Frequency : rare, 2.4 kernels only
5435 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5436 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5437 before unmounting their filesystem in order to avoid deadlock.
5438 The kernel_locked() function in 2.4 kernels only checks whether
5439 the kernel lock is held, not whether it is this process that is
5440 holding it as 2.6 kernels do.
5445 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5446 Details : The client ptlrpc code may be trying to reconnect to a down
5447 server before a previous connection attempt has timed out.
5448 Increase the reconnect interval to be longer than the connection
5449 timeout interval to avoid sending duplicate connections to
5453 Frequency : echo_client brw_test command
5455 Description: fix echo_client to work with OST preallocated code
5456 Details : OST preallocation code (5137) didn't take echo_client IO path
5457 into account: echo_client calls filter methods outside of any
5458 OST thread and, hence, there is no per-thread preallocated
5459 pages and buffers to use. Solution: hijack pga pages for IO. As
5460 a byproduct, this avoids unnecessary data copying.
5464 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5465 Description: Client can oops in mdc_commit_close() after open replay
5466 Details : It was possible for the MDS to return an open request with no
5467 transaction number in mds_finish_transno() if the client was
5468 evicted, but without actually returning an error. Clients
5469 would later try to replay that open and may trip an assertion
5470 Simplify the client close codepath, and always return an error
5471 from the MDS in case the open is not successful.
5474 Frequency : rare, 2.6 OSTs only
5476 Description: OST may deadlock under high load on fragmented files
5477 Details : If there was a heavy load and highly-fragmented OST filesystems
5478 it was possible to have all the OST threads deadlock waiting on
5479 allocation of biovecs, because the biovecs were not released
5480 until the entire RPC IO was completed. Instead, release biovecs
5481 as soon as they are complete to ensure forward IO progress.
5483 Severity : enhancement
5485 Description: Support for specifying external journal device at mount
5486 Details : If an OST or MDS device is formatted with an external journal
5487 device, this device major/minor is stored in the ext3 superblock
5488 and may not be valid for failover. Allow detecting and
5489 specifying the external journal at mount time.
5494 Description: Mounting an MDS with pending unlinked files may cause oops
5495 Details : target_finish_recovery() calls mds_postrecov() which returned
5496 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5497 considers this an error and immediately begins cleaning up the
5498 lov, just after starting the mds_lov process
5500 Severity : enhancement
5502 Description: Implement 'lfs df' to report actual free space on per-OST basis
5503 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5504 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5505 usage of objects; '-h' to report in human readable format.
5507 ------------------------------------------------------------------------------
5509 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5516 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5517 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5518 at startup during orphan unlinking. Instead, skip buddy update
5519 before setup, as it will be regenerated after recovery is complete.
5524 Description: If an OST is inactive, its locks might reference stale inodes.
5525 Details : lov_change_cbdata() must iterate over all namespaces, even if
5526 they are inactive to clear inode references from the lock.
5528 Severity : enhancement
5529 Frequency : occasional, if non-standard max_dirty_mb used
5531 Description: Client will block write RPCs if not enough grant
5532 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5533 then the client will block writes while waiting for another RPC
5534 to complete instead of consuming its dirty limit. With change
5535 we get improved performance when max_dirty_mb is small.
5537 Severity : enhancement
5538 Bugzilla : 3389, 6253
5539 Description: Add support for supplementary groups on the MDS.
5540 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5541 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5542 lookups for user supplementary groups into a cache.
5546 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5547 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5548 for liblustre O_EXCL is enforced by MDS.
5552 Description: readdir never returns NULL in liblustre.
5553 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5554 dentry in struct dirent.
5558 Frequency : liblustre only, depends on application IO pattern
5559 Description: liblustre clients evicted if not contacting servers
5560 Details : Don't put liblustre clients into the ping_evictor list, so
5561 they will not be evicted by the pinger ever.
5563 Severity : enhancement
5565 Description: Add ability to evict clients by NID from MDS.
5566 Details : By echoing "nid:$NID" string into
5567 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5568 $NID would be instantly evicted from this MDS and from all active
5569 OSTs connected to it.
5573 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5574 Details : lookup_it_finish() used to query file size from OSTs that was not
5579 Description: service threads change working directory to that of init
5580 Details : Starting lustre service threads may pin the working directory
5581 of the parent thread, making that filesystem busy. Threads
5582 now change to the working directory of init to avoid this.
5586 Frequency : during shutdown only
5587 Description: shutdown with a failed MDS or OST can cause unmount to hang
5588 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5593 Frequency : 2.6 only
5594 Description: chmod/chown may include an extra supplementary group
5595 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5596 supplementary group and if none is specified this is used.
5599 Bugzilla : 5479 (6816)
5601 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5602 Details : If lookup is in progress on a file that is unlinked we might try
5603 to revalidate the inode and fail in revalidate after lookup is
5604 complete and ll_file_open() enqueues the open again but
5605 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5608 Frequency : always, if lconf --abort_recovery used
5610 Description: lconf --abort_recovery fails with 'Operation not supported'
5611 Details : lconf was attempting to abort recovery on the MDT device and not
5614 ------------------------------------------------------------------------------
5616 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5621 Frequency : rare (only unsupported configurations with a node running as an
5623 Bugzilla : 6514, 5137
5624 Description: Mounting a Lustre file system on a node running as an OST could
5626 Details : OSTs now preallocates memory needed to write out data at
5627 startup, instead of when needed, to avoid having to
5628 allocate memory in possibly low memory situations.
5629 Specifically, if the file system is mounted on on OST,
5630 memory pressure could force it to try to write out data,
5631 which it needed to allocate memory to do. Due to the low
5632 memory, it would be unable to do so and the node would
5633 become unresponsive.
5635 Severity : enhancement
5637 Description: Addition of lconf --service command line option
5638 Details : lconf now accepts a '--service <arg>' option, which is
5639 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5641 Severity : enhancement
5643 Description: Failover mode is now the default for OSTs.
5644 Details : By default, OSTs will now run in failover mode. To return to
5645 the old behaviour, add '--failout' to the lmc line for OSTs.
5647 Severity : enhancement
5649 Description: Health checks are now provided for MDS and OSTs
5650 Details : Additional detailed health check information on MSD and OSTs
5651 is now provided through the procfs health_check value.
5654 Frequency : occasional, depends on IO load
5656 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5657 after numerous create/delete cycles
5658 Details : The ext3 inode allocation policy would not allocate new inodes
5659 very well on the OSTs because there are no new directories
5660 being created. Instead we look for groups with free space if
5661 the parent directories are nearly full.
5666 Description: Network or server problems during mount may cause partially
5667 mounted clients instead of returning an error.
5668 Details : The config llog parsing code may overwrite the error return
5669 code during mount error handling, returning success instead
5675 Description: MDS can fail to allocate large reply buffers
5676 Details : After long uptimes the MDS can fail to allocate large reply
5677 buffers (e.g. zconf client mount config records) due to memory
5678 fragmentation or consumption by the buffer cache. Preallocate
5679 some large reply buffers so that these replies can be sent even
5680 under memory pressure.
