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 4.18.0-372.16.1.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 3.10.0-1160.71.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.18.0-348.23.1.el8 (RHEL8.5)
19 4.4.120-92.70 (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 4.18.0-372.16.1.el8 (RHEL8.6)
31 5.4.0-37 (Ubuntu 20.04)
32 5.3.18-59.27 (SLES15 SP3)
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 3.10.0-1160.71.1.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 4.18.0-348.23.1.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 4.12.14-122.116 (SLES12 SP5)
51 4.12.14-197.75 (SLES15 SP1)
52 5.3.18-24.96 (SLES15 SP2)
53 4.4.0-131 (Ubuntu 16.04)
54 4.15.0-48 (Ubuntu 18.04)
55 5.8.0-53 (Ubuntu 20.04.2 HWE)
56 5.11.0-31 (Ubuntu 20.04.3 HWE)
57 5.11.0 (vanilla kernel.org)
58 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
59 * Recommended ZFS version: 2.1.2
60 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
61 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
62 stack overflow. For more information, please refer to bugzilla 17630.
63 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
64 the re-exporting nodes to be mounted with "32bitapi" mount option
66 --------------------------------------------------------------------------------
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 4.18.0-348.2.1.el8 (RHEL8.5)
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 3.10.0-1160.49.1.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 4.4.120-92.70 (SLES12 SP2)
87 4.4.180-94.100 (SLES12 SP3)
88 4.4.0-131 (Ubuntu 16.04)
89 4.15.0-32 (Ubuntu 18.04)
90 5.4.0-48 (Ubuntu 20.04)
91 vanilla linux 5.4.0 (ZFS + ldiskfs)
92 vanilla linux 5.4.21 (ZFS + ldiskfs)
93 vanilla linux 5.4.136 (ZFS + ldiskfs)
94 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
95 * Client primary kernels built and tested during release cycle:
96 4.18.0-348.2.1.el8 (RHEL8.5)
97 5.4.0-37 (Ubuntu 20.04)
98 5.3.18-59.27 (SLES15 SP3)
99 * Other clients known to build on these kernels at some point (others may also work):
100 3.10.0-862.14.4.el7 (RHEL7.5)
101 3.10.0-957.27.2.el7 (RHEL7.6)
102 4.14.0-49.13.1.el7a (RHEL7.5)
103 4.14.0-115.2.2.el7a (RHEL7.6)
104 3.10.0-1062.18.1.el7 (RHEL7.7)
105 3.10.0-1127.19.1.el7 (RHEL7.8)
106 3.10.0-1160.49.1.el7 (RHEL7.9)
107 4.18.0-80.11.2.el8 (RHEL8.0)
108 4.18.0-147.8.1.el8 (RHEL8.1)
109 4.18.0-193.28.1.el8 (RHEL8.2)
110 4.18.0-240.22.1.el8 (RHEL8.3)
111 4.18.0-305.25.1.el8 (RHEL8.4)
112 4.4.120-92.70 (SLES12 SP2)
113 4.4.180-94.100 (SLES12 SP3)
114 4.12.14-95.48 (SLES12 SP4)
115 4.12.14-122.91 (SLES12 SP5)
116 4.12.14-197.75 (SLES15 SP1)
117 5.3.18-24.96 (SLES15 SP2)
118 4.4.0-131 (Ubuntu 16.04)
119 4.15.0-48 (Ubuntu 18.04)
120 5.8.0-53 (Ubuntu 20.04.2 HWE)
121 5.11.0-31 (Ubuntu 20.04.3 HWE)
122 5.11.0 (vanilla kernel.org)
123 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
124 * Recommended ZFS version: 2.0.7
125 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
126 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
127 stack overflow. For more information, please refer to bugzilla 17630.
128 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
129 the re-exporting nodes to be mounted with "32bitapi" mount option
131 --------------------------------------------------------------------------------
135 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
136 for currently supported client and server kernel versions.
137 * Server primary kernels built and tested during release cycle:
138 3.10.0-1062.18.1.el7 (RHEL7.7)
139 3.10.0-1127.19.1.el7 (RHEL7.8)
140 3.10.0-1160.6.1.el7 (RHEL7.9)
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 4.18.0-80.11.2.el8 (RHEL8.0)
145 4.18.0-147.8.1.el8 (RHEL8.1)
146 4.18.0-193.28.1.el8 (RHEL8.2)
147 4.18.0-240.1.1.el8 (RHEL8.3)
148 4.4.120-92.70 (SLES12 SP2)
149 4.4.180-94.100 (SLES12 SP3)
150 4.4.0-131 (Ubuntu 16.04)
151 4.15.0-32 (Ubuntu 18.04)
152 5.4.0-48 (Ubuntu 20.04)
153 vanilla linux 5.4.0 (ZFS + ldiskfs)
154 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
155 * Client primary kernels built and tested during release cycle:
156 3.10.0-1062.18.1.el7 (RHEL7.7)
157 3.10.0-1127.19.1.el7 (RHEL7.8)
158 3.10.0-1160.6.1.el7 (RHEL7.9)
159 4.12.14-95.48 (SLES12 SP4)
160 4.12.14-122.20 (SLES12 SP5)
161 4.12.14-197.75 (SLES15 SP1)
162 5.3.18-24.24 (SLES15 SP2)
163 4.15.0-48 (Ubuntu 18.04)
164 * Other clients known to build on these kernels at some point (others may also work):
165 3.10.0-862.14.4.el7 (RHEL7.5)
166 3.10.0-957.27.2.el7 (RHEL7.6)
167 4.14.0-49.13.1.el7a (RHEL7.5)
168 4.14.0-115.2.2.el7a (RHEL7.6)
169 4.18.0-80.11.2.el8 (RHEL8.0)
170 4.18.0-147.8.1.el8 (RHEL8.1)
171 4.18.0-193.28.1.el8 (RHEL8.2)
172 4.18.0-240.1.1.el8 (RHEL8.3)
173 4.4.120-92.70 (SLES12 SP2)
174 4.4.180-94.100 (SLES12 SP3)
175 4.4.0-131 (Ubuntu 16.04)
176 5.4.0-37 (Ubuntu 20.04)
177 5.4.0 (vanilla kernel.org)
178 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
179 * Recommended ZFS version: 2.0.0
180 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
181 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
182 stack overflow. For more information, please refer to bugzilla 17630.
183 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
184 the re-exporting nodes to be mounted with "32bitapi" mount option
186 --------------------------------------------------------------------------------
190 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
191 for currently supported client and server kernel versions.
192 * Server primary kernels built and tested during release cycle:
193 3.10.0-1062.1.1.el7 (RHEL7.7)
194 * Other server kernels known to build and work at some point (others may also work):
195 3.10.0-862.14.4.el7 (RHEL7.5)
196 3.10.0-957.27.2.el7 (RHEL7.6)
197 4.18.0-80.7.1.el8 (RHEL8.0)
198 4.4.120-92.70 (SLES12 SP2)
199 4.4.180-94.100 (SLES12 SP3)
200 4.4.0-131 (Ubuntu 16.04)
201 4.15.0-32 (Ubuntu 18.04)
202 vanilla linux 4.6.7 (ZFS only)
203 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
204 * Client primary unpatched kernels built and tested during release cycle:
205 3.10.0-1062.1.1.el7 (RHEL7.7)
206 4.12.14-95.29 (SLES12 SP4)
207 4.15.0-48 (Ubuntu 18.04)
208 * Other clients known to build on unpatched kernels at some point (others may also work):
209 3.10.0-862.14.4.el7 (RHEL7.5)
210 3.10.0-957.27.2.el7 (RHEL7.6)
211 4.14.0-49.13.1.el7a (RHEL7.5)
212 4.14.0-115.2.2.el7a (RHEL7.6)
213 4.18.0-80.7.1.el8 (RHEL8.0)
214 4.4.120-92.70 (SLES12 SP2)
215 4.4.180-94.100 (SLES12 SP3)
216 4.4.0-131 (Ubuntu 16.04)
217 vanilla linux 4.15.0 (ZFS only)
218 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
219 * Recommended ZFS / SPL version: 0.7.13
220 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
221 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
222 stack overflow. For more information, please refer to bugzilla 17630.
223 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
224 the re-exporting nodes to be mounted with "32bitapi" mount option
225 * Note for Ubuntu 14.04.5:
226 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
227 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
229 --------------------------------------------------------------------------------
233 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
234 for currently supported client and server kernel versions.
235 * Server known to build on the following kernels (others may also work):
236 2.6.32-696.18.7.el6 (RHEL6.9)
237 3.10.0-862.14.4.el7 (RHEL7.5)
238 3.10.0-957.el7 (RHEL7.6)
239 4.4.120-92.70 (SLES12 SP2)
240 4.4.162-94.69 (SLES12 SP3)
241 4.4.0-131 (Ubuntu 16.04)
242 4.15.0-32 (Ubuntu 18.04)
243 vanilla linux 4.6.7 (ZFS only)
244 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
245 * Client known to build on unpatched kernels (others may also work):
246 2.6.32-696.18.7.el6 (RHEL6.9)
247 2.6.32-754.6.3.el6 (RHEL6.10)
248 3.10.0-862.14.4.el7 (RHEL7.5)
249 3.10.0-957.el7 (RHEL7.6)
250 4.14.0-49.13.1.el7a (RHEL7.5)
251 4.14.0-115.2.2.el7a (RHEL7.6)
252 4.4.120-92.70 (SLES12 SP2)
253 4.4.162-94.69 (SLES12 SP3)
254 4.4.0-131 (Ubuntu 16.04)
255 4.15.0-32 (Ubuntu 18.04)
256 vanilla linux 4.15.0 (ZFS only)
257 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
258 * Recommended ZFS / SPL version: 0.7.9
259 * Tested with ZFS / SPL version: 0.7.9
260 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
261 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
262 stack overflow. For more information, please refer to bugzilla 17630.
263 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
264 the re-exporting nodes to be mounted with "32bitapi" mount option
265 * Note for Ubuntu 14.04.5:
266 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
267 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
269 --------------------------------------------------------------------------------
271 2018-04-03 Intel Corporation
273 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
274 for currently supported client and server kernel versions.
275 * Server known to build on patched kernels:
276 2.6.32-431.29.2.el6 (RHEL6.5)
277 2.6.32-504.30.3.el6 (RHEL6.6)
278 2.6.32-573.26.1.el6 (RHEL6.7)
279 2.6.32-642.15.1.el6 (RHEL6.8)
280 2.6.32-696.18.7.el6 (RHEL6.9)
281 3.10.0-693.21.1.el7 (RHEL7.4)
282 3.0.101-0.47.71 (SLES11 SP3)
283 3.0.101-107 (SLES11 SP4)
284 3.12.74-60.64.40 (SLES12 SP1)
285 4.4.114-92.64 (SLES12 SP2)
286 4.4.114-94.11 (SLES12 SP3)
287 3.13.0-101 (Ubuntu 14.04, ZFS only)
288 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
289 4.4.0-116 (Ubuntu 16.04)
290 vanilla linux 4.6.7 (ZFS only)
291 * Client known to build on unpatched kernels:
292 2.6.32-431.29.2.el6 (RHEL6.5)
293 2.6.32-504.30.3.el6 (RHEL6.6)
294 2.6.32-573.26.1.el6 (RHEL6.7)
295 2.6.32-642.15.1.el6 (RHEL6.8)
296 2.6.32-696.18.7.el6 (RHEL6.9)
297 3.10.0-693.21.1.el7 (RHEL7.4)
298 3.0.101-0.47.71 (SLES11 SP3)
299 3.0.101-107 (SLES11 SP4)
300 3.12.74-60.64.40 (SLES12 SP1)
301 4.4.114-92.64 (SLES12 SP2)
302 4.4.114-94.11 (SLES12 SP3)
303 3.13.0-101 (Ubuntu 14.04)
304 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
305 4.4.0-116 (Ubuntu 16.04)
307 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
308 * Recommended ZFS / SPL version: 0.7.6
309 * Tested with ZFS / SPL version: 0.7.6
310 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
311 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
312 stack overflow. For more information, please refer to bugzilla 17630.
313 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
314 the re-exporting nodes to be mounted with "32bitapi" mount option
315 * Note for Ubuntu 14.04.5:
316 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
317 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
319 --------------------------------------------------------------------------------
321 2017-07-05 Intel Corporation
323 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
324 for currently supported client and server kernel versions.