5684 Frequency : rare (liblustre)
5685 Description: fsx running with liblustre complained that using truncate() to
5686 extend the file doesn't work. This patch corrects that issue.
5687 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5688 ATTR_SIZE and lsm use in llu_setattr_raw.
5692 Frequency : rare, only 2.6 kernels
5693 Description: Unusual file access patterns on the MDS may result in inode
5694 data being lost in very rare circumstances.
5695 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5696 code in the 2.6 kernel caused the inode and/or EA data not to
5697 be read from disk, causing single-file corruption.
5701 Frequency : rare, only 2.6 filesystems using extents
5702 Description: Heavy concurrent write and delete load may cause data corruption.
5703 Details : It was possible under high-load situations to have an extent
5704 metadata block in the block device cache from a just-unlinked
5705 file overwrite a newly-allocated data block. We now unmap any
5706 metadata buffers that alias just-allocated data blocks.
5710 Frequency : filesystems with default stripe_count larger than 77
5711 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5712 Details : lconf specifies an inode size of 4096 bytes when the default
5713 stripe_count is larger than 77. This conflicts with the default
5714 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5715 this case to avoid pinning too much memory for large EAs.
5717 ------------------------------------------------------------------------------
5719 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5724 Frequency : rare (extremely heavy IO load with hundreds of clients)
5726 Description: Client is evicted, gets IO error writing to file
5727 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5728 caused clients to be evicted for AST timeouts. The fixes in
5729 bug 5192 mean we no longer need to have such short AST timeouts
5730 so ldlm_timeout has been increased.
5733 Frequency : occasional during --force or --failover shutdown under load
5734 Bugzilla : 5949, 4834
5735 Description: Server oops/LBUG if stopped with --force or --failover under load
5736 Details : a collection of import/export refcount and cleanup ordering
5737 issues fixed for safer force cleanup
5740 Frequency : only filesystems larger than 120 OSTs
5741 Bugzilla : 5990, 6223
5742 Description: lfs getstripe would oops on a very large filesystem
5743 Details : lov_getconfig used kfree on vmalloc'd memory
5746 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5747 Bugzilla : 6242, 6243
5748 Description: reading from files that had been truncated to a non-zero size
5749 but never opened returned no data
5750 Details : ll_file_read() reads zeros from no-object files to EOF
5755 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5756 Details : There's little harm in aborting MDS/OSS recovery and letting it
5757 try again, so I removed the LASSERT and return an error instead.
5759 Severity : enhancement
5761 Description: New debugging infrastructure for tracking down data corruption
5762 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5763 (b) cover more of the client-side code path, and (c) try to narrow
5764 down where problems occurred
5768 Bugzilla : 3819, 4364, 4397, 6313
5769 Description: Racing close and eviction MDS could cause assertion in mds_close
5770 Details : It was possible to get multiple mfd references during close and
5771 client eviction, leading to one thread referencing a freed mfd.
5773 Severity: : enhancement
5774 Bugzilla : 3262, 6359
5775 Description: Attempts to reconnect to servers are now more aggressive.
5776 Details : This builds on the enhanced upcall-less recovery that was added
5777 in 1.4.2. When trying to reconnect to servers, clients will
5778 now try each server in the failover group every 10 seconds. By
5779 default, clients would previously try one server every 25 seconds.
5784 Description: After recovery, certain operations trigger a failed
5785 assertion on a client.
5786 Details : Failing over an mds, using lconf -d --failover, while a
5787 client was doing a readdir() call would cause the client to
5788 LBUG after recovery completed and the readdir() was resent.
5790 Severity : enhancement
5792 Description: Default groups are now added by lconf
5793 Details : You can now run lconf --group <servicename> without having to
5794 manually add groups with lmc.
5797 Frequency : occasional
5799 Description: Nodes with an elan id of 0 trigger a failed assertion
5802 Frequency : always when accessing e.g. tty/console device nodes
5804 Description: tty and some other devices nodes cannot be used on lustre
5805 Details : file's private_data field is used by device data and lustre
5806 values in there got lost. New field was added to struct file to
5807 store fs-specific private data.
5810 Frequency : when exporting Lustre via NFS
5812 Description: NFSD failed occasionally when looking up a path component
5813 Details : NFSD is looking up ".." which was broken in ext3 directories
5814 that had grown large enough to become hashed.
5817 Frequency : Clusters with multiple interfaces not on the same subnet
5819 Description: Nodes will repeatedly try to reconnect to an interface which it
5820 cannot reach and report an error to the log.
5821 Details : Extra peer list entries will be created by lconf with some peers
5822 unreachable. lconf now validates the peer before adding it.
5825 Frequency : Only if a default stripe is set on the filesystem root.
5827 Description: Setting a default stripe on the filesystem root prevented the
5828 filesystem from being remounted.
5829 Details : The client was sending extra request flags in the root getattr
5830 request and did not allocate a reply buffer for the dir EA.
5833 Frequency : occasional, higher if lots of files are accessed by one client
5834 Bugzilla : 6159, 6097
5835 Description: Client trips assertion regarding lsm mismatch/magic
5836 Details : While revalidating inodes the VFS looks up inodes with ifind()
5837 and in rare cases can find an inode that is being freed.
5838 The ll_test_inode() code will free the lsm during ifind()
5839 when it finds an existing inode and then the VFS later attaches
5840 this free lsm to a new inode.
5844 Bugzilla : 6422, 7030
5845 Description: MDS deadlock between mkdir and client eviction
5846 Details : Creating a new file via mkdir or mknod (starting a transaction
5847 and getting the ns lock) can deadlock with client eviction
5848 (gets ns lock and trying to finish a synchronous transaction).
5851 Frequency : occasional
5852 Description: While starting a server, the fsfilt_ext3 module could not be
5854 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5855 kernels. Previously, lconf would still use the ext3 name
5856 when trying to load modules. Now, it will correctly use
5857 ext3 on 2.4 and ldiskfs on 2.6.
5859 Severity : enhancement
5860 Description: The default stripe count has been changed to 1
5861 Details : The interpretation of the default stripe count (0, to lfs
5862 or lmc) has been changed to mean striping across a single
5863 OST, rather than all available. For general usage we have
5864 found a stripe count of 1 or 2 works best.
5866 Severity : enhancement
5867 Description: Add support for compiling against Cray portals.
5868 Details : Conditional compiling for some areas that are different
5872 Frequency : occasional
5873 Bugzilla : 6409, 6834
5874 Description: Creating files with an explicit stripe count may lead to
5875 a failed assertion on the MDS
5876 Details : If some OSTs are full or unavailable, creating files may
5877 trigger a failed assertion on the MDS. Now, Lustre will
5878 try to use other servers or return an error to the
5882 Frequency : occasional
5884 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5885 caused lock timeout during readahead (since 1.4.2).
5886 Details : Processes doing readahead might match a lock that hasn't been
5887 granted yet if there are overlapping and conflicting lock
5888 requests. The readahead process waits on ungranted lock
5889 (original lock is CBPENDING), while OST waits for that process
5890 to cancel CBPENDING read lock and eventually evicts client.
5892 Severity : enhancement
5894 Description: Initial enabling of flock support for clients
5895 Details : Implements fcntl advisory locking and file status functions.
5896 This feature is provided as an optional mount flag (default
5897 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5898 locking are implemented yet, and those that are are not guaranteed
5899 to be completely correct in production environments.
5900 mount -t lustre -o [flock|noflock] ...
5903 Frequency : occasional
5905 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5906 assertion in the ext3 JBD (journaling) layer.
5907 Details : The b_committed_data struct is protected by the big kernel lock
5908 in 2.4 kernels, serializing journal_commit_transaction() and
5909 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5910 access to this struct. In 2.6 kernels there is finer grained
5911 locking to improve SMP performance of the JBD layer.
5915 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5920 Description: merge and cleanup kernel patches.
5923 -----------------------------------------------------------------------------
5925 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5927 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5928 versions of Lustre. Please update all servers and clients to
5929 version 1.4.2 or later at the same time. You must also run
5930 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5931 to update the configuration logs.
5933 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5934 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5935 - swab llog records of type '0' so we get proper header size/idx (5861)
5936 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5937 - fix rename of one directory over another leaking an inode (5953)
5938 - avoid SetPageDirty on 2.6 (5981)
5939 - don't re-add just-being-destroyed locks to the waiting list (5653)
5940 - when creating new directories, inherit the parent's custom
5941 striping settings if present parent (3048)
5942 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5943 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5944 - don't hold client locks on temporary worklist from l_lru (5666)
5945 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5946 - automatically evict dead clients (5921)
5947 - Update file size properly in create+truncate+fstat case (6196)
5948 - Do not unhash mountpoint dentries, do not allow removal of
5950 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5951 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5952 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5953 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5954 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5955 - Correctly handle reads of files with no objects (6243)
5956 - lctl recover will also mark a device active if deactivate used (5933)
5958 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5959 - allow --write-conf on an MDS with different nettype than client (5619)
5960 - don't write config llogs to MDS for mounts not from that MDS (5617)
5961 - lconf should create multiple TCP connections from a client (5201)
5962 - init scripts are now turned off by default; run chkconfig --on
5963 lustre and chkconfig --on lustrefs to use them
5964 - upcalls are no longer needed for clients to recover to failover
5966 - add --abort-recovery option to lconf to abort recovery on device
5968 - add support for an arbitrary number of OSTs (3026)
5969 - Quota support protocol changes.
5970 - forward compatibility changes to wire structs (6007)
5971 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5972 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5973 - improved "lustre status" script
5974 - initialize blocksize for non-regular files (6062)
5975 - added --disable-server and --disable-client configure options (5782)
5976 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5977 - Vanilla 2.4.29 support
5978 - increase maximum number of obd devices to 520 (6242)
5979 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5980 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5981 SLES 9 2.6.5 kernels
5982 - limit stripes per file to 160 (the maximum EA size) (6093)
5984 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5987 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5988 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5989 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5990 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5991 - small ext3 extents cleanups and fixes (5733)
5992 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5993 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5994 - avoid needless client->OST connect, fix handle mismatch (5317)
5995 - fix DLM error path that led to out-of-sync client, long delays (5779)
5996 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5997 - fix several locking issues related to i_size (5492,5624,5654,5672)
5998 - don't move pending lock onto export if it is already evicted (5683)
5999 - fix kernel oops when creating .foo in unlinked directory (5548)
6000 - fix deadlock in obdfilter statistics vs. object create (5811)
6001 - use time_{before,after} to avoid timer jiffies wrap (5882)
6002 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6003 - Do not leak request if server was not able to process it (5154)
6004 - If mds_open unable to find parent dir, make that negative lookup(5154)
6005 - don't create new directories with extent-mapping (5909, 5936)
6007 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6008 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6009 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6010 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6011 - support for RHEL4 kernel on i686 (5773)
6012 - provide error messages when incompatible logs are encountered (5898)
6014 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6015 * version 1.4.0.10 (1.4.1 release candidate 1)
6017 - don't keep a lock reference when lock is not granted (4238)
6018 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6019 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6020 - drop import inflight refcount on signal_completed_replay error (5255)
6021 - unlock page after async write error during send (3677)
6022 - handle missing objects in filter_preprw_read properly (5265)
6023 - no transno return for symlink open, don't save no-trasno open (3440)
6024 - don't try to complete elan receive that already failed (4012)
6025 - free RPC server reply state on error (5406)
6026 - clean up thread from ptlrpc_start_thread() on error (5160)
6027 - readahead could read extra page into cache that wasn't ejected (5388)
6028 - prevent races in class_attach/setup/cleanup/detach (5260)
6029 - don't dereference de->d_inode after l_dput of de (5458)
6030 - use "int" for stripe value returned from lock_to_stripe (5544)
6031 - mballoc allocation and error-checking fixes in 2.6 (5504)
6032 - block device patches to fix I/O request sizes in 2.6 (5482)
6033 - look up hostnames for IB nals (5602)
6034 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6035 - don't start multiple acceptors for the same port (5277)
6036 - fix incorrect LASSERT in mds_getattr_name (5635)
6037 - export a proc file for general "ping" checking (5628)
6038 - fix "lfs check" to not block when the MDS is down (5628)
6040 - service request history (4965)
6041 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6042 - create an "evict_client" /proc entry on OSTs, like the MDS has
6043 - fix mount usage message, return errors per mount(8) (5168)
6044 - change grep [] to grep "[]" in tests so they work in more UMLs
6045 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6046 - remove ancient LOV_MAGIC_V0 EA support (5047)
6047 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6048 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6049 - allow readahead allocations to fail when low on memory (5383)
6050 - mmap locking landed again, after considerable improvement (2828)
6051 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6053 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6056 - send OST transaction number in read/write reply to free req (4966)
6057 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6058 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6059 - the watchdog thread now runs as interruptible (5246)
6060 - flock/lockf fixes (but it's still disabled, pending 5135)
6061 - don't use EXT3 constants in llite code (5094)
6062 - memory shortage at startup could cause assertion (5176)
6064 - reorganization of lov code
6065 - single portals codebase
6067 - add extents/mballoc support (5025)
6068 - direct I/O reads in the obdfilter (4048)