325 * Server known to build on patched kernels:
326 2.6.32-431.29.2.el6 (RHEL6.5)
327 2.6.32-504.30.3.el6 (RHEL6.6)
328 2.6.32-573.26.1.el6 (RHEL6.7)
329 2.6.32-642.15.1.el6 (RHEL6.8)
330 2.6.32-696.3.1.el6 (RHEL6.9)
331 3.10.0-514.21.1.el7 (RHEL7.3)
332 3.0.101-0.47.71 (SLES11 SP3)
333 3.0.101-100 (SLES11 SP4)
334 3.12.74-60.64.40 (SLES12 SP1)
335 4.4.59-92.17 (SLES12 SP2)
336 vanilla linux 4.6.7 (ZFS only)
337 * Client known to build on unpatched kernels:
338 2.6.32-431.29.2.el6 (RHEL6.5)
339 2.6.32-504.30.3.el6 (RHEL6.6)
340 2.6.32-573.26.1.el6 (RHEL6.7)
341 2.6.32-642.15.1.el6 (RHEL6.8)
342 2.6.32-696.3.1.el6 (RHEL6.9)
343 3.10.0-514.21.1.el7 (RHEL7.3)
344 3.0.101-0.47.71 (SLES11 SP3)
345 3.0.101-100 (SLES11 SP4)
346 3.12.74-60.64.40 (SLES12 SP1)
347 4.4.59-92.17 (SLES12 SP2)
349 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
350 * Recommended ZFS / SPL version: 0.6.5.9
351 * Tested with ZFS / SPL version: 0.6.5.9
352 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
353 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
354 stack overflow. For more information, please refer to bugzilla 17630.
355 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
356 the re-exporting nodes to be mounted with "32bitapi" mount option
358 --------------------------------------------------------------------------------
360 12-07-2016 Intel Corporation
362 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
363 for currently supported client and server kernel versions.
364 * Server known to build on patched kernels:
365 2.6.32-431.29.2.el6 (RHEL6.5)
366 2.6.32-504.30.3.el6 (RHEL6.6)
367 2.6.32-573.26.1.el6 (RHEL6.7)
368 2.6.32-642.11.1.el6 (RHEL6.8)
369 3.10.0-514.el7 (RHEL7.3)
370 3.0.101-0.47.71 (SLES11 SP3)
371 3.0.101-80 (SLES11 SP4)
372 3.12.62-60.64.8 (SLES12 SP1)
373 vanilla linux 4.5.7 (ZFS only)
374 * Client known to build on unpatched kernels:
375 2.6.32-431.29.2.el6 (RHEL6.5)
376 2.6.32-504.30.3.el6 (RHEL6.6)
377 2.6.32-573.26.1.el6 (RHEL6.7)
378 2.6.32-642.11.1.el6 (RHEL6.8)
379 3.10.0-514.el7 (RHEL7.3)
380 3.0.101-0.47.71 (SLES11 SP3)
381 3.0.101-80 (SLES11 SP4)
382 3.12.62-60.64.8 (SLES12 SP1)
384 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
385 * Recommended ZFS / SPL version: 0.6.5.8
386 * Tested with ZFS / SPL version: 0.6.5.8
387 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
388 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
389 stack overflow. For more information, please refer to bugzilla 17630.
390 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
391 the re-exporting nodes to be mounted with "32bitapi" mount option
393 --------------------------------------------------------------------------------
395 02-29-2016 Intel Corporation
397 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
398 for currently supported client and server kernel versions.
399 * Server known to build on patched kernels:
400 2.6.32-431.29.2.el6 (RHEL6.5)
401 2.6.32-504.30.3.el6 (RHEL6.6)
402 2.6.32-573.12.1.el6 (RHEL6.7)
403 3.10.0-327.3.1.el7 (RHEL7.2)
404 3.0.101-0.47.71 (SLES11 SP3)
405 3.0.101-68 (SLES11 SP4)
406 vanilla linux 4.2.1 (ZFS only)
407 * Client known to build on unpatched kernels:
408 2.6.32-431.29.2.el6 (RHEL6.5)
409 2.6.32-504.30.3.el6 (RHEL6.6)
410 2.6.32-573.12.1.el6 (RHEL6.7)
411 3.10.0-327.3.1.el7 (RHEL7.2)
412 3.0.101-0.47.71 (SLES11 SP3)
413 3.0.101-68 (SLES11 SP4)
416 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
417 * Recommended ZFS / SPL version: 0.6.4.2
418 * Tested with ZFS / SPL version: 0.6.4.2
419 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
420 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
421 stack overflow. For more information, please refer to bugzilla 17630.
422 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
423 the re-exporting nodes to be mounted with "32bitapi" mount option
425 --------------------------------------------------------------------------------
427 03-10-2015 Intel Corporation
429 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
430 for currently supported client and server kernel versions.
431 * Server known to build on patched kernels:
432 2.6.32-431.29.2.el6 (RHEL6.5)
433 2.6.32-504.8.1.el6 (RHEL6.6)
434 3.0.101-0.46 (SLES11 SP3)
435 * Client known to build on unpatched kernels:
436 2.6.32-431.29.2.el6 (RHEL6.5)
437 2.6.32-504.8.1.el6 (RHEL6.6)
438 3.10.0-123.20.1.el7 (RHEL7)
439 3.0.101-0.46 (SLES11 SP3)
440 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
441 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
442 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
443 stack overflow. For more information, please refer to bugzilla 17630.
444 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
445 the re-exporting nodes to be mounted with "32bitapi" mount option
447 Severity : enhancement
449 Description: control OST-index in IDIF via ROCOMPAT flag.
450 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
451 last_rcvd file. For new formatted OST device, it will be auto set;
452 for the case of upgrading from old OST device, you can enable it
453 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
454 enabled, for new created OST-object, its IDIF-in-LMA will contain
455 the OST-index; for the existing OST-object, the OSD will convert
456 old format IDIF as new format IDIF with OST-index stored in the
457 LMA EA when accessing such OST-object or via OI scrub. Once such
458 flag is enabled, it cannot be reverted back, so the system cannot
459 be downgraded to the orignal incompatible version.
461 --------------------------------------------------------------------------------
463 07-30-2014 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 2.6.32-431.20.3.el6 (RHEL6.5)
469 3.0.101-0.31 (SLES11 SP3)
470 * Client known to build on unpatched kernels:
471 2.6.32-431.20.3.el6 (RHEL6.5)
472 3.0.101-0.31 (SLES11 SP3)
475 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
476 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
477 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
478 stack overflow. For more information, please refer to bugzilla 17630.
479 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
480 the re-exporting nodes to be mounted with "32bitapi" mount option
482 --------------------------------------------------------------------------------
484 10-31-2013 Intel Corporation
486 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
487 * Server builds for kernels:
488 2.6.32-358.18.1.el6 (RHEL6.4)
489 * Client builds for unpatched kernels:
490 2.6.32-358.18.1.el6 (RHEL6.4)
491 3.0.92-0.8 (SLES11 SP3)
493 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
494 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
495 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
496 stack overflow. For more information, please refer to bugzilla 17630.
497 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
498 the re-exporting nodes to be mounted with "32bitapi" mount option
500 --------------------------------------------------------------------------------
502 05-30-2013 Intel Corporation
504 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
505 * Server builds for kernels:
506 2.6.32-358.6.2.el6 (RHEL6.4)
507 * Client builds for unpatched kernels:
508 2.6.32-358.6.2.el6 (RHEL6.4)
509 3.0.74-0.6.6 (SLES11 SP2)
511 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
512 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
513 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
514 stack overflow. For more information, please refer to bugzilla 17630.
515 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
516 the re-exporting nodes to be mounted with "32bitapi" mount option
518 --------------------------------------------------------------------------------
520 9-30-2012 Whamcloud, Inc.
522 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
523 * Server support for kernels:
524 2.6.32-279.5.1.el6 (RHEL6)
525 * Client support for unpatched kernels:
526 2.6.32-279.5.1.el6 (RHEL6)
528 2.6.18-238.19.1.el5 (RHEL5)
529 2.6.16.60-0.69.1 (SLES10)
530 * Recommended e2fsprogs version: 1.41.90.wc4
531 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
532 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
533 stack overflow. For more information, please refer to bugzilla 17630.
534 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
535 the re-exporting nodes to be mounted with "32bitapi" mount option
537 Severity : enhancement
539 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
541 Severity : enhancement
543 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
545 Severity : enhancement
547 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
549 Severity : enhancement
551 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
553 Severity : enhancement
555 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
557 --------------------------------------------------------------------------------
559 04-30-2012 Whamcloud, Inc.
561 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
562 * Server support for kernels:
563 2.6.32-220.4.2.el6 (RHEL6)
564 * Client support for unpatched kernels:
565 2.6.32-220.4.2.el6 (RHEL6)
567 2.6.18-238.19.1.el5 (RHEL5)
568 2.6.16.60-0.69.1 (SLES10)
569 * Recommended e2fsprogs version: 1.41.90.wc4
570 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
571 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
572 stack overflow. For more information, please refer to bugzilla 17630.
573 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
574 the re-exporting nodes to be mounted with "32bitapi" mount option
576 Severity : enhancement
578 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
580 Severity : enhancement
582 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
584 Severity : enhancement
586 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
588 Severity : enhancement
590 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
592 --------------------------------------------------------------------------------
594 09-30-2011 Whamcloud, Inc.
596 * Server support for kernels:
597 2.6.18-238.19.1.el5 (RHEL5)
598 2.6.32-131.6.1.el6 (RHEL6)
599 * Client support for unpatched kernels:
600 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
601 2.6.18-238.19.1.el5 (RHEL5)
602 2.6.32-131.6.1.el6 (RHEL6)
603 2.6.29.4-167.fc11 (FC11)
604 2.6.16.60-0.69.1 (SLES10)
606 * Recommended e2fsprogs version: 1.41.90.wc3
607 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
608 Lustre filesystem with 4K stack may cause a stack overflow. For more
609 information, please refer to bugzilla 17630.
610 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
611 mounted with "32bitapi" mount option
613 Severity : enhancement
614 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
616 Severity : enhancement
617 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
619 Severity : enhancement
620 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
623 Frequency : only with SLES10
625 Description: Use OFED "KMP" provided by Novell
626 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
627 Lustre is now built against this package. That means you need to
628 install the ofed-kmp package from Novell for the patchless client.
629 Note that the ofed-kmp that Novell ships may not exactly match the
630 kernel version but should still be compatible.
632 Severity : enhancement
634 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
635 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
637 --------------------------------------------------------------------------------
639 2010-07-15 Oracle, Inc.
641 * Server support for kernels:
642 2.6.18-164.11.1.el5 (RHEL 5)
643 2.6.18-164.11.1.0.1.el5 (OEL 5)
644 * Client support for unpatched kernels:
645 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
646 2.6.18-164.11.1.el5 (RHEL 5),
647 2.6.18-164.11.1.0.1.el5 (OEL 5)
648 2.6.16.60-0.42.8 (SLES 10),
650 2.6.29.4-167.fc11 (FC11)
651 * Recommended e2fsprogs version: 1.41.10-sun2
653 Severity : enhancement
655 Description: Distribute lfs_migrate script with lustre
656 Details : The lfs_migrate mentioned in the manual is now included in the
657 lustre tree in lustre/scripts and will distributed in lustre RPMs.
659 Severity : enhancement
661 Description: Failover nid list added to import proc file.
663 Severity : enhancement
665 Description: "som_preview" mount option required to use SOM
666 Details : The clients will need to use som_preview mount option
670 Description: Lustre does not do 1MB IOs to HW RAID
671 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
673 Severity : enhancement
675 Description: lfs quota output is not very convenient for awk/sed-parsing
676 Details : Some positions in lfs quota output table could be empty or
677 non-empty which made it hard to parse it with scripts, now
678 a dash is put instead of space where there is not supposed
683 Description: Update to OFED 1.5.1
684 Details : OFED 1.5.1 will be used in environments that don't use
685 vendor supplied OFED.
689 Description: Remove lnet. prefix from lctl params display
690 Details : Also renamed lnet.memused to lnet_memused to
695 Description: Use conf_param -d to remove permanant settings.
699 Description: Add list_param -R
700 Details : lctl list_param now supports recursive mode.
703 Bugzilla : 21829, 22312
704 Description: Fix llobdstat.
705 Details : 1. Now updating the write_bytes counter
706 2. Port script fixes from 1.8.
710 Description: decrease the usage of memory on clients.
711 Details : 1. On clients, recycle dentries and inodes unused.
712 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
717 Description: llobdstat fix and enhancement
718 Details : add a counter to set a limit to how many samples will be returned
719 fix a wildcard in the path to limit to obdfilter stats only
723 Description: lfs check only allowed for root
724 Details : Code cleanups and added a regresion test of non-root use of
729 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
730 Details : Fix hr->hr_index overload.
734 Description: Fix FIEMAP functionality.
738 Description: FC11 patchless client support added
742 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
743 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
747 Description: New acc-sm script and various test-framework enhancments.
751 Description: Previously landed patch was reverted.
755 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
759 Description: Add ldiskfs maxdirsize mount option.
763 Description: Fix mkfs.lustre support for 16TB LUNs.
767 Description: Print mmp_check_interval
768 Details : print mmp_check_interval and make it possible to abort
769 mount operation in case it takes too long.
773 Description: Add cascading_rw to lustre mpi tests.
777 Description: Store SOM data in EA.
778 Details : Also several SOM related cleanups.
782 Description: Add refcount to md_open_data
786 Description: Use refcount on lov_request_set to prevent memory corruption.
790 Description: Statahead: use parent dir's i_mutex for synchronization.
794 Description: loadgen improvements.
798 Description: Enable peer health detection in ksocklnd and ko2iblnd.
802 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
806 Description: Fix race on obd_devcie:md_stats
810 Description: Fix SLES 11 client builds.
811 Details : Resolved uninitalized pointer gcc warnings.
815 Description: Allow non-root access for "lfs check".
816 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
820 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
822 Severity : enhancement
824 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
825 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
829 Description: Sanity test_27p fails.
830 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
831 Sanity tests 27 were cleaned up and improved.
835 Description: Unify req format on client/servers
836 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
837 buffers and swabbers approach..
842 Description: drop join file stuff.
844 Severity : enhancement
846 Description: Add 2.6.27 and SLES11 (patchless client) support
850 Description: Fix in ptlrpc_expire_one_request() to print the signed time
853 Severity : enhancement
855 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
858 Severity : enhancement
860 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
862 Severity : enhancement
864 Description: correctly handle big reply message.
865 Details : send LNet event if reply is bigger than buffer and adjust this buffer
870 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
872 Severity : enhancement
874 Description: Add support for OEL5.
878 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
880 Severity : enhancement
882 Description: "weak-modules" support
883 Details : Implement "weak-modules" support which enables kernel modules
884 to be used with any kernel that implements the same kABI. In
885 order to achieve this modules are now installed in
886 /lib/modules/$(uname -r)/updates/kernel on all distributions.
890 Description: client could not reconnect to OST because of an active request.
891 Details : The client now retries to reconnect to the same server, if a
892 connect request failed with EBUSY or -EAGAIN. Test case 35b
893 conf-sanity.sh has been created.
897 Description: Don't allow make backward step on assignin osc next id.
898 Details : race between allocation next id and ll_sync thread can be cause
899 of set wrong osc next id and can be kill valid ost objects.
903 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
907 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
911 Description: Error handling in osc_statfs_interpret() has been improved.
912 Details : Check in osc_statfs_interpret() for EBADR.
916 Description: Do not send statfs() requests to OSTs disabled by administrator.
917 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
919 Severity : enhancement
921 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
923 Severity : enhancement
925 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
927 Severity : enhancement
929 Description: Enhancement for lfs(1) command to use numeric uid/gid.
933 Description: Improved cleanup procedure for conf-sanity test_46a.
937 Description: Do not start bulk transfer, if client tried to reconnect.
938 Details : Do not start bulk transfer, if client tried to reconnect,
939 but failed due to active rpcs. Check for exp_abort_active_req
940 has been implemented.
944 Description: Prevent inconsistences between linux and lustre mount structures.
945 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
946 Make the sleep interruptible.
948 Severity : enhancement
950 Description: provide server to client comms path
951 Details : server to client communications path via new LDLM_SET_INFO rpc
955 Description: truncate starts GFP_FS allocation under transaction causing deadlock
956 Details : ldiskfs_truncate calls grab_cache_page which may start page
957 allocation under an open transaction. This may lead to
958 calling prune_icache with consequent lustre reentrance.
962 Description: send correctly lsm on open replay
963 Details : MDS is trust to LSM size on replay open, but client can set wrong size
968 Description: enable client interface failover
969 Details : When a child reconnects from another NID, properly update export
970 nid hash position and ldlm reverse import.
972 Severity : enhancement
974 Description: Communicate OST degraded/readonly state via statfs to MDS
975 Details : Flags in the statfs returned from OSTs indicate whether the
976 OST is in a degraded RAID state, or if the filesystem has
977 turned read-only after a filesystem error is detected.
982 Description: don't panic if EPROTO was hit when reading symlink
983 Details : correctly handling request reference in error cases.
985 Severity : enhancement
987 Description: Add LustreNetLink, a kernel-userspace communcation path.
989 Severity : enhancement
991 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
996 Description: access to llog context before init.
997 Details : move handling CATALOGS file at osc layer and forbid access to llog
1001 Frequency : rare, only when client is writing to OST on local node
1003 Description: Avoid deadlock for local client writes
1004 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1005 memory freeing context. This allows OST threads to set the
1006 PF_MEMALLOC flag on task structures in order to allocate memory
1007 from reserved pools and complete IO.
1008 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1009 so that the OST threads generate memory pressure and allow
1010 inactive pages to be reclaimed.
1012 Severity : enhancement
1014 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1019 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1020 Details : move ldlm namespace creation in setup phase to avoid grab
1021 _lprocfs_lock with cli_sem held.
1025 Description: Temporarily disable grant shrink.
1026 Details : Disable the feature for debugging.
1029 Frequency : only when formatting test filesystems
1031 Description: Unable to run several mkfs.lustre on loop devices at the same time
1032 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1033 formatting. The solution is to proper handle the error.
1035 Severity : enhancement
1037 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1039 Severity : enhancement
1041 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1046 Description: resolve race between obd_disconnect and class_disconnect_exports
1047 Details : if obd_disconnect will be called to already disconnected export he
1048 forget release one reference and osc module can't unloaded.
1050 Severity : enhancement
1052 Description: Allow tuning service thread via /proc
1053 Details : For each service a new
1054 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1055 created that can be used to set min/max thread counts, and get the
1056 current number of running threads.
1061 Descriptoin: don't return error if have partially created objects for file.
1062 Details : lov_update_create_set uses set->set_success as index for created
1063 objects, so if some requests will be failed, they will have hole
1064 at end of array and qos_shrink_lsm for allocate correct lsm.
1066 Severity : enhancement
1068 Description: Update OFED support to 1.4.1
1073 Description: Reduce small size read RPC
1074 Details : Set read-ahead limit for every file and only do read-ahead when
1075 available read-ahead pages are bigger than 1M to avoid small size
1078 Severity : enhancement
1080 Description: move AT tunable parameters for more consistent usage
1081 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1083 Severity : enhancement
1085 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1086 Details : allow skip disconnected ost for send statfs request and hide error
1092 Description: fix racy locking of mballoc block bitmaps causing BUG
1093 Details : The locking of the mballoc buddy bitmap and the in-memory
1094 block bitmap was using two different spin locks in some
1095 cases. This made it possible to incorrectly access the
1096 mballoc bitmap while another process was modifying it,
1097 causing a sanity assertion to fail. While no on-disk corruption
1098 was reported, there was some risk of this happening.
1101 Frequency : rare, on llog test 6
1103 Descriptoin: don't allow connect to already connected import
1104 Details : allowing connect to already connected import is hide connecting problem.
1109 Descriptoin: allow kill process which wait statahead result
1110 Details : for some reasons 'ls' can stick in waiting result from statahead,
1111 in this case need way for kill this process.
1113 Severity : enhancement
1115 Description: Add state history info file, enhance import info file
1116 Details : Track import connection state changes in a new osc/mdc proc file;
1117 add overview-type data to the osc/mdc import proc file.
1119 Severity : enhancement
1121 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1123 Severity : enhancement
1125 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1130 Descriptoin: OST grant too much space to client even there are not enough space.
1131 Details : Client will shrink its grant cache to OST if there are no write
1132 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1133 this grant cache if there are already not enough avaible space
1134 (left_space < total_clients * 32M).
1137 Frequency : start MDS on uncleanly shutdowned MDS device
1139 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1140 Details : stay in waiting mds<>ost recovery finished produce random bugs
1141 due race between two ll_sync thread for one lov target. send
1142 ACTIVATE event only if connect realy finished and import have
1146 Frequency : rare, connect and disconnect target at same time
1148 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1149 Details : don't call obd_disconnect under lov_lock. this long time
1150 operation and can block ptlrpcd which answer to connect request.
1155 Descriptoin: don't lose wakeup for imp_recovery_waitq
1156 Details : recover_import_no_retry or invalidate_import and import_close can
1157 both sleep on imp_recovery_waitq, but we was send only one wakeup
1161 Frequency : always with long access acl
1163 Descriptoin: mds can't pack reply with long acl.
1164 Details : mds don't control size of acl but they limited by reint/getattr
1167 Severity : enhancement
1169 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1171 Severity : enhancement
1173 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1176 Frequency : start MDS on uncleanly shutdowned MDS device
1178 Descriptoin: aborting recovery hang on MDS
1179 Details : don't throttle destroy RPCs for the MDT.
1182 Frequency : on remount
1184 Description: external journal device not working after the remount
1185 Details : clear dev_rdonly flag for external journal devices in
1191 Description: shutdown vs evict race
1192 Details : client_disconnect_export vs connect request race.
1193 if client will evicted at this time - we start invalidate
1194 thread without referece to import and import can be freed
1198 Frequency : rare, need acl's on inode.
1200 Description: client can't handle ost additional correctly
1201 Details : if ost was added after client connected to mds client can have
1202 hit lnet_try_match_md ... to big messages to wide striped files.
1203 in this case need teach client to handle config events about add
1204 lov target and update client max ea size at that event.
1206 Severity : enhancement
1208 Description: Changelogs
1209 Details : Changelogs are a lightweight mechanism to track filesystem
1210 metadata and namespace changes. The changelog is recorded
1211 permanently on the MDTs, and is periodically "consumed" / purged
1212 when records are no longer needed.
1214 Severity : enhancement
1216 Description: compact fld format with extents
1217 Details : Store range of seq rather than every seq in FLD. Seq
1218 controller update FLD rather than clients. In Case of CMD, mdt0
1219 has FLD, all other metadata server act as non persistent proxy
1220 for FLD queries and cache fld entries in fld cache.
1225 Description: don't skip ost target if they assigned to file
1226 Details : Drop slow OSCs if we can, but not for requested start idx.
1227 This means "if OSC is slow and it is not the requested
1228 start OST, then it can be skipped, otherwise skip it only
1229 if it is inactive/recovering/out-of-space.
1233 Description: more cleanup in mds_lov
1234 Details : not send LOV EA under replay, we can't know about they size at this
1235 time. Don't allow client connect to mds before any ost connected,
1236 for avoid problems with LOV EA size and returning EIO to client.
1238 Severity : enhancement
1240 Description: Interoperability at server side (Disk interoperability)
1242 Severity : enhancement
1244 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1246 Severity : enhancement
1248 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1250 Severity : enhancement
1252 Description: New client IO stack (CLIO).
1254 Severity : enhancement
1256 Description: Commit on sharing. Eliminate inter-client dependencies between
1257 uncommitted transactions by doing transaction commits.
1258 Thereby clients may recovery independently.
1261 Frequency : Create a symlink file with a very long name
1263 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1264 Details : If there is no extra space in the request for early cancels,
1265 ldlm_req_handles_avail() returns 0 instead of a negative value.
1267 Severity : enhancement
1269 Description: Add /proc entry for import status
1270 Details : The mdc, osc, and mgc import directories now have
1271 an import directory that contains useful import data for debugging
1272 connection problems.
1274 Severity : enhancement
1276 Description: Re-disable certain /proc logging
1277 Details : Enable and disable client's offset_stats, extents_stats and
1278 extents_stats_per_process stats logging on the fly.
1281 Frequency : Only on FC kernels 2.6.22+
1283 Description: oops in statahead
1284 Details : Do not drop reference count for the dentry from VFS when lookup,
1285 VFS will do that by itself.
1287 Severity : enhancement
1289 Description: Generic /proc file permissions
1290 Details : Set /Proc file permissions in a more generic way to enable non-
1291 root users operate on some /proc files.
1295 Description: Hitting mdc_commit_close() ASSERTION
1296 Details : Properly handle request reference release in
1297 ll_release_openhandle().
1301 Description: quota recovery deadlock during mds failover
1302 Details : This patch includes att18982, att18236, att18237 in bz14840.
1304 1. fix osts hang when mds does failover with quotaon
1305 2. prevent watchdog storm when osts threads wait for the
1310 Frequency : only patchless client
1311 Description: add workaround for race between add/remove dentry from hash
1313 Severity : enhancement
1315 Description: Allow OST glimpses to return PW locks
1319 Description: LBUG when llog conf file is full
1320 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1324 Description: Prevent import from entering FULL state when server in recovery
1328 Description: service mount cannot take device name with ":"
1329 Details : Only when device name contains ":/" will mount treat it as
1335 Description: replace ptlrpcd with the statahead thread to interpret the async
1336 statahead RPC callback
1340 Frequency : on recovery
1341 Description: I/O failures after umount during fail back
1342 Details : if client reconnected to restarted server we need join to recovery
1343 instead of find server handler is changed and process self
1344 eviction with cancel all locks.
1346 Severity : enhancement
1348 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1350 Severity : enhancement
1352 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1354 Severity : enhancement
1356 Description: Upcall on Lustre log has been dumped
1357 Details : Allow for a user mode script to be called once a Lustre log has
1358 been dumped. It passes the filename of the dumped log to the
1359 script, the location of the script can be specified via
1360 /proc/sys/lnet/debug_log_upcall.
1365 Description: avoid idr_remove called for id which is not allocated.
1366 Details : Move assigment s_dev for clustered nfs to end of initialization,
1367 to avoid problem with error handling.
1372 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1373 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1374 into CONN_USED_HASH and this prodice warning when put connection
1375 again in unused hash.
1378 Severity : enhancement
1380 Description: File striping can now be set to use an arbitrary pool of OSTs.
1382 Severity : enhancement
1384 Description: Export bytes_read/bytes_write count on OSC/OST.
1388 Description: Early reply size mismatch, MGC loses connection
1389 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1390 the connect flags are properly negotiated.