6069 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6071 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6074 - send OST transaction number in read/write reply to free req (4966)
6075 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6076 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6077 - the watchdog thread now runs as interruptible (5246)
6078 - handle missing objects in filter_preprw_read properly (5265)
6079 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6080 - drop import inflight refcount on signal_completed_replay error (5255)
6081 - unlock page after async write error during send (3677)
6082 - return original error code on reconstructed replies (3761)
6083 - no transno return for symlink open, don't save no-trasno open (3440)
6085 - add pid to ldlm debugging output (4922)
6086 - bump the watchdog timeouts -- we can't handle 30sec yet
6087 - extra debugging for orphan dentry/inode bug (5259)
6089 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6092 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6093 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6094 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6095 - fix NULL dereference in /proc/sys/portals/routes (4827)
6096 - allow failed mdc_close() operations to be interrupted (4561)
6097 - stop precreate on OST before MDS would time out on it (4778)
6098 - don't send partial-page writes before EOF from client (4410)
6099 - discard client grant for sub-page writes on large-page clients (4520)
6100 - don't free dentries not owned by NFS code, check generation (4806)
6101 - fix lsm leak if mds_create_objects() fails (4801)
6102 - limit debug_daemon file size, always print CERROR messages (4789)
6103 - use transno after validating reply (3892)
6104 - process timed out requests if import state changes (3754)
6105 - update mtime on OST during writes, return in glimpse (4829)
6106 - add mkfsoptions to LDAP (4679)
6107 - use ->max_readahead method instead of zapping global ra (5039)
6108 - don't interrupt __l_wait_event() during strace
6110 - add software watchdogs to catch hung threads quickly (4941)
6111 - make lustrefs init script start after nfs is mounted
6112 - fix CWARN/ERROR duplication (4930)
6113 - return async write errors to application if possible (2248)
6114 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6115 - print NAL number in %x format (4645)
6116 - update barely-supported suse-2.4.21-171 series (4842)
6117 - support for sles 9 %post scripts
6118 - support for building 2.6 kernel-source packages
6119 - support for sles km_* packages
6121 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6124 - ignore -ENOENT errors in osc_destroy (3639)
6125 - notify osc create thread that OSC is being cleaned up (4600)
6126 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6127 - reconstruct ost_handle() like mds_handle() (4657)
6128 - create a new thread to do import eviction to avoid deadlock (3969)
6129 - let lconf resolve symlinked-to devices (4629)
6130 - don't unlink "objects" from directory with default EA (4554)
6131 - hold socknal file ref over connect in case target is down (4394)
6132 - allow more than 32000 subdirectories in a single directory (3244)
6133 - fix blocks count for O_DIRECT writes (3751)
6134 - OST returns ENOSPC from object create when no space left (4539)
6135 - don't send truncate RPC if file size isn't changing (4410)
6136 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6137 - bind to privileged port in socknal and tcpnal (3689)
6139 - rate limit CERROR/CWARN console message to avoid overload (4519)
6140 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6141 - basic mmap support (3918)
6142 - kernel patch series update from b1_4 (4711)
6144 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6147 - avoid crash during MDS cleanup with OST shut down (2775)
6148 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6149 - don't use bad inodes on the MDS (3744)
6150 - dynamic object preallocation to improve recovery speed (4236)
6151 - don't hold spinlock over lock dumping or change debug flags (4401)
6152 - don't zero obd_dev when it is force cleaned (3651)
6153 - print grants to console if they go negative (4431)
6154 - "lctl deactivate" will stop automatic recovery attempts (3406)
6155 - look for existing locks in ldlm_handle_enqueue() (3764)
6156 - don't resolve lock handle twice in recovery avoiding race (4401)
6157 - revalidate should check working dir is a directory (4134)
6159 - don't always mark "slow" obdfilter messages as errors (4418)
6161 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6164 - don't close LustreDB during write_conf until it is done (3860)
6165 - fix typo in lconf for_each_profile (3821)
6166 - allow dumping logs from multiple threads at one time (3820)
6167 - don't allow multiple threads in OSC recovery (3812)
6168 - fix debug_size parameters (3864)
6169 - fix mds_postrecov to initialize import for llog ctxt (3121)
6170 - replace config semaphore with spinlock (3306)
6171 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6172 - don't allow enqueue to complete on a destroyed export (3822)
6173 - down write_lock before checking llog header bitmap (3825)
6174 - recover from lock replay timeout (3764)
6175 - up llog sem before sending rpc (3652)
6176 - reduce ns lock hold times when setting kms (3267)
6177 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6178 - fix NULL deref and obd_dev leak on setup error (3312)
6179 - replace some LBUG about llog ops with error handling (3841)
6180 - don't match INVALID dentries from d_lookup and spin (3784)
6181 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6182 - fix invalid assertion in ptlrpc_set_wait (3880)
6184 - add libwrap support for the TCP acceptor (3996)
6185 - add /proc/sys/portals/routes for non-root route listing (3994)
6186 - allow setting MDS UUID in .xml (2580)
6187 - print the stack of a process that LBUGs (4228)
6189 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6192 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6193 - make sure to unset replay flag from failed open requests (3440)
6194 - if default stripe count is 0, use OST count for inode size (3636)
6195 - update parent mtime/ctime on client for create/unlink (2611)
6196 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6197 - free recovery state on server during a forced cleanup (3571)
6198 - unregister_reply for resent reqs (3063)
6199 - loop back devices mounting and status check on 2.6 (3563)
6200 - fix resource-creation race that can provoke i_size == 0 (3513)
6201 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6202 - more debugging for page-accounting assertion (3746)
6203 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6204 - avoid deadlock after precreation failure (3758)
6205 - fix race and lock order deadlock in orphan handling (3450, 3750)
6206 - add validity checks when grabbing inodes from l_ast_data (3599)
6208 - add /proc/.../recovery_status to obdfilter (3428)
6209 - lightweight CDEBUG infrastructure, debug daemon (3668)
6210 - change default OSC RPC parameters to be better on small clusters
6211 - turn off OST read cache for files smaller than 32MB
6212 - install man pages and include them in rpms (3100)
6213 - add new init script for (un)mounting lustre filesystems (2593)
6214 - run chkconfig in %post for init scripts (3701)
6215 - drop scimac NAL (unmaintained)
6217 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6220 - clean kiobufs before and after use (3485)
6221 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6222 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6223 - change init script to fail more clearly if not run as root (1528)
6224 - allow clients to reconnect during replay (1742)
6225 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6226 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6227 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6228 - fix cond_resched (3554)
6229 - only evict unfinished clients after recovery (3515)
6230 - allow bulk resend, prevent data loss (3570)
6231 - dynamic ptlrpc request buffer allocation (2102)
6232 - don't allow unlinking open directory if it isn't empty (2904)
6233 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6234 - remove extraneous obd dereference causing LASSERT failure (3334)
6235 - don't use get_cycles() when creating temp. files on the mds (3156)
6236 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6237 - handle EEXIST for set-stripe, set proper directory name (3336)
6239 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6240 - fix ksocknal_fmb_callback() error messages (2918)
6242 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6245 - don't copy lvb into (possibly NULL) reply on error (2983)