1395 Description: quotas are not honored with O_DIRECT
1396 Details : all writes with the flag O_DIRECT will use grants which leads to
1397 this problem. Now using OBD_BRW_SYNC to guard this.
1401 Description: add quota statistics
1402 Details : 1. sort out quota proc entries and proc code.
1403 2. add quota statistics
1405 Severity : enhancement
1407 Description: enable quota support for HEAD.
1411 Description: Properly propagate oinfo flags from lov to osc for statfs
1412 Details : restore missing copy oi_flags to lov requests.
1414 Severity : enhancement
1416 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1417 : l_getgroups(8), lst(8), routerstat(8)
1418 Details : included man pages for llobdstat(8), llstat(8),
1419 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1421 Severity : enhancement
1423 Description: configure's --enable-quota should check the
1424 : kernel .config for CONFIG_QUOTA
1425 Details : configure is terminated if --enable-quota is passed but
1426 : no quota support is in kernel
1430 Description: Remove portals compatibility
1431 Details : Remove portals compatibility, not interoperable with releases
1436 Description: Resolve device initialization race
1437 Details : Prevent proc handler from accessing devices added to the
1438 obd_devs array but yet be intialized.
1440 Severity : enhancement
1442 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1444 Severity : enhancement
1446 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1451 Description: Using wrong pointer in osc_brw_prep_request
1452 Details : Access to array[-1] can produce panic if kernel compiled with
1453 CONFIG_PAGE_ALLOC enabled
1457 Description: Client runs out of low memory
1458 Details : Consider only lowmem when counting initial number of llap pages
1462 Description: *optional* service tags registration
1463 Details : if the "service tags" package is installed on a Lustre node
1464 When the filesystem is mounted, a local-node service tag will
1465 be created. See http://inventory.sun.com/ for more information
1466 about the Service Tags asset management system.
1470 Description: Kernel BUG tries to release flock
1471 Details : Lustre does not destroy flock lock before last reference goes
1472 away. So always drop flock locks when client is evicted and
1473 perform unlock regardless of successfulness of speaking to MDS.
1477 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1481 Description: sanity test 65a fails if stripecount of -1 is set
1482 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1487 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1488 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1491 Severity : enhancement
1493 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1497 Description: do not process already freed flock
1498 Details : flock can possibly be freed by another thread before it reaches
1499 to ldlm_flock_completion_ast.
1503 Description: LBUG during stress test
1504 Details : Need properly lock accesses the flock deadlock detection list.
1508 Description: oops in page fault handler
1509 Details : kernel page fault handler can return two special 'pages' in error
1510 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1514 Description: timeout with invalidate import.
1515 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1516 should be handled by ptlrpcd. This produce long age waiting and
1517 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1519 Severity : enhancement
1521 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1525 Description: enable MGS and MDT services start separately
1526 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1527 a co-located MGS without starting the MGS, which is a complement
1528 to 'nosvc' mount option.
1532 Frequency : after recovery
1533 Description: precreate to many object's after del orphan.
1534 Details : del orphan st in oscc last_id == next_id and this triger growing
1535 count of precreated objects. Set flag LOW to skip increase count
1536 of precreated objects.
1540 Frequency : rare, on clear nid stats
1541 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1542 Details : when clean nid stats sometimes try destroy live entry,
1543 and this produce panic in free.
1547 Description: Stack overflow during MDS log replay
1548 ease stack pressure by using a thread dealing llog_process.
1552 Description: wait until IO finished before start new when do lock cancel.
1553 Details : VM protocol want old IO finished before start new, in this case
1554 need wait until PG_writeback is cleared until check dirty flag and
1555 call writepages in lock cancel callback.
1557 Severity : enhancement
1559 Description: using special macro for print time and cleanup in includes.
1563 Description: mds_mfd_close() ASSERTION(rc == 0)
1564 Details : In mds_mfd_close(), we need protect inode's writecount change
1565 within its orphan write semaphore to prevent possible races.
1569 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1573 Frequency : rare, on shutdown ost
1574 Description: don't hit live lock with umount ost.
1575 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1576 use shrink_dcache_sb instead.
1580 Description: don't panic with use echo client
1581 Details : echo client pass NULL as client nid pointer and this produce null
1582 pointer dereference.
1586 Description: fix build on ppc32
1587 Details : compile code with -m64 flag produce wrong object file for ppc32.
1591 Description: add message levels for liblustreapi
1595 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1596 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1597 will result in an -EAGAIN and in the exit call path we call
1598 alloc_rr() which will with increasing aggressiveness attempt to
1599 aquire precreated objects on the minimum number of required OSCs.
1603 Description: Use old size assignment to avoid deadlock
1604 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1605 the scheduling while holding a spinlock. We do not need locking
1606 for size in ll_update_inode() because size is only updated from
1607 the MDS for directories or files without objects, so there is no
1608 other place to do the update, and concurrent access to such inodes
1609 are protected by the inode lock.
1613 Description: resolve "_IOWR redefined" build error on SLES10
1617 Description: dump the memory debugging after all modules are unloaded to
1618 suppress false negative in conf_sanity test 39
1620 Severity : enhancement
1622 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1627 Description: fix SLES kernel versioning
1628 Details : the kernel version for our SLES 10 kernel did not include a "-"
1629 before the "smp" at the end. while this was not a problem in
1630 general, it did mean that software trying to use the kernel
1631 version to try to detect a vendor specific kernel would fail.
1632 this was most evident by the OFED build scripts.
1636 Description: Don't update lov_desc members until making sure they are valid
1637 Details : When updating lov_desc members via proc fs, need fix their
1638 validities before doing the real update.
1642 Description: don't put request into delay list while invalidate in flight.
1643 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1644 import in flight. this produce timeout for invalidate and sometimes
1645 can cause stale data.
1649 Frequency : on ppc only
1650 Description: not convert ost objects for directory because it's not exist.
1651 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1655 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1657 Severity : enhancement
1659 Description: Update to support 2.6.22.14 vanilla kernel.
1663 Frequency : rare, on recovery
1664 Description: read procfs can produce deadlock in some situation
1665 Details : Holding lprocfs lock which send rpc can produce block for destroy
1666 obd objects and this also block reconnect with -EALREADY. This
1667 isn't fix all lprocfs bugs - but make it rare.
1669 Severity : enhancement
1671 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1674 Frequency : frequent on X2 node
1676 Description: mdc_set_open_replay_data LBUG
1677 Details : Set replay data for requests that are eligible for replay.
1681 Description: lustre_mgs: operation 101 on unconnected MGS
1682 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1683 MGC, and late on MGC cannot successfully connect to MGS and a lot
1684 of the error messages complaining that MGS is not connected.
1688 Frequency : on network error
1689 Description: panic with double free request if network error
1690 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1691 but it's true only for synchronus enqueue, for async enqueue
1692 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1695 Severity : enhancement
1697 Description: client-side metadata stat-ahead during readdir(directory readahead)
1698 Details : perform client-side metadata stat-ahead when the client detects
1699 readdir and sequential stat of dir entries therein
1702 Frequency : on start mds
1704 Description: Implement get_info(last_id) in obdfilter.
1707 Frequency : occasional
1709 Description: Correctly check stale fid, not start epoch if ost not support SOM
1710 Details : open with flag O_CREATE need set old fid in op_fid3 because
1711 op_fid2 was overwritten with new generated fid, but mds can answer
1712 with one of these two fids and both is not stale. Setattr
1713 incorrectly started an epoch and assume will be called
1714 done_writing, but without SOM done_writing ever being called.
1717 Frequency : rare, depends on device drivers and load
1719 Description: MDS or OSS nodes crash due to stack overflow
1720 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1721 In some cases, in conjunction with device drivers that use a lot
1722 of stack the MDS (or possibly OSS) service threads could overflow
1723 the stack. One change which was identified to consume additional
1724 stack has been reworked to avoid the extra stack usage.
1727 Frequency : occasional
1729 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1730 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1731 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1732 oscc_has_objects and see OSC has no objests and call
1733 oscc_internal_create to resend create request.
1735 Severity : enhancement
1737 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1739 Severity : enhancement
1741 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1744 Frequency : very rare
1746 Description: Fix replay if there is an un-replied request and open
1747 Details : In some cases, older replay request will revert the
1748 mcd->mcd_last_xid on MDS which is used to record the client's
1749 latest sent request.
1751 Severity : enhancement
1753 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1755 Severity : enhancement
1757 Description: Add rhel5 support to HEAD.
1759 Serverity : enhancement
1761 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1766 Description: Don't allow skipping OSTs if index has been specified.
1767 Details : Don't allow skipping OSTs if index has been specified, make
1768 locking in internal create lots better.
1772 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1773 Details : access to bitfield in structure is always rounded to long
1774 and this produce problem with not atomic change any bit.
1778 Description: Lustre make rpms failed.
1779 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1780 builds Lustre rpms from tarball.
1783 Frequency : rare on shutdown ost
1785 Description: If llog cancel was not send before clean_exports phase, this can
1786 produce deadlock in llog code.
1787 Details : If llog thread has last reference to obd and call class_import_put
1788 this produce deadlock because llog_cleanup_commit_master wait when
1789 last llog_commit_thread exited, but this never success because was
1790 called from llog_commit_thread.
1794 Description: allow userland application know is lost one of stripes.
1795 Details : fill lvb_blocks with error code on ost and return it to
1796 application if error flag found.
1800 Description: NULL lov_tgts causing MDS oops
1801 Details : more safe checks for NULL lov_tgts for avoid oops.
1803 Severity : enhancement
1805 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1809 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1810 Details : If an OST is down the MDS will hang indefinitely in
1811 obd_statfs_rqset() waiting for the statfs data. While for
1812 MDS QOS usage of statfs, it should not stuck in waiting.
1814 Severity : enhancement
1816 Description: remote_acl support
1817 Details : Support ACL-based permission check for remote user.
1818 Support setfacl/getfacl for remote user with the utils
1819 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1820 the system "{s,g}etfacl" utils.
1822 Severity : enhancement
1824 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1826 Severity : enhancement
1828 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1832 Description: make mgs_setparam() handle fsname containing dash
1833 Details : fsname containing a dash does not work with lctl conf_param
1835 Severity : enhancement
1837 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1839 Severity : enhancement
1841 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1845 Frequency : rare, at shutdown
1846 Description: access already free / zero obd_namespace.
1847 Details : if client_disconnect_export was called without force flag set,
1848 and exist connect request in flight, this can produce access to
1849 NULL pointer (or already free pointer) when connect_interpret
1850 store ocd flags in obd_namespace.
1854 Frequency : only at startup
1855 Description: not alloc memory with spinlock held.
1856 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1857 if any spinlock held.
1859 Severity : enhancement
1861 Description: make lustre randomly fail allocating memory
1862 Details : Make lustre randomly failed allocating memory for testing purpose.
1864 Severity : enhancement
1866 Description: lost problems with lov objid file
1867 Details : Fixes some scability and access to not inited memory problems
1868 in work with lov objdid file.
1873 Description: lfs find does not continue on file error
1874 Details : Continue other files processing when a file/dir is absent.
1878 Description: Inconsistent usage of lustre_pack_reply()
1879 Details : Standardize the usage of lustre_pack_reply() such that it
1880 always generate a CERROR on failure.
1883 Frequency : occasional
1885 Description: MDS hang or stay in waiting lock
1886 Details : If client receive lock with CBPENDING flag ldlm need send lock
1887 cancel as separate rpc, to avoid situation when cancel request
1888 can't processed due all i/o threads stay in wait lock.
1892 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1896 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1900 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1904 Description: Update to RHEL4 latest kernel.
1906 Severity : enhancement
1908 Description: Build SLES10 patchless client fails
1909 Details : The configure was broken by run ./configure with
1910 --with-linux-obj=.... argument for patchless client. When the
1911 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1912 can't search header adequately. Use absolute path such as
1913 -I($LINUX)/include instead.
1917 Description: interrupt oig_wait produce painc on resend.
1918 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1919 private set, and this produce situation when rq_ptlrpcd_data not
1920 copyed to new allocated request and triggered LBUG on assert
1921 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1922 ptlrpcd set if request is changed and can be safety changed to use
1927 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1928 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1929 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1930 these macros and replace them with their difinitions instead.
1934 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1935 Details : Update which_patch & target file for SLES10 latest kernel.
1938 Description: Remove client patches from SLES 10 kernel.
1939 Details : This causes SLES 10 clients to behave as patchless clients
1940 even on a Lustre-patched (server) kernel.
1942 Severity : enhancement
1944 Description: self-adjustable client's lru lists
1945 Details : use adaptive algorithm for managing client cached locks lru
1946 lists according to current server load, other client's work
1947 pattern, memory activities, etc. Both, server and client
1948 side namespaces provide number of proc tunables for controlling
1951 Severity : enhancement
1953 Description: light-weight GSS support
1954 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1955 and reduce performance overhead.
1957 Severity : enhancement
1959 Description: Linux keyring support
1960 Details : Support using service of Linux keyring for Lustre GSS internal
1961 context refresh/cache mechanism.
1965 Description: Fix errors in lfs documentation
1966 Details : Fixes man pages
1970 Description: don't use GFP_* in generic Lustre code.
1971 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1975 Description: obdclass is limited by single OBD_ALLOC(idarray)
1976 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1980 Description: warnings with build patchless client with vanila 2.6.19 and up
1981 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1982 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1986 Frequency : Only for SLES
1987 Description: sanity_quota fail test_1
1988 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1989 file, which makes TSTID[2] inunique.
1993 Frequency : Only for SLES9
1994 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1995 Details : Update target/ChangeLog/which_patch .
1999 Frequency : Only for test_10 in sanity-quota.sh
2000 Description: a bug in quota test code
2001 Details : checking a test flag in a wrong place
2005 Frequency : Only for test_13 in sanity-quota.sh
2006 Description: a bug in quota test script
2007 Details : shouldn't check overfull allocated quota
2011 Description: jbd statistics
2012 Details : Port older jbd statistics patch for sles10
2014 Severity : enhancement
2016 Description: Add printing inode info into message about error in writepage.
2021 Description: reply_lock_interpret crash due to race with it and lock cancel.
2022 Details : Do not replay locks that are being cancelled. Do not reference
2023 locks by their address during replay, just by their handle.
2027 Frequency : When flocks are used.
2028 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2029 Details : Flock locks might destroy just granted lock if it could be merged
2030 with another existing flock, this is done in completion handler,
2031 so teach ldlm_cli_enquque_fini that this is a valid case for
2037 Description: Oops in read and write path when failing to allocate lock.
2038 Details : Check if lock allocation failed and return error back.
2042 Description: lstripe command fails for valid OST index
2043 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2044 instead of lov->desc.ld_active_tgt_count.
2048 Description: sanity.sh failed test 103
2049 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2054 Description: df doesn't work properly if diskfs blocksize != 4K
2055 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2059 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2060 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2062 Severity : enhancement
2064 Description: Adaptive timeouts
2065 Details : RPC timeouts adapt to changing server load and network
2066 conditions to reduce resend attempts and improve recovery time.
2070 Description: llapi_file_create() does not allow some changes
2071 Details : add llapi_file_open() that allows specifying the mode and
2072 open flags, and also returns an open file handle.
2076 Description: merge and cleanup kernel patches.
2077 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2081 Description: Add journal checksum support.(Kernel part)
2082 Details : The journal checksum feature adds two new flags i.e
2083 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2084 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2085 indicates that the commit block contains the checksum for
2086 the blocks described by the descriptor blocks. Now commit
2087 record can be sent to disk without waiting for descriptor
2088 blocks to be written to disk. This behavior is controlled
2089 using JBD2_FEATURE_ASYNC_COMMIT flag.
2093 Description: OSS needs mutliple precreate threads
2094 Details : Add ability to start more than one create thread per OSS.
2098 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2099 Details : execute lfs setstripe on client
2103 Description: mds_obd_create error creating tmp object
2104 Details : When the user sets quota on root, llog will be affected and can't
2105 create files and write files.
2108 Frequency : Always on ia64 patchless client, and possibly others.
2110 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2111 Details : This allows the patchless client to be loaded on architectures
2112 without this export.
2116 Description: RedHat Update kernel for RHEL5
2117 Details : Add the kernel config file for RHEL5.
2121 Description: RedHat Update kernel for RHEL5
2122 Details : Modify the kernel config file more closer RHEL5.
2126 Description: Build failure against Centos5 (RHEL5)
2127 Details : Define PAGE_SIZE when it isn't present.
2131 Description: test_7 on sanity-quota.sh can't be used on separate machines
2132 Details : add facet to handle it
2136 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2137 Details : with NFS, the anon dentry's parent was set to itself in
2138 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2139 corresponding dentry other than use rec->ur_name.
2141 Severity : enhancement
2143 Description: lfs setstripe enhancement
2144 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2148 Description: enable data checksumming by default
2149 Details : enable checksum by default, allow --disable-checksum
2150 configure option and "-o nochecksum" mount option. Checksums
2151 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2155 Description: lustre support for RHEL5
2156 Details : Add support for RHEL5.
2160 Description: Client eviction due to failover config
2161 Details : after a connection loss, the lustre client should attempt to
2162 reconnect to the last active server first before trying the
2163 other potential connections.
2167 Description: when mds and osts use different quota unit(32bit and 64bit),
2168 quota will be released repeatly.
2169 Details : void sending multiple quota reqs to mds, which will keep the
2170 status between the reqs.
2174 Description: osts not allocated evenly to files
2175 Details : change the condition to increase offset_idx
2178 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2180 Description: Data corruption for OSTs that are formatted larger than 2TB
2182 Details : When generating the bio request for lustre file writes the
2183 sector number would overflow a temporary variable before being
2184 used for the IO. The data reads correctly from Lustre (which
2185 will overflow in a similar manner) but other file data or
2186 filesystem metadata may be corrupted in some cases.
2190 Description: Tune the kernel for good SCSI performance.
2191 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2192 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2197 Description: rewrite ext2-derived code in obdclass/uuid.c
2198 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2199 from scratch preserving functionality.
2203 Description: "lfs find -obd UUID" prints directories
2204 Details : "lfs find -obd UUID" will return all directory names instead
2205 of just file names. It is incorrect because the directories
2206 do not reside on the OSTs.
2210 Description: use i_size_read and i_size_write in 2.6 port
2211 Details : replace inode->i_size access with i_size_read/write()
2214 Frequency : when removing large files
2216 Description: scheduling issue during removal of large Lustre files
2217 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2218 It causes scheduling issues when removing large files (17TB in the
2222 Frequency : only with liblustre clients on XT3
2224 Description: evictions taking too long
2225 Details : allow llrd to evict clients directly on OSTs
2228 Frequency : only on ppc
2230 Description: /proc/fs/lustre/devices broken on ppc
2231 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2232 We should make sure the type of 'index' is loff_t and then cast
2233 explicitly as needed below. Do not assign an explicitly cast
2237 Frequency : only for rhel5
2239 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2240 Details : Modify the target file & which_kernel.
2242 Severity : enhancement
2244 Description: omit set fsid for export NFS
2245 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2246 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2250 Frequency : Always, for kernels after 2.6.16
2251 Description: Fix warning idr_remove called for id=.. which is not allocated.
2252 Details : Last kernels save old s_dev before kill super and not allow
2253 to restore from callback - restore it before call kill_anon_super.
2257 Description: buffer overruns could theoretically occur
2258 Details : llapi_semantic_traverse() modifies the "path" argument by
2259 appending values to the end of the origin string, and a
2260 overrun may occur. Adding buffer overrun check in liblustreapi.
2264 Description: Fix error on 'ls .' at the top of the Lustre mount.
2265 Details : Don't revalidate dentry if it is a root dentry.
2269 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2270 Details : Modify vm-tunables-rhel4.patch.
2274 Description: Kernel config for 2.6.18-vanilla.
2275 Details : Modify targets/2.6-vanilla.target.in.
2276 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2277 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2278 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2279 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2283 Description: improve handling recoverable errors
2284 Details : if request processig with error which can be recoverable on server
2285 request should be resend, otherwise page released from cache and
2290 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2291 Details : Modify target file & which_patch.
2292 A flaw was found in the IA32 system call emulation provided
2293 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2294 value could be stored in the %RAX register, which could trigger an
2295 out-of-bounds system call table access. An untrusted local user
2296 could exploit this flaw to run code in the kernel
2297 (ie a root privilege escalation). (CVE-2007-4573).
2301 Description: O_DIRECT bypasses client statistics.
2302 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2303 read_ahead_stats were updated. copied the stats section from
2304 osc_send_oap_rpc() into async_internal().
2308 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2312 Description: change order of libsysio includes
2313 Details : '#include sysio.h' should always come before '#include xtio.h'
2317 Description: handle "serious error: objid * already exists" more gracefully
2318 Details : If LAST_ID value on disk is smaller than the objects existing in
2319 the O/0/d* directories, it indicates disk corruption and causes an
2320 LBUG(). If the object is 0-length, then we should use the existing
2321 object. This will help to avoid a full fsck in most cases.
2323 Severity : enhancement
2325 Description: adapt the lustre_config script to support the upgrade case
2326 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2327 server targets to 1.6 in parallel.
2331 Description: To avoid grant space > avaible space when the disk is almost
2332 full. Without this patch you might see the error "grant XXXX >
2333 available" or some LBUG about grant, when the disk is almost
2335 Details : In filter_check_grant, for non_grant cache write, we should
2336 check the left space by if (*left > ungranted + bytes), instead
2337 of (*left > ungranted), because only we are sure the left space
2338 is enough for another "bytes", then the ungrant space should be
2339 increase. In client, we should update cl_avail_grant only there
2340 is OBD_MD_FLGRANT in the reply.
2344 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2345 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2346 described in CVE-2007-4573. This problem could allow untrusted
2347 local users to gain root access.
2350 Frequency : when using O_DIRECT and quotas
2352 Description: Incorrect file ownership on O_DIRECT output files
2353 Details : block usage reported by 'lfs quota' does not take into account
2354 files that have been written with O_DIRECT.
2359 Description: touch file failed when fs is not full
2360 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2361 otherwise we can get ENOSP while fs is not full.
2365 Description: parallel lock callbacks
2366 Details : Instead of sending blocking and completion callbacks as separated
2367 requests, adding them to a set and sending in parallel.
2370 Frequency : only for Cray XT3
2371 Bugzilla : 12829/13455
2372 Description: Changing primary group doesn't change the group lustre assigns to
2374 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2375 overridden with the primary group provided by the group upcall,
2376 whereas the supplied fsgid can be trusted if it is in the list of
2377 supplementary groups returned by the group upcall.
2379 Severity : enhancement
2381 Description: Allow masking D_WARNING, D_ERROR messages from console
2382 Details : Console messages can now be disabled via lnet.printk.
2386 Description: User code with malformed file open parameter crashes client node
2387 Details : Before packing join_file req, all the related reference should be
2388 checked carefully in case some malformed flags cause fake
2389 join_file req on client.
2393 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2394 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2395 mode is equal to the granted mode. However, it should also check
2396 that LDLM_FL_CP_REQD is not set.
2400 Description: Heavy nfs access might result in deadlocks
2401 Details : After ELC code landed, it is now improper to enqueue any mds
2402 locks under och_sem, because enqueue might want to decide to
2403 cancel open locks for same inode we are holding och_sem for.
2407 Description: Client eviction while running blogbench
2408 Details : A lot of unlink operations with concurrent I/O can lead to a
2409 deadlock causing evictions. To address the problem, the number of
2410 oustanding OST_DESTROY requests is now throttled to
2411 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2412 callbacks are processed in priority.
2416 Description: enable ACLs on MDS by default
2417 Details : ACLs must be enabled on MDS by default.
2420 Frequency : PPC/PPC64 only
2422 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2423 Details : fix duplicated definitions between asm-ppc64/types.h and
2424 lustre_types.h on PPC.
2427 Frequency : PPC/PPC64 only
2429 Description: asm-ppc/segment.h does not exist
2430 Details : fix compile issue on PPC.
2434 Description: better handle error messages in extents code
2437 Frequency : RHEL4 only
2439 Description: mkfs is very slow on IA64/RHEL4
2440 Details : A performance regression has been discovered in the MPT Fusion
2441 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2442 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2443 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2445 Severity : enhancement
2447 Description: SNMP support enhancement
2448 Details : Adding total number of sampled request for an MDS node in snmp
2451 Severity : enhancement
2453 Description: Optimize ldlm waiting list processing for PR extent locks
2454 Details : When processing waiting list for read extent lock and meeting read
2455 lock that is same or wider to it that is not contended, skip
2456 processing rest of the list and immediatelly return current
2457 status of conflictness, since we are guaranteed there are no
2458 conflicting locks in the rest of the list.
2462 Description: Time out and refuse to reconnect
2463 Details : When the failover node is the primary node, it is possible
2464 to have two identical connections in imp_conn_list. We must
2465 compare not conn's pointers but NIDs, otherwise we can defeat
2466 connection throttling.
2470 Description: port llog fixes from b1_6 into HEAD
2471 Details : Port llog reference couting and some llog cleanups from b1_6
2472 (bug 10800) into HEAD, for protect from panic and access to already
2473 free llog structures.
2477 Description: Detect stride IO mode in read-ahead
2478 Details : When a client does stride read, read-ahead should detect that and
2479 read-ahead pages according to the detected stride pattern.
2483 Description: data checksumming impacts single node performance
2484 Details : add support for several checksum algorithm. Currently, only CRC32
2485 and Adler-32 are supported. The checksum type can be changed on
2486 the fly via /proc/fs/lustre/osc/*/checksum_type.