6246 - don't deref dentry after dput, don't free lvb on error (2922)
6247 - use the kms to determine writeback rpc length (2947)
6248 - increment oti_logcookies when osc is inactive (2948)
6249 - update client's i_blocks count via lvb messages (2543)
6250 - handle intent open/close of special files properly (1557)
6251 - mount MDS with errors=remount-ro, like obdfilter (2009)
6252 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6253 - don't use cancelling-locks' kms values (2947)
6254 - use highest lock extent for kms, not last one (2925)
6255 - don't dereference ERR_PTR() dentry in error handling path (3107)
6256 - fix thread race in portals_debug_dumplog() (3122)
6257 - create lprocfs device entries at setup instead of at attach (1519)
6258 - common AST error handler, don't evict client on completion race (3145)
6259 - zero nameidata in detach_mnt in 2.6 (3118)
6260 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6261 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6262 - initialize RPC timeout timer earlier for 2.6 (3219)
6263 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6264 - print nal/nid for unknown nid (3258)
6265 - additional checks for oscc recovery before doing precreate (3284)
6266 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6267 - don't crash in mdc_close for bad permissions on open (3285)
6268 - zero i_rdev for non-device files (3147)
6269 - clear page->private before handing to FS, better assertion (3119)
6270 - tune the read pipeline (3236)
6271 - fix incorrect decref of invalidated dentry (2350)
6272 - provide read-ahead stats and refine rpc in flight stats (3328)
6273 - don't hold journal transaction open across create RPC (3313)
6274 - update atime on MDS at close time (3265)
6275 - close LDAP connection when recovering to avoid server load (3315)
6276 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6277 - don't leak open file on MDS after open resend (3325)
6278 - serialize filter_precreate and filter_destroy_precreated (3329)
6279 - loop device shouldn't call sync_dev() for nul device (3092)
6280 - clear page cache after eviction (2766)
6281 - resynchronize MDS->OST in background (2824)
6282 - refuse to mount the same filesystem twice on same mountpoint (3394)
6283 - allow llmount to create routes for mounting behind routers (3320)
6284 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6285 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6286 - fix rare problem with rename on htree directories (3417)
6288 - allow default OST striping configuration per directory (1414)
6289 - fix compilation for qswnal for 2.6 kernels (3125)
6290 - increase maximum number of MDS request buffers for large systems
6291 - change liblustreapi to be useful for external progs like lfsck (3098)
6292 - increase local configuration timeout for slow disks (3353)
6293 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6295 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6298 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6299 - don't overwrite extent policy data in reply if lock was blocked (2901)
6300 - drop filter export grants atomically with removal from device (2663)
6301 - del obd_self_export from work_list in class_disconnect_exports (2908)
6302 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6303 - swab reply message in mdc_close, other PPC fixes (2464)
6304 - fix destroying of named logs (2325)
6305 - overwrite old logs when running lconf --write_conf (2264)
6306 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6307 - fix race in target_handle_connect (2898)
6308 - mds_reint_create() should take same inode create lock (2926)
6309 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6310 - don't close files for self_export to avoid uninitialized obd (2936)
6311 - allow MDS with the same name as client node (2939)
6312 - hold dentry reference for closed log files for unlink (2325)
6313 - reserve space for all logs during transactions (2059)
6314 - don't evict page beyond end of stripe extent (2925)
6315 - don't oops on a deleted current working directory (2399)
6316 - handle hard links to targets without a parent properly (2517)
6317 - don't dereference NULL lock when racing during eviction (2867)
6318 - don't grow lock extents when lots of conflicting locks (2919)
6320 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6323 - account for cache space usage on clients to avoid data loss (974)
6324 - lfsck support in lustre kernel code (2349)
6325 - reduce journal credits needed for BRW writes (2370)
6326 - orphan handling to avoid losing space on client/server crashes
6327 - ptlrpcd can be blocked, stopping ALL progress (2477)
6328 - use lock value blocks to assist in proper KMS, faster stat (1021)
6329 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6330 - recovery for initial connections (2355)
6331 - fixes for mds_cleanup_orphans (1934)
6332 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6333 - block all file creations until orphan recovery completes (1901)
6334 - client remove rq_connection from request struct (2423)
6335 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6336 - recovery timer race (2670)
6337 - mdc_close recovey bug (2532)
6338 - ptlrpc cleanup bug (2710)
6339 - mds timeout on local locks (2588)
6340 - namespace lock held during RPCs (2431)
6341 - handle interrupted sync write properly (2503)
6342 - don't try to handle a message that hasn't been replied to (2699)
6343 - client assert failure during cleanup after abort recovery (2701)
6344 - leak mdc device after failed mount (2712)
6345 - ptlrpc_check_set allows timedout requests to complete (2714)
6346 - wait for inflight reqs when ptlrpcd finishes (2710)
6347 - make sure unregistered services are removed from the srv_list
6348 - reset bulk XID's when resending them (caught by 1138 test)
6349 - unregister_bulk after timeout
6350 - fix lconf error (2694)
6351 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6352 - readahead locks pages, leaves pending causing memory pressure (2673)
6353 - increase OST request buffers to 4096 on large machines (2729)
6354 - fix up permission of existing directories in simple_mkdir (2661)
6355 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6356 - don't assign transno to errored transactions (2742)
6357 - don't delete objects on OST if given a bogus objid from MDS (2751)
6358 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6359 - if rq_no_resend, then timeout request after recovery (2432)
6360 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6361 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6362 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6363 - MDS recovery shouldn't delete objects that it has given out (2730)
6364 - if enqueue arrives after completion, don't clobber LVB (2819)
6365 - don't unlock pages twice when trigger_group_io returns error (2814)
6366 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6367 - don't write pages to disk if there was an error (1450)
6368 - don't ping imports that have recovery disabled (2676)
6369 - take buffered bytes into account when balancing socknal conn (2817)
6370 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6371 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6372 - remove little-endian swabbing of llog records (1987)
6373 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6374 - retry reposting request buffers if they fail (1191)
6375 - grow extent at grant time to avoid granting a revoked lock (2809)
6376 - lock revoke doesn't evict page if covered by a second lock (2765)
6377 - disable VM readahead to avoid reading outside lock extents (2805)
6379 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6380 - updated kernel patches for hp-2.4.20 kernel (2681)
6382 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6385 - fix truncated write corruption (2366)
6386 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6388 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6389 - ensure len(uuid) < 37 in lmc (1171)
6390 - fix ia64 OOPS in llog_test (2255)
6391 - zero end of page at obdfilter for partial page writes (2648)
6392 - don't leave stale dentries around after renames (bug 2428)
6393 - fix timeouts when evicting a client with a single lock held (2642)
6394 - set deadline for the initial HELLO message to drain (2634)
6395 - print out dotted-quad IP addresses in the socknal (2302)
6397 - additional debugging for MDS client eviction problem (2443)
6398 - fix mkfsoptions support for osts (2603, 2604)
6400 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6403 - add series for the vanilla 2.6.0 kernel
6404 - add series for the vanilla 2.4.24 kernel
6405 - add series for a cray x86/64 UL kernel drop
6406 - fix xattr patches for the vanilla 2.4.19 series