2490 Description: use adler32 for page checksums
2491 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2496 Description: build for x2 fails
2497 Details : fix compile issue on Cray systems.
2501 Description: Properly match for duplicate locks
2502 Details : Due to different lock order from skiplists code, we need to
2503 traverse entire list for now
2506 Frequency : only on PPC/SLES10
2508 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2509 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2510 be defined. Add a hack in mkfs_lustre.c to work around this compile
2515 Description: LASSERT on MDS when client holding flock lock dies
2516 Details : ldlm pool logic depends on number of granted locks equal to
2517 number of released locks which is not true for flock locks, so
2518 just exclude such locks from consideration.
2522 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2523 Details : Use fsfilt_sync() for both the whole filesystem sync and
2524 individual file sync to eliminate dangerous inode locking
2525 with I_LOCK that can lead to a deadlock.
2529 Description: performance in 1.6.3
2530 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2531 sure that 1MB requests are not fragmented by the block layer.
2533 Severity : enhancement
2535 Description: organize the server-side client stats on per-nid basis
2536 Details : Change the structure of stats under obdfilter and mds to
2544 The "uuid"s file would list the uuids of _active_ exports.
2545 And the clear entry is to clear all stats and stale nids.
2547 Severity : enhancement
2549 Description: eliminate client locks in face of contention
2550 Details : file contention detection and lockless i/o implementation
2551 for contended files.
2555 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2558 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2560 Description: MDS may lose file striping (and hence file data) in some cases
2561 Details : If there are additional extended attributes stored on the MDS,
2562 in particular ACLs, SELinux, or user attributes (if user_xattr
2563 is specified for the client mount options) then there is a risk
2564 of attribute loss. Additionally, the Lustre file striping
2565 needs to be larger than default (e.g. striped over all OSTs),
2566 and an additional attribute must be stored initially in the
2567 inode and then increase in size enough to be moved to the
2568 external attribute block (e.g. ACL growing in size) for file
2573 Description: skiplist implementation simplification
2574 Details : skiplists are used to group compatible locks on granted list
2575 that was implemented as tracking first and last lock of each
2576 lock group the patch changes that to using doubly linked lists
2580 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2581 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2582 because LASSERT on both the data supplied by a client, and the
2583 data on disk is dangerous and incorrect.
2585 Severity : enhancement
2587 Description: Slow truncate/writes to huge files at high offsets.
2588 Details : Directly associate cached pages to lock that protect those pages,
2589 this allows us to quickly find what pages to write and remove
2590 once lock callback is received.
2594 Description: more ldlm soft lockups
2595 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2596 starve other threads from the resource lock for a long time in
2597 case of long waiting queue, so change the debug level from
2598 D_OTHER to the less frequently used D_INFO.
2600 Severity : enhancement
2602 Description: add -gid, -group, -uid, -user options to lfs find
2606 Description: Hung threads in invalidate_inode_pages2_range
2607 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2608 once one is completed. As a result, some RPCs are stuck in the
2609 queue and are never sent.
2613 Description: filter threads hungs on waiting journal commit
2614 Details : Cleanup filter group llog code, then only filter group llog will
2615 be only created in the MDS/OST syncing process.
2619 Description: Procfs and llog threads access destoryed import sometimes.
2620 Details : Sync the import destoryed process with procfs and llog threads by
2621 the import refcount and semaphore.
2623 Severity : enhancement
2625 Description: openlock cache of b1_6 port to HEAD
2630 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2631 Details : If insertion of an extent fails, then discard the inode
2632 preallocation and free data blocks else it can lead to duplicate
2637 Description: don't always update ctime in ext3_xattr_set_handle()
2638 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2639 In some cases the ctime should not be updated, for example for
2640 2.0->1.8 compatibility it is necessary to delete an xattr and it
2641 should not update the ctime.
2645 Bugzilla : 15713/16362
2646 Description: Assertion in iopen_connect_dentry in 1.6.3
2647 Details : looking up an inode via iopen with the wrong generation number can
2648 populate the dcache with a disconneced dentry while the inode
2649 number is in the process of being reallocated. This causes an
2650 assertion failure in iopen since the inode's dentry list contains
2651 both a connected and disconnected dentry.
2655 Description: assertion failure in ldlm_handle2lock()
2656 Details : fix a race between class_handle_unhash() and class_handle2object()
2657 introduced in lustre 1.6.5 by bug 13622.
2662 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2663 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2668 Description: oops in sd_iostats_seq_show()
2669 Details : unloading/reloading the scsi low level driver triggers a kernel
2670 bug when trying to access the sd iostat file.
2675 Description: Kernel panics during QLogic driver reload
2676 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2677 patch, causing memory corruption.
2682 Description: journal_dev option does not work in b1_6
2683 Details : pass mount option during pre-mount.
2685 Severity : enhancement
2687 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2688 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2689 extent information of a file. It can be used to map logical blocks
2690 in a file to physical blocks in the block device.
2694 Description: LDLM soft lockups - improvement
2695 Details : It is be possible to send the lock handle along with each read
2696 or write request because the client is already doing a lock match
2697 itself so there isn't any reason the OST should have to re-do that
2701 Frequency : only on Cray X2
2703 Description: X2 build failures
2704 Details : fix build failures on Cray X2.
2708 Description: xid & resent requests
2709 Details : Initialize RPC XID from clock at startup (randomly if clock is
2712 Severity : enhancement
2714 Description: Add lustre_start utility to start or stop multiple Lustre servers
2719 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2720 Details : In case of memory pressure, list_del() can be called twice on
2721 req->rq_history_list, causing a kernel oops.
2725 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2726 Details : fix stack overflow in the distributed lock manager by defering
2727 export eviction after a failed AST to the elt thread instead of
2728 handling it in the dlm interpret routine.
2732 Description: Convert some comments to new format.
2733 Details : Update documenting comments to match doxygen conventions.
2737 Description: Grammar fixes.
2738 Details : A couple of trivial spelling fixes.
2742 Description: OSD_COUNTERS-mandatory
2743 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2744 to simplify the code.
2748 Description: simplify cmm_device freeing logic.
2749 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2753 Description: Add lockdep support to dt_object_operations locking interface.
2754 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2755 parameter indicating lock ordering. Update mdd code to use new
2760 Description: Introduce failloc constants for lockless IO tests.
2761 Details : Add two new failloc constants to test lockless IO. Only one of
2762 them in implemented---another is checked in yet to be landed
2767 Description: Add lockdep support for inode mutex.
2768 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2769 in the situations where Lustre has to lock more than one inode
2774 Description: Add optional invariants checking support.
2775 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2776 configure switch. This macro is to be used for consistency and
2777 sanity checks that are too expensive to be left in `production'
2782 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2783 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2788 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2789 Details : Use spin_lock_nested() in (the only) situation where more than
2790 one ldlm_lock is locked simultaneously. Also, fix possible
2791 dead-lock in ldlm_lock_change_resource() by enforcing particular
2796 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2797 Details : Instead of specifying each ldlm_lock call-back through separate
2798 parameter, wrap them into struct ldlm_callback_suite.
2802 Description: Kill join_lru obd method and its callers.
2803 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2804 in memory---a requirement implemented through ->o_join_lru() obd
2805 method in HEAD. Remove this method and its users.
2809 Description: Add asynchronous ldlm ENQUEUE completion handler.
2810 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2811 case that stock ldlm_completion_ast() cannot handle as it waits
2812 until lock is granted. Introduce new ldlm_completion_ast_async()
2813 for this. Also comment ldlm_completion_ast().
2817 Description: ldlm_error <-> errno conversion.
2818 Details : Add functions to map (rather arbitrary) between LDLM error codes
2819 and standard errno values. CLIO needs this to prevent LDLM specific
2820 constants from escaping ldlm and osc.
2824 Description: Kill unused ldlm_handle2lock_ns() function.
2825 Details : Kill unused ldlm_handle2lock_ns() function.
2829 Description: Add lu_ref support to ldlm_lock
2830 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2831 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2832 and ldlm_resource. LDLM interface has to be changed, because code that
2833 releases a reference on a lock, has to "know" what reference this is.
2834 In the most frequent case
2836 lock = ldlm_handle2lock(handle);
2838 LDLM_LOCK_PUT(lock);
2840 no changes are required. When any other reference (received _not_ from
2841 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2842 instead of LDLM_LOCK_PUT().
2844 Arguably, changes are pervasive, and interface requires some discipline
2845 for proper use. On the other hand, it was very instrumental in finding
2846 a few leaked lock references.
2850 Description: Add ldlm_lock_addref_try().
2851 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2852 attempts to addref a lock that might be being canceled
2857 Description: Add ldlm_weigh_callback().
2858 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2859 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2860 hard-coded `number of pages' logic.
2864 Description: Add lockdep annotations to llog code.
2865 Details : Use appropriately tagged _nested() locking calls in the places
2866 where llog takes more than one ->lgh_lock lock.
2870 Description: Add loi_kms_set().
2871 Details : Wrap kms updates into a helper function.
2875 Description: Constify instances of struct lsm_operations.
2876 Details : Constify instances of struct lsm_operations.
2880 Description: lu_conf support.
2881 Details : On a server, a file system object is uniquely identified
2882 by a fid, which is sufficient to locate and load all object
2883 state (inode). On a client, on the other hand, more data are
2884 necessary instantiate an object. Change lu_object_find() and
2885 friends to take additional `lu_conf' argument describing object.
2886 Typically this includes layout information.
2890 Description: lu_context fixes.
2891 Details : Introduce new lu_context functions that are needed on the client
2892 side, where some system threads (ptlrpcd) are shared by multiple
2893 modules, and so cannot be stopped during module shutdown.
2897 Description: Add start and stop methods to lu_device_type_operations.
2898 Details : Introduce two new methods in lu_device_type_operations, that are
2899 invoked when first instance of a given type is created and last one
2900 is destroyed respectively. This is need by CLIO.
2904 Description: Add lu_ref support to struct lu_device.
2905 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2906 track leaked references.
2910 Description: Introduce lu_kmem_descr.
2911 Details : lu_kmem_descr and its companion interface allow to create
2912 and destroy a number of kmem caches at once.
2916 Description: Fix lu_object finalization race.
2917 Details : Fix a race between lu_object_find() finding an object and its
2918 concurrent finalization. This race is (most likely) not possible
2919 on the server, but might happen on the client.
2923 Description: Introduce lu_ref interface.
2924 Details : lu_ref is a debugging module allowing to track references to
2925 a given object. It is quite cpu expensive, and has to be
2926 explicitly enabled with --enable-lu_ref. See usage description
2931 Description: Factor lu_site procfs stats into a separate function.
2932 Details : Separate lu_site stats printing code into a separate function
2933 that can be reused on a client.
2937 Description: Constify instances of struct {lu,dt,md}_device_operations.
2938 Details : Constify instances of struct {lu,dt,md}_device_operations.
2942 Description: Introduce struct md_site and move meta-data specific parts of
2943 struct lu_site here.
2944 Details : Move md-specific fields out of struct lu_site into special struct
2945 md_site, so that lu_site can be used on a client.
2949 Description: Kill mdd_lov_destroy().
2950 Details : Remove unused mdd code.
2954 Description: Add st_block checking to multistat.c.
2955 Details : Add st_block checking to multistat.c.
2959 Description: Add lu_ref support to struct obd_device.
2960 Details : Add lu_ref tracking to obd_device.
2964 Description: Kill obd_set_fail_loc().
2965 Details : Remove unused code.
2969 Description: Add special type for ptlrpc_request interpret functions.
2970 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2971 there. Actual usage will be in CLIO.
2975 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2976 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2977 doesn't work with lockdep.
2981 Description: Add rwv.c test program.
2982 Details : New testing program exercising readv(2) and writev(2) (Qian).
2986 Description: Add sendfile.c test program.
2987 Details : New testing program exercising sendfile(2) (Jay).
2991 Description: Ratelimit a message that can be very frequent.
2992 Details : Ratelimit a memory allocation failure message that can
2997 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2998 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3002 Description: Liblustre build fixes.
3003 Details : Liblustre build fixes.
3007 Description: libcfs: add cfs_{need,cond}_resched() interface.
3008 Details : libcfs: add cfs_{need,cond}_resched() definition and
3009 implementations for Linux, NT, and liblustre.
3011 Severity : enhancement
3013 Description: More exported tunables for mballoc
3014 Details : Add support for tunable preallocation window and new tunables for
3015 large/small requests
3019 Description: Detect corruption of block bitmap and checking for preallocations
3020 Details : Checks validity of on-disk block bitmap. Also it does better
3021 checking of number of applied preallocations. When corruption is
3022 found, it turns filesystem readonly to prevent further corruptions.
3026 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3027 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3028 window to 0. If it does hit cache miss, and read-pattern is still
3029 stride-io mode, does not reset the stride window, but also does not
3030 increase the stride window length in this case.
3034 Frequency : only for big-endian servers
3035 Description: Check if system is big-endian while mounting fs with extents feature
3036 Details : Mounting a filesystem with extents feature will fail on big-endian
3037 systems since ext3-based ldiskfs is not supported on big-endian
3038 systems. This can be over-riden with "bigendian_extents" mount option.