6408 - generate true UUIDs in lmc (1171)
6409 - have portals stack dumping break in UML (2466)
6410 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6411 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6412 - wake callback waiting threads on client eviction (2460)
6413 - Add --ptldebug and --subsystem to lmc (1719)
6414 - update assertion to allow safe interrupt allocation
6415 - set rq_no_resend for cancel requests (2432)
6416 - recalculate ptlrpcd timeout after resend (2494)
6417 - call vfs_rmdir when removing pending directories (2368)
6418 - fix renaming a file to itself (2429)
6419 - lmc creates a default one-stripe lov (2454)
6420 - expand procfs space to handle large clusters (2326)
6421 - increase UML stack to avoid overflow
6422 - update lconf's list of debug and subsystem masks
6423 - fix lfs find --obd (2510)
6424 - /proc tunable for disabling filter read caching (2591)
6425 - stop rpm packages from altering slapd.conf (2301)
6426 - disable nagle in the socknal under 0conf (2578)
6427 - choose mds inode size based on stripe count (2572)
6428 - fix kernel-source rpm problems (2516)
6430 - add --disable-doc to avoid pdf generation (2421)
6431 - update documentation, tests, type-os, comments
6432 - avoid format warnings on ia64
6433 - remove the TOE NAL
6434 - tiny code cleanups by removing unused fields
6436 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6439 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6440 - remove the most blatant lies from BUILDING file (2371)
6441 - change default debug level to reasonable production setting
6442 - reduce client side cache size to reduce cache flush time
6443 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6444 - make TCP zerocopy and pinger support enabled by default (2476)
6445 - sync writes completed after process exits caused crashes (2319)
6446 - maintain correct mount count on the MDS (2356)
6447 - backout 1557, because 2316 wasn't really fixed
6448 - better file I/O statistics gathering in /proc
6449 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6450 - another kernel patch to fix zero-copy TCP function export
6451 - don't take duplicate lock when processing re-sent getattr (2420)
6452 - lctl uses obd_self_export instead of creating new conn (2353)
6453 - MDS/OST recovery case which requires object creation asserted (2425)
6454 - move lfs from /usr/sbin to /usr/bin in packages
6455 - fix race between mds_client_add and mds_client_free (2417)
6456 - use kmalloc instead of slabs in portals (2430)
6457 - don't create duplicate records when a failover MDS is present (2442)
6458 - remove unnecessary mount age check (2332)
6459 - don't remove directory inodes from locks prematurely (2451)
6460 - don't break if MDS service name is the same as hostname (2103)
6461 - fix races in client write RPC generation when cache full (2482)
6463 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6466 - remove now-unused request->rq_obd (278)
6467 - if an allocation fails, print out how much memory we've used (1933)
6468 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6469 - add the "configurable stack size" patch to most series files (1256)
6470 - ability to write large log records, for 100+ OST configs (2306)
6471 - fix NULL deref when filter_prep fails (2314)
6472 - fix operator precedence error in filter_sync
6473 - dynamic allocation of socknal TX descriptors (2315)
6474 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6475 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6476 - ensure that CWARN messages in Portals always get to the syslog
6477 - __init/__exit are not for prototype decls (ldlm_init/exit)
6478 - x86-64 compile warning fixes
6479 - fix gateway LMC keyword conflict (2318)
6480 - fix MDS lock inversions in getattr/reint paths (1844)
6481 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6482 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6483 - DLM race condition prevented some lock evictions (2328)
6484 - ENOMEM detection and retry on socknal sends (2230)
6485 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6486 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6487 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6488 - better error messages when a client is rejected during recovery (1505)
6489 - avoid cancelling locks which were never granted, after failure (2330)
6490 - fix i_sem/journal inversion in mds_client_add (2333)
6491 - fix truncate/getattr lock cycle deadlock (2334)
6492 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6493 - fix two rare exit paths which could leak an l_lock() ref (2321)
6494 - fencepost error in MDS/OST orphan recovery (2226)
6495 - make log record alignment 8 bytes (1988)
6496 - lstripe now fails when requested offset > ost_count (2237)
6497 - ensure that all kernel series have a complete list.h (1607)
6498 - fix crashes in special-file operations (2316)
6499 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6501 - allow configurable automake binary, for testing new versions
6502 - small update to the lfs documentation
6504 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6506 * fix negative export reference count in fsfilt_sync (2312)
6508 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6509 * release candidate 0.9.1
6511 - orphans are moved into the PENDING directory for possible recovery
6512 - replayed opens now open by fid for orphan/rename safety (1042)
6513 - last close of an orphan inode generates a transno (683)
6514 - chdir() and mount() now pin the directory entry (1020)
6515 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6516 - discard very old requests without processing them (1502)
6517 - remove some common, well-understood CERRORs (1505)
6518 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6519 - clear "grant" flags in OST replies until OST grant code lands (1644)
6520 - fix read performance by not clobbering i_blksize on client (1598)
6521 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6522 - make LRU size a /proc tunable, clears locks when reduced (707)
6523 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6524 - server locks take references on exports now (1558)
6525 - build fixes for 2.4.20-rh trees (1663)
6526 - return an error from lov_create if all OSCs are inactive (1751)
6527 - fix import levels when a reconnect happens without a timeout (1597)
6528 - exit early from mds_open if we get a lookup error (1749)
6529 - partial page read at EOF wouldn't wait for disk before sending (1642)
6530 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6531 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6532 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6533 - write/truncate lock inversion (1639)
6534 - Don't auto-load obdclass, portals modules during cleanup (1495)
6535 - fix timestamps from jumping to "now" (1763)
6536 - extra journal assertions (1648)
6537 - add an extra multiunlink test (1771)
6538 - fix read_record/write_record API (1776)
6539 - fix leak of offset_extent, possible incorrect i_size later (1772)
6540 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6541 - Debugging for the kqswnal_get_idle_tx problems (1820)
6542 - Allow recovery to be attempted multiple times (1536)
6543 - Write out MDS last_rcvd file after it is first created (1600)
6544 - Fix tx_descriptor leak in failed transmit situations (1827)
6545 - ext3 journaling fixes for assertion failure after IO error (1871)
6546 - class_export_put() on freed export after completion AST error (1896)
6547 - Fix revalidate looping in VFS (1322)
6548 - Don't access a freed export during MDS_REINT timeout (1521)
6549 - Add open-unlink recovery support on the MDS (1673,1764)
6550 - Return an error if no MDS data was read from last_rcvd (1946)
6551 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6552 - Don't setup a disk device that doesn't match exported UUID (317)
6553 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6554 - avoid committing NULL handle in force close
6555 - local.sh is now a one-stripe LOV configuration
6556 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6557 - fix ext3 htree duplicate directory entry corruption (1516)
6558 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6559 - update to new LOV EA format (2097)
6560 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6562 2003-06-15 Phil Schwan <phil@clusterfs.com>
6565 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6566 - per-import/export recovery handling (958, 931, 959)
6567 - multiple last-rcvd slots, for serving multiple FSes (949)
6568 - connections are again shared between multiple imp/exports (963, 964)
6569 - "umount -f" would hang if any requests needed to be sent (393, 978)
6570 - avoid pinning large req buffer by copying for queued messages (989)
6571 - add "uuid" to "lctl device" command to help upcalls (991)
6572 - "open" RPCs with transnos would confuse recovery counters (1037)
6573 - do proper endian conversion of all wire messages (288, 340, 891)
6574 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6575 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6576 - fix ptlrpc_connection leak in target_handle_connect (1174)
6577 - fix import refcounting bug in OST and MDS cleanup (1134)
6578 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6579 - fix very unlikely obd_types race condition (501)
6580 - remove osc_open hack for echo_client (1187)
6581 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6582 - a failure in read_inode2 leads to deadlock (1139)
6583 - cancel ack-locks as soon as transaction is committed (1072)
6584 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6585 - make sure to commitrw after any preprw to avoid deadlock (1162)
6586 - failing to execute a file in a lustre FS would lock inode (1203)
6587 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6588 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6589 - obd_uuid2tgt would walk off the end of the list (1255)
6590 - on IA64 the osc would give portals incorrect bulk size (1258)
6591 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6592 - fix lock inversion caused by new llite matching code (1282)
6593 - limit the number of dirty pages on a client to 10MB (1286)
6594 - timed out locks were not being corrected cancelled (1289)
6595 - fix O_DIRECT above 4GB on IA-32 (1292)
6596 * major user-visible changes
6597 - fail out/fail over policy now controlled by the upcall (993)
6599 - add OBD_PING to check server availability and failure (954)
6600 - lustre messages are now sent in sending host order (288, 340, 891)
6601 - add eadatalen to MDS getattr reply (340)
6602 - OST read replies may contain second buffer, with per-page status (593)
6604 2003-03-11 Phil Schwan <phil@clusterfs.com>
6607 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6608 - failed open()s could cause deadlock; fixed (867, 869)
6609 - stop cancelling OST locks when files are closed (481)
6610 - overlapping XID spaces caused network corruption (851, 853)
6611 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6612 - setattr_raw added, to do single-RPC, server-side setattrs
6613 - lmc/lconf syntax change for OST UUIDs
6614 - fix crashy race condition between ptlrpc_free_req and osc_close
6615 - don't use request in mdc_enqueue if we hit a timeout (889)
6616 - don't set the inode i_size for regular files from the MDS (896)
6617 - handle out of order completion AST (842)
6618 - don't LBUG if a lock request times out after receiving AST (913)
6619 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6620 - if a bad lock AST arrives, send an error instead of dropping entirely
6621 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6622 - fix leak in bulk IO when only partially completed (899, 900, 926)
6623 - fix O_DIRECT for ia64 (55)
6624 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6625 - C-z after timeout could hang a process forever; fixed (977)
6627 - client-side I/O cache (678, 924, 929, 941, 970)
6629 - READPAGE and SETATTRs which don't take server-side locks get
6632 2003-02-11 Phil Schwan <phil@clusterfs.com>
6635 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6636 - Fix fsfilt_extN_readpage() to read a full page of directory
6637 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6638 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6639 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6640 - Most intent operations were removed, in favour of a new RPC mode
6641 that does a single RPC to the server and bypasses most of the VFS
6642 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6643 - Aggressively cancel local locks on DLM servers
6644 - mds_reint_unlink sends EA to the client if it's the last nlink.
6645 client uses that EA to unlink OST objects.
6646 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6647 - recursive symlinks were fixed (439)
6648 - fixed NULL deref in DEBUG_REQ
6649 - filter_update_lastobjid no longer calls sync, which annoyed extN
6650 - fixed multi-client small-writes to a single file problem (445)
6651 - fixed mtime updates during file writes (607)
6652 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6653 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6654 - fixed Linux OST size reporting problem (444, 656)
6655 - OST now updates object mtime with writes or setattr (607, 619)
6656 - client verifies file size before zeroing page past EOF (445)
6657 - OST now writes last allocated objid to disk with allocation (108)
6658 - LOV on echo now works (409)
6660 - mds_reint_unlink sends a new buffer, with the EA included. this
6661 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6662 set if a regular file was being unlinked, and it was the last link
6663 - use PtlGet from the target for bulk writes (315)
6664 - OST now updates object mtime with writes or setattr (607, 619)
6665 - LDLM now has a grant-time callback to revalidate locked items, if
6667 - Many MDS operations were reorganized to combat race conditions
6669 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6670 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6672 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6675 - Fully reactivate OST imports after reconnection (512, others)
6676 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6677 - More graceful error handling for truncating on dead OST (515)
6678 - Don't error out unless we're actually accessing dead stripes (474)
6679 - Fix garbage sizes when stripes are missing (410)
6680 - LRU counters were broken, causing constant lock purge (433, 432)
6681 - garbage on read from stripes with failed OSTs (441)
6682 - mark OSCs as active before reconnecting during recovery (438)
6683 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6684 - lfind did not preserve OST order in output (443)
6685 - symlinks cause hung clients, incorrect data (439)
6686 - stop dereferencing request after dropping refcount (457)
6687 - don't LASSERT(spin_is_locked) on non-SMP (455)
6688 - fixes for many rename() bugs
6689 - fstat didn't correctly synchronize attributes (399)
6690 - server must handle lock cancellation during blocking AST prep (487)
6691 - bulk descriptors were free()d too soon (511)
6692 - fix paths in lconf, which would load incorrect modules (451, 507)
6693 - fix confusing lconf 'host not found' error message (386)
6694 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6695 - fix race condition in mdc_blocking_ast() for inode access (526)
6696 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6697 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6698 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6699 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6700 - fix lstripe in several ways (kernel side) (527)
6701 - fix request leak in ldlm_cli_enqueue (262)
6702 - incorrect OSC was marked inactive after OST failure
6703 - call mds_fs_cleanup before unmounting filesystem (524)
6704 - fix races between taking ns_lock and ldlm_lock_change_resource
6705 - fix races updating LOV export open file list
6706 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6707 - fix recovery NULL deref in ldlm_cli_cancel_unused
6708 - fix some DLM races by using new hash table for lock handles (419)
6709 - permit the client to specify desired inodes, at replay
6710 - duplicate requests when we queue them for replay reintegration
6711 - fix last_rcvd offset calculation
6712 - sync after each recovered transaction, so we always make progress
6713 - never, not always, ERESTART requests without transnos
6714 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6716 - skip replay if the MDS says that the client is already connected
6717 - don't check for a recovery-enabled export to match lctl's UUID
6718 - don't INC_USE_COUNT for phantom exports
6719 - don't crash when cleaning up phantom exports (567)
6720 - don't double-finish or set replay data for errored mdc_open requests
6721 - abort requests when they time out, so we don't get old replies
6722 - send/receive replies for AST messages again
6723 - if the client says that it doesn't have the lock, cancel it on the
6725 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6726 mark it as destroyed, it will all work out when decref is called
6727 - fix module use counts (22, 581)
6729 - ASTs now expect a reply (server cancels lock on error reply)
6731 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6734 - fix many simultaneous client startup (392)
6735 - fix dentry->d_it clobbering