3040 Severity : enhancement
3042 Description: The root squash functionality
3043 Details : A security feature, which is to prevent users from being able
3044 to mount lustre on their desktop, run as root, and delete
3045 all of the files in the filesystem. The goal is accomplished by
3046 remapping user id (UID) and group id (GID) of the root user to
3047 a UID and GID specified by the system administartor via Lustre
3048 configuration management server (MGS). The functionality also
3049 allows to specify sets of clients for which the remapping does
3054 Description: Excessive recovery window
3055 Details : With AT enabled, the recovery window can be excessively long (6000+
3056 seconds). To address this problem, we no longer use
3057 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3058 timeout no longer depends on the service time, it is set to
3059 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3060 time via pb_service_time.
3064 Description: Watchdog triggered on MDS failover
3065 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3066 the OSTs know that the MDS "UUID" can be reused for the same export
3067 from a different NID, so we do not need to wait for the export to be
3071 Frequency : rare, only if using MMP with Linux RAID
3073 Description: MMP doesn't work with Linux RAID
3074 Details : While using HA for Lustre servers with Linux RAID, it is possible
3075 that MMP will not detect multiple mounts. To make this work we
3076 need to unplug the device queue in RAID when the MMP block is being
3077 written. Also while reading the MMP block, we should read it from
3078 disk and not the cached one.
3080 Severity : enhancement
3082 Description: open file using fid
3083 Details : A file can be opened using just its fid, like
3084 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3087 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3090 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3091 Details : add proper locking for creation of jbd_4k slab cache
3095 Description: MMP check in ext3_remount() fails without displaying any error
3096 Details : When multiple mount protection fails during remount, proper error
3099 Severity : enhancement
3101 Description: Allow stripe size to be up to 4G-64k
3102 Details : Fix math logic to allow large stripe sizes.
3106 Description: add check for >8TB ldiskfs filesystems
3107 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3108 Don't allow >8TB ldiskfs filesystems to be mounted without
3109 force_over_8tb mount option
3114 Descriptoin: OSS complains "too many blocks" and crashes
3115 Details : In filter_commitrw_write():
3116 1. lquota_chkquota();
3117 2. obdo_from_inode();
3118 3. lquota_pending_commit();
3119 "2." has changed uid and gid in this case, then caused
3120 lquota_chkquota() and lquota_pending_commit() to use
3121 different uid and gid. That is the root of the bug.
3125 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3126 Details : Detect and print out the 'stripe count = -1' case properly.
3130 Description: Out or order replies might be lost on replay
3131 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3132 request with tid smaller than one currently being added, add it
3133 to the start, not end of the list.
3135 --------------------------------------------------------------------------------
3137 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3139 * Support for kernels:
3140 2.6.5-7.283 (SLES 9),
3141 2.6.9-55.EL (RHEL 4),
3142 2.6.16.46-0.14 (SLES 10),
3143 2.6.18.8 vanilla (kernel.org)
3144 * Client support for unpatched kernels:
3145 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3146 2.6.16 - 2.6.22 vanilla (kernel.org)
3147 * Due to recently discovered recovery problems, we do not recommend
3148 using patchless RHEL 4 clients with this or any earlier release.
3149 * Recommended e2fsprogs version: 1.39.cfs8
3150 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3151 * Starting with this release, the ldiskfs backing filesystem required
3152 by Lustre is now in its own package, lustre-ldiskfs. This package
3153 should be installed. It is versioned separately from Lustre and
3154 may be released separately in future.
3158 Description: block reactivating mgc import until all deactivates complete
3159 Details : Fix race when failing back MDT/MGS to itself (testing)
3162 Frequency : at statup only
3164 Description: mds_lov_synchronize race leads to various problems
3165 Details : simultaneous MDT->OST connections at startup can cause the
3166 sync to abort, leaving the OSC in a bad state.
3168 Severity : enhancement
3170 Description: add optional extra BUILD_VERSION info
3171 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3172 to override the lustre version.
3175 Frequency : 2.6.18 servers only
3177 Description: ll_kern_mount() doesn't release the module reference
3178 Details : The ldiskfs module reference count never drops down to 0
3179 because ll_kern_mount() doesn't release the module reference.
3184 Description: server LBUG when using old ost_num_threads parameter
3185 Details : Accept the old ost_num_threads parameter but warn that it
3186 is deprecated, and fix an off-by-one error that caused an LBUG.
3191 Description: Transient SCSI error results in persistent IO issue
3192 Details : iobuf->dr_error is not reinitialized to 0 between two
3196 Frequency : sometimes when underlying device returns I/O errors
3198 Description: OSTs not going read-only during write failures
3199 Details : OSTs are not remounted read-only when the journal commit threads
3200 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3201 instead of the ext3 wrappers.
3205 Description: poor connect scaling with increasing client count
3206 Details : Don't run filter_grant_sanity_check for more than 100 exports
3207 to improve scaling for large numbers of clients.
3210 Frequency : SLES10 only
3212 Description: sanity-quota.sh quotacheck failed: rc = -22
3213 Details : Quotas cannot be enabled on SLES10.
3216 Frequency : liblustre clients only
3218 Description: getdirentries does not give error when run on compute nodes
3219 Details : getdirentries does not fail when the size specified as an argument
3220 is too small to contain at least one entry
3222 Severity : enhancement
3224 Description: Add LNET router traceability for debug purposes
3225 Details : If a checksum failure occurs with a router as part of the
3226 IO path, the NID of the last router that forwarded the bulk data
3227 is printed so it can be identified.
3232 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3233 Details : Due to a race condition, liblustre clients were occasionally
3234 evicted incorrectly.
3236 Severity : enhancement
3238 Description: lfs setstripe use optional parameters instead of postional
3241 Severity : enhancement
3243 Description: Nanosecond timestamp support for ldiskfs
3244 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3245 resolution timestamps. There is not yet support for this at
3246 the Lustre filesystem level.
3249 Frequency : during server recovery
3251 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3252 Details : request with rq_no_resend flag not awake l_wait_event if they get
3256 Frequency : nfs export on patchless client
3258 Description: connectathon hang when test nfs export over patchless client
3259 Details : Disconnected dentry cannot be found with lookup, so we do not need
3260 to unhash it or make it invalid
3263 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3264 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3265 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3270 Description: open req refcounting wrong on reconnect
3271 Details : If reconnect happened between getting open reply from server and
3272 call to mdc_set_replay_data in ll_file_open, we will schedule
3273 replay for unreferenced request that we are about to free.
3274 Subsequent close will crash in variety of ways.
3275 Check that request is still eligible for replay in
3276 mdc_set_replay_data().
3281 Description: disable writes to filesystem when reading health_check file
3282 Details : the default for reading the health_check proc file has changed
3283 to NOT do a journal transaction and write to disk, because this
3284 can cause reads of the /proc file to hang and block HA state
3285 checking on a healthy but otherwise heavily loaded system. It
3286 is possible to return to the previous behaviour during configure
3287 with --enable-health-write.
3289 Severity : enhancement
3291 Description: 64-bit inode version
3292 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3293 to the inode. This will be required for version-based recovery.
3298 Description: MDS fails to start if a duplicate client export is detected
3299 Details : in some rare cases it was possible for a client to connect to
3300 an MDS multiple times. Upon recovery the MDS would detect this
3301 and fail during startup. Handle this more gracefully.
3303 Severity : enhancement
3305 Description: Add -o localflock option to simulate old noflock
3307 Details : This will achieve local-only flock/fcntl locks
3313 Description: log_commit_thread vs filter_destroy race leads to crash
3314 Details : Take import reference before releasing llog record semaphore
3319 Description: Wrong request locking in request set processing
3320 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3321 imp_delayed_list, in this place should be used imp_lock.
3324 Frequency : when reconnection
3326 Description: Grant Leak when osc reconnect to OST
3327 Details : When osc reconnect ost, OST(filter) should check whether it
3328 should grant more space to client by comparing fed_grant and
3329 cl_avail_grant, and return the granted space to client instead
3330 of "new granted" space, because client will call osc_init_grant
3331 to update the client grant space info.
3334 Frequency : when client reconnect to OST
3336 Description: Grant Leak when osc do resend and replay bulk write
3337 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3338 bulk write request, because the grant info while be sync between
3339 OSC and OST when reconnect, and we should ignore the grant info
3340 these of resend/replay write req.
3345 Description: Grant space more than available left space sometimes.
3346 Details : When then OST is about to be full, if two bulk writing from
3347 different clients came to OST. Accord the avaliable space of the
3348 OST, the first req should be permitted, and the second one
3349 should be denied by ENOSPC. But if the seconde arrived before
3350 the first one is committed. The OST might wrongly permit second
3351 writing, which will cause grant space > available space.
3354 Frequency : when client is evicted
3356 Description: Grant might be wrongly erased when osc is evicted by OST
3357 Details : when the import is evicted by server, it will fork another
3358 thread ptlrpc_invalidate_import_thread to invalidate the
3359 import, where the grant will be set to 0. While the original
3360 thread will update the grant it got when connecting. So if
3361 the former happened latter, the grant will be wrongly errased
3362 because of this race.
3367 Description: Checking Stale with correct fid
3368 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3369 to check whether it is stale, because sometimes, we want the
3370 enqueue happened anyway, and op_data.fid2 will not be initialized.
3372 Severity : enhancement
3374 Description: update patchless client
3375 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3378 Frequency : only with 2.4 kernel
3380 Description: random memory corruption
3381 Details : size of struct ll_inode_info is to big for union inode.u and this
3382 can be cause of random memory corruption.
3387 Description: Memory leak in recovery
3388 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3389 It should also check obd_fail before fsfilt_start, otherwise if
3390 fsfilt_start return -EROFS,(failover mds during mds recovery).
3391 then the req will return with repmsg->transno = 0 and rc = EROFS.
3392 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3393 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3394 be freed no matter whether fsfilt_commit success or not.
3397 Frequency : only with huge count clients
3399 Description: Prevents from taking the superblock lock in llap_from_page for
3401 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3402 for llap_from_page call in ll_removepage prevents from taking the
3403 superblock lock for a soon died page.
3408 Description: Not check open intent error before release open handle
3409 Details : in some rare cases, the open intent error is not checked before
3410 release open handle, which may cause
3411 ASSERTION(open_req->rq_transno != 0), because it tries to release
3412 the failed open handle.
3417 Description: Set cat log bitmap only after create log success.
3418 Details : in some rare cases, the cat log bitmap is set too early. and it
3419 should be set only after create log success.
3423 Description: Accessing a block bevice can re-enable I/O when Lustre is
3424 tearing down a device.
3425 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3430 Description: service threads may hog cpus when there are a lot of requests
3431 Details : Insert cond_resched to give other threads a chance to use some CPU
3436 Description: the cat log was not initialized in recovery
3437 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3438 unlink log on mds will not be initialized until mds post
3439 recovery. And also in mds post recovery, the unlink log will
3440 initialization will be done asynchronausly, so there will be race
3441 between add unlink log and unlink log initialization.
3445 Description: brw_stats were being printed incorrectly
3446 Details : brw_stats were being printed as log2 but all of them were not
3447 recorded as log2. Also remove some code duplication arising from
3448 filter_tally_{read,write}.
3452 Frequency : rare, only in recovery.
3453 Description: ASSERTION(req->rq_type != LI_POISON) failed
3454 Details : imp_lock should be held while iterating over imp_sending_list for
3455 prevent destroy request after get timeout in ptlrpc_queue_wait.
3459 Description: replay-single.sh test 52 fails
3460 Details : A lock's skiplist need to be cleanup when it being unlinked
3461 from its resource list.
3465 Description: Short directio read returns full requested size rather than
3467 Details : Direct I/O operations should return actual amount of bytes
3468 transferred rather than requested size.
3470 Severity : enhancement
3472 Description: metadata RPC reduction (e.g. for rm performance)
3473 Details : decrease the amount of synchronous RPC between clients and servers
3474 by canceling conflicing lock before the operation on the client
3475 and packing thier handles into the main operation RPC to server.
3477 Severity : enhancement
3479 Description: Async OSC create to avoid the blocking unnecessarily.
3480 Details : If an OST has no remaining object, system will block on the
3481 creation when it needs to create a new object on this OST. Now,
3482 ways use pre-created objects when available, instead of blocking on
3483 an empty osc while others are not empty. If we must block, we
3484 block for the shortest possible period of time.
3488 Description: improve handling recoverable errors
3489 Details : if request processig with error which can be recoverable on server
3490 request should be resend, otherwise page released from cache and
3493 Severity : enhancement
3495 Description: refine locking for avoid write wrong info into lov_objid file
3496 Details : fix possible races with add new target and write/update data in
3499 --------------------------------------------------------------------------------
3501 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3506 Frequency : on some architectures
3508 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3509 Details : Uninitialized flags sometimes cause configuration commands to
3513 Frequency : patchless clients only
3515 Description: missing __iget() symbol export
3516 Details : The __iget() symbol export is missing. To avoid the need for
3517 this on patchless clients the deathrow inode reaper is turned
3518 off, and we depend on the VM to clean up old inodes. This
3519 dependency was during via the fix for bug 12181.
3521 --------------------------------------------------------------------------------
3523 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3525 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3526 INTEROPERATE with older versions automatically. In many cases a
3527 special upgrade step is needed. Please read the
3528 user documentation before upgrading any part of a live system.