6736 - credentials weren't being shipped for readdir/getattr operations
6737 - remove invalid assertions triggered during some concurrent MD
6739 - proper Lustre versions added (336, 389)
6740 - fix memory leak for create error case (398)
6741 - fix LOV locking bug that would get cli/srv out of sync
6742 - fix echo client over LOV (409)
6743 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6744 - fix double-O_EXCL intent crash (424)
6745 - avoid sending multiple lock CANCELs (352)
6747 - MDS can do multi-client recovery (modulo bugs in new code)
6749 - many updates, edits, cleanups
6751 2002-11-18 Phil Schwan <phil@clusterfs.com>
6754 - fix null d_it dereference (346)
6755 - fix full OST/dbench hang (333)
6756 - fix permission problem with file removal (286)
6757 - fix removal of OSCs from LOV when they fail
6758 - fix NULL deref during bulk timeout (214)
6759 - fix problems related to multiple filesystems on one MDS (241)
6760 - fixed serious subtle metadata locking bugs
6761 - free locks on clients when inodes are removed due to memory
6763 - fix inode pointer in lock data (285)
6764 - partial support for multiple MDS on a single host (241)
6765 - data locks weren't cancelled at clear_inode time (290, 311)
6766 - intent locks could lead to unbounded lock growth (205)
6767 - added a maximum lock count, an LRU list, and a flusher
6768 - fix multiple rename (365)
6769 - properly abstracted the echo client
6770 - OSC locked 1 byte too many; fixed
6771 - rewrote brw callback code:
6772 - fixed recovery bugs related to LOVs (306)
6773 - fixed too-many-pages-in-one-write crash (191)
6774 - fixed (again) crash in sync_io_timeout (214)
6775 - probably fixed callback-related race (385)
6777 - Add capability to MDS protocol
6778 - LDLM cancellations and callbacks on different portals
6780 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6783 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6784 - defer open object destruction to close time (601981, 138)
6785 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6786 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6787 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6788 - don't call obd_setattr in ll_file_release for destroyed objects
6789 * protocol change to lustre_msg: move |version| and add |flags|
6790 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6791 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6792 * added replay of create, unlink, link and rename operations during
6793 MDS failover; recovery should be much more robust now
6794 * remove failed OSCs from LOVs (only lov_create uses this so far)
6795 * the lustre-HOWTO was brought (more) up to date (582544)
6797 2002-10-23 Phil Schwan <phil@clusterfs.com>
6800 - in-use dentries weren't being reused properly (617851)
6801 - prevent multiple LDLM setup (599178)
6802 - fix LOV size calculations for truncate (617853)
6803 - fix client handling of MDS intent errors (POSIX)
6804 - fix permission bug in lovstripe.c test (624321)
6805 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6806 - truncate past end of file could corrupt data
6807 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6808 - a race in recovery could return ETIMEDOUT to apps (623947)
6809 - building outside the source directory was fixed
6810 * the lustre-HOWTO was brought (more) up to date (582544)
6811 * major progress was made on recovery functionality
6813 2002-10-10 Phil Schwan <phil@clusterfs.com>
6816 - recovery deadlock fix
6817 - rm -rf causes LBUG fix (617817)
6818 - file open by multiple tasks fix (618962)
6819 - directory permissions bugs (602707 and 620007)
6820 - journal_stop fixed with locking (611313)
6821 - O_APPEND failures resolved (618273, perhaps 614459)
6822 - lconf PATH fix (619770)
6823 - IA64 build fix (621450)
6824 - RPC buffer sizes scale with amount of memory
6826 2002-10-01 Phil Schwan <phil@clusterfs.com>
6829 - locks would be cancelled without throwing away data pages,
6830 resulting in inconsistent data (605627)
6831 - inode attributes were not always being refreshed (605627, 612449)
6832 - lconf now continues to cleanup after lctl reports an error
6833 - MDS now enforces user permissions (602707)
6834 - lprocfs cleanup fixed, but not yet enabled (614157)
6835 - fixed infinite server hang, should a client not respond to an AST
6836 - avoid going into recovery if user calls readlink() with a buffer
6837 that's too small (613941)
6838 - AST RPCs no longer require replies (614867) -- this may be changed
6839 - don't crash server if client sends an IOV that's too big (611336)
6840 - fixed lock conversion deadlock (611892)
6841 - fixed the following of symlinks (614622)
6842 * recovery: the server can remove locks from a client that dies, other
6843 clients can make progress
6844 * more extN patch fixes
6845 * compile-time configurable ptlrpc buffer allocations
6847 - collaborative read cache document
6848 - Lustre Lite Performance CDR document-in-progress
6850 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6853 - fix typo in patch-2.4.18
6855 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6858 - clear ptlrpc request each time in handle_incoming_request()
6859 - unlink of files now destroys the object on the OST
6861 2002-09-19 Peter Braam <braam@clusterfs.com>
6863 * add hard link support
6864 * change obdfile creation method
6865 * kernel patch changed
6867 2002-09-19 Peter Braam <braam@clusterfs.com>
6870 - stack overflow bug in extN fixed
6872 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6874 * documentation updates
6875 - add man pages for config tools
6876 - update tests/README to describe testing with new config tools
6877 - finish metadata API descriptions
6878 * bug fixes and cleanups
6879 - statfs workaround for 16TB limit
6880 - LOV stripe allocation improved, can stripe on subset of OSTs
6881 - LOV file size/IO offset was wrong for files > 4GB in size
6882 - object EA data was being dropped, caused files to be unreadable
6883 - memory overflow with non-LOV OST caused memory corruption
6884 - fixed regression tests to work with new config tools, obdfilter
6885 - fixed bug when directory size became larger than 1 block
6886 - fixed bug (for single client case) when PWD was deleted
6887 - invalidate local directory pages when doing intent-based ops
6888 - avoid LDLM oops when lock callback contained bad data
6890 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6892 * documentation updates
6893 * bug fixes and cleanups
6894 - configuration tools
6897 - 64-bit compile warnings
6898 - 64-bit internal statfs data
6900 * test_brw on persistent OST devices
6902 * lprocfs (disabled)
6904 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6906 * documentation updates
6907 * bug fixes and cleanups
6908 * configuration tools
6910 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6913 * many small fixes to 0_5_4
6914 * io/network handling
6915 * thinkos in MDS operations
6917 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6920 * crucial basic fixes to 0.5.3
6921 * IOR, Iozone work over Elan
6924 2002-08-07 Phil Schwan <phil@clusterfs.com>
6925 * version 0_5_3, our first alpha
6926 * we use the new Portals iovs
6927 * documentation updates
6928 * bug fixes and cleanups
6929 * small changes in the DLM wire protocol
6931 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6932 * version 0_5_1 with some initial stability,
6933 * locking on MD and file I/O.
6934 * documentation updates
6935 * several bug fixes since 0.5.0
6936 * small changes in wire protocol
6938 2002-07-18 Phil Schwan <phil@clusterfs.com>
6940 * delivered as Lustre Light Alpha
6941 * fixed a crash after handling invalid MDS requests
6942 * fixed directory pages for architectures with non-4k pages sizes
6944 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6945 * release version v0_4_4
6946 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6948 * Fixes a number of interruption problems with OST operations.
6949 * Update documentation for portals header changes
6950 * Move all wire protocol structs/defines to lustre_idl.h
6951 * Fixes symlink length bug.
6952 * Add tcpdump to repository.
6954 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6955 * release version v0_4_3
6956 * Fixes statfs for inodes on extN.
6957 * Fixes bug in runtests which would delete /etc/hosts.
6958 * Use 64-bit object IDs wherever possible (not into VFS though)
6959 Remove ost_get_info, which is unused by lustre, and out of date.
6961 2002-07-03 Peter Braam <braam@clusterfs.com>
6962 * release version v0_4_2 Fixes a lookup error (type not passed)
6963 * move forward to head of Portals
6964 * move forward to latest Lustre kernel
6966 2002-06-25 Peter Braam <braam@clusterfs.com>
6967 * release version v0_4_1. Hopefully stable on single node use.