3529 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3530 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3531 * WARNING: Lustre configuration and startup changes are required with
3532 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3536 Severity : enhancement
3538 Description: MountConf
3539 Details : Lustre configuration is now managed via mkfs and mount
3540 commands instead of lmc and lconf. New obd types (MGS, MGC)
3541 are added for dynamic configuration management. See
3542 https://mail.clusterfs.com/wikis/lustre/MountConf for
3545 Severity : enhancement
3547 Description: dynamic OST addition
3548 Details : OSTs can now be added to a live filesystem
3550 Severity : enhancement
3552 Description: startup order invariance
3553 Details : MDTs and OSTs can be started in any order. Clients only
3554 require the MDT to complete startup.
3556 Severity : enhancement
3558 Description: parallel, asynchronous orphan cleanup
3559 Details : orphan cleanup is now performed in separate threads for each
3560 OST, allowing parallel non-blocking operation.
3562 Severity : enhancement
3564 Description: optimized stripe assignment
3565 Details : stripe assignments are now made based on ost space available,
3566 ost previous usage, and OSS previous usage, in order to try
3567 to optimize storage space and networking resources.
3569 Severity : enhancement
3571 Description: Permanently set tunables
3572 Details : All writable /proc/fs/lustre tunables can now be permanently
3573 set on a per-server basis, at mkfs time or on a live system.
3575 Severity : enhancement
3577 Description: Lustre message v2
3578 Details : Add lustre message format v2.
3580 Severity : enhancement
3582 Description: client OST exclusion list
3583 Details : Clients can be started with a list of OSTs that should be
3584 declared "inactive" for known non-responsive OSTs.
3588 Description: SPEC SFS validation failure on NFS v2 over lustre.
3589 Details : Changes the blocksize for regular files to be 2x RPC size,
3590 and not depend on stripe size.
3592 Severity : enhancement
3594 Description: Multiple MD RPCs in flight.
3595 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3596 To avoid overly-overloading MDS, introduce a limit on number of
3597 MDS RPCs in flight for a single client and add /proc controls
3598 to adjust this limit.
3600 Severity : enhancement
3602 Description: client read/write statistics
3603 Details : Add client read/write call usage stats for performance
3604 analysis of user processes.
3605 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3606 file access. extents_stats shows chunk size distribution.
3607 extents_stats_per_process show chunk size distribution per
3610 Severity : enhancement
3612 Description: mds statistics
3613 Details : Add detailed mds operations statistics in
3614 /proc/fs/lustre/mds/*/stats.
3618 Description: Failure of copying files with lustre special EAs.
3619 Details : Client side always return success for setxattr call for lustre
3620 special xattr (currently only "trusted.lov").
3625 Description: Refcount LNET uuids
3626 Details : The global LNET uuid list grew linearly with every startup;
3627 refcount repeated list entries instead of always adding to
3630 Severity : enhancement
3632 Description: Dynamic service threads
3633 Details : Within a small range, start extra service threads
3634 automatically when the request queue builds up.
3637 Frequency : mixed-endian client/server environments
3639 Description: mixed-endian crashes
3640 Details : The new msg_v2 system had some failures in mixed-endian
3643 Severity : enhancement
3645 Description: Easy OST removal
3646 Details : OSTs can be permanently deactivated with e.g. 'lctl
3647 conf_param lustre-OST0001.osc.active=0'
3649 Severity : enhancement
3651 Description: MGS proc entries
3652 Details : Added basic proc entries for the MGS showing what filesystems
3655 Severity : enhancement
3657 Description: provide MGS failover
3658 Details : Added config lock reacquisition after MGS server failover.
3660 Severity : enhancement
3662 Description: add Linux 2.4 support
3663 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3667 Description: a large application tries to do I/O to the same resource and dies
3668 in the middle of it.
3669 Details : Check the req->rq_arrival time after the call to
3670 ost_brw_lock_get(), but before we do anything about
3671 processing it & sending the BULK transfer request. This
3672 should help move old stale pending locks off the queue as
3673 quickly as obd_timeout.
3676 Frequency : when an incorrect nid is specified during startup
3678 Description: ptlrpc connect to non-existant node causes kernel crash
3679 Details : LNET can't be re-entered from an event callback, which
3680 happened when we expire a message after the export has been
3681 cleaned up. Instead, hand the zombie cleanup off to another
3684 Severity : enhancement
3686 Description: plain/inodebits lock performance improvement
3687 Details : Grouping plain/inodebits in granted list by their request modes
3688 and bits policy, thus improving the performance of search through
3692 Frequency : only if OST filesystem is corrupted
3694 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3695 Details : for a short time RPCs with bulk IO are in the replay list,
3696 but replay of bulk IOs is unimplemented. If the OST filesystem
3697 is corrupted due to disk cache incoherency and then replay is
3698 started it is possible to trip an assertion. Avoid putting
3699 committed RPCs into the replay list at all to avoid this issue.
3702 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3704 Description: System hang on startup
3705 Details : This bug allowed the liblustre (e.g. catamount) client to
3706 return to the app before handling all startup RPCs. This
3707 could leave the node unresponsive to lustre network traffic
3708 and manifested as a server ptllnd timeout.
3710 Severity : enhancement
3712 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3713 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3714 boolean to control whether to print peer debug info when a
3715 client's RPC times out.
3718 Frequency : only for kernels with patches from Lustre below 1.4.3
3720 Description: Remove old rdonly API
3721 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3724 Frequency : only for devices with external journals
3726 Description: Set external device read-only also
3727 Details : During a commanded failover stop, we set the disk device
3728 read-only while the server shuts down. We now also set any
3729 external journal device read-only at the same time.
3732 Frequency : when upgrading from 1.4 while trying to change parameters
3734 Description: The wrong (new) MDC name was used when setting parameters for
3735 upgraded MDT's. Also allows changing of OSC (and MDC)
3736 parameters if --writeconf is specified at tunefs upgrade time.
3739 Frequency : when setting specific ost indicies
3741 Description: QOS code breaks on skipped indicies
3742 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3743 created with --index need not be sequential.
3745 Severity : enhancement
3747 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3748 Details : The uninit_groups feature works in conjunction with the kernel
3749 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3750 up the pass1 processing of e2fsck. This is a read-only feature
3751 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3752 mount filesystems that have had this feature enabled.
3754 Severity : enhancement
3756 Description: Improve multi-block allocation algorithm to avoid fragmentation
3757 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3758 allocation locality and avoid filesystem fragmentation.
3760 ------------------------------------------------------------------------------
3762 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3764 * Support for kernels:
3765 2.4.21-47.0.1.EL (RHEL 3)
3766 2.6.5-7.283 (SLES 9)
3767 2.6.9-42.0.10.EL (RHEL 4)
3768 2.6.12.6 vanilla (kernel.org)
3769 2.6.16.27-0.9 (SLES 10)
3770 * Recommended e2fsprogs version: 1.39.cfs5
3772 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3776 Frequency : occasional, depends on client load and configuration
3777 Bugzilla : 12181, 12203
3778 Description: data loss for recently-modified files
3780 Details : In some cases it is possible that recently written or created
3781 files may not be written to disk in a timely manner (this should
3782 normally be within 30s unless client IO load is very high).
3783 The problem appears as zero-length files or files that are a
3784 multiple of 1MB in size after a client crash or client eviction
3785 that are missing data at the end of the file.
3787 This problem is more likely to be hit on clients where files are
3788 repeatedly created and unlinked in the same directory, clients
3789 have a large amount of RAM, have many CPUs, the filesystem has
3790 many OSTs, the clients are rebooted frequently, and/or the files
3791 are not accessed by other nodes after being written.
3793 The presence of the problem can be detected by looking at
3794 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3795 smaller than the second (nr_unused) then dirty files will not
3796 be flushed automatically to disk. "sync; sleep 10" should be
3797 run several times on the node before unmounting it to update
3798 Lustre (this is also safe to run on nodes without this problem).
3800 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3801 that can cause this same problem, so customers using that kernel
3802 also need to update the kernel in addition to Lustre. In order
3803 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3805 It is normal that files written just before a client crash (less
3806 than 30s) may not yet have been flushed to disk, even for local
3810 Frequency : frequent on thin XT3 nodes
3812 Description: UUID collision on thin XT3 Linux nodes
3813 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3814 randomly, since we relied on an insufficiently-seeded PRNG.
3819 Description: OSS hangs after "All ost request buffers busy"
3820 Details : A deadlock between quota and journal operations caused OSS
3821 hangs after printing "All ost request buffers busy."
3824 Frequency : always on liblustre builds
3826 Description: Cleanup compiler warnings on liblustre
3829 Frequency : always on liblustre builds on XT3
3831 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3835 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3836 > 32000 subdirectories
3837 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3838 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3839 subdirectory count crosses 32000. This will aid e2fsck to
3840 correctly handle more than 32000 subdirectories.
3843 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3845 Description: System hang on startup
3846 Details : This bug allowed the liblustre (e.g. catamount) client to
3847 return to the app before handling all startup RPCs. This
3848 could leave the node unresponsive to lustre network traffic
3849 and manifested as a server ptllnd timeout.
3851 Severity : enhancement
3853 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3854 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3855 boolean to control whether to print peer debug info when a
3856 client's RPC times out.
3861 Description: make O_SYNC working on 2.6 kernels
3862 Details : 2.6 kernels use different method for mark pages for write,
3863 so need add a code to lustre for O_SYNC work.
3868 Description: Failure to close file and release space on NFS
3869 Details : Put inode details into lock acquired in ll_intent_file_open.
3870 Use mdc_intent_lock in ll_intent_open to properly
3871 detect all kind of errors unhandled by mdc_enqueue.
3876 Description: proc file read during shutdown sometimes raced obd removal,
3878 Details : Add lock to prevent obd access after proc file removal.
3881 Frequency : Only for files larger than 4GB on 32-bit clients.
3883 Description: improperly doing page alignment of locks
3884 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3885 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3890 Description: RPCs being resent when they shouldn't be
3891 Details : Some RPCs that should not be resent are being resent. This
3892 can cause inconsistencies in the RPC state machine. Do not
3893 resend such requests.
3896 Frequency : rare, only with NFS export
3898 Description: Crash on NFS re-export node
3899 Details : under very unusual load conditions an assertion is hit in
3900 ll_intent_file_open()
3903 Frequency : only if OST filesystem is corrupted
3905 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3906 Details : for a short time RPCs with bulk IO are in the replay list,
3907 but replay of bulk IOs is unimplemented. If the OST filesystem
3908 is corrupted due to disk cache incoherency and then replay is
3909 started it is possible to trip an assertion. Avoid putting
3910 committed RPCs into the replay list at all to avoid this issue.
3915 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3916 Details : Large single O_DIRECT read and write calls can fail to allocate
3917 a sufficiently large buffer to process the request. In case of
3918 allocation failure the allocation is retried with a smaller
3919 buffer and broken into smaller requests.
3921 Severity : enhancement
3923 Description: Add -o localflock option to simulate old noflock behaviour.
3924 Details : This will achieve local-only flock/fcntl locks coherentness.
3929 Description: versioning check is incomplete
3930 Details : Checking the version difference of client vs. server, report
3931 error if the gap is too big.
3936 Description: add support PG_writeback bit
3937 Details : add support for PG_writeback bit for Lustre, for more carefull
3938 work with page cache in 2.6 kernel. This also fix some deadlocks
3939 and remove hack for work O_SYNC with 2.6 kernel.
3941 Severity : enhancement
3943 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3944 Details : The uninit_groups feature works in conjunction with the kernel
3945 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3946 up the pass1 processing of e2fsck. This is a read-only feature
3947 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3948 mount filesystems that have had this feature enabled.
3950 Severity : enhancement
3952 Description: Improve multi-block allocation algorithm to avoid fragmentation
3953 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3954 allocation locality and avoid filesystem fragmentation.
3956 ------------------------------------------------------------------------------
3958 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3960 * Support for kernels:
3961 2.6.9-42.0.3.EL (RHEL 4)
3962 2.6.5-7.276 (SLES 9)
3963 2.4.21-47.0.1.EL (RHEL 3)
3964 2.6.12.6 vanilla (kernel.org)
3965 2.6.16.21-0.8 (SLES10)
3966 * Recommended e2fsprogs version: 1.39.cfs2-0
3968 * The backwards-compatible /proc/sys/portals symlink has been removed
3969 in this release. Before upgrading, please ensure that you change
3970 any configuration scripts or /etc/sysctl.conf files that access
3971 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3972 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3973 in advance of the upgrade on any system running Lustre 1.4.6 or
3974 newer, since /proc/sys/lnet was added in that version.
3975 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3979 Frequency : only when quota is used
3981 Description: avoid scanning export list for quota master
3982 Details : Change the algorithms to avoid scanning export list in order
3983 to improve the efficiency.
3986 Frequency : MDS failover only, very rarely
3988 Description: "went back in time" messages on mds failover
3989 Details : The greatest transno may be lost when the current operation
3990 finishes with an error (transno==0) and the client's last_rcvd
3991 record is over-written. Save the greatest transno in the
3992 mds_last_transno for this case.
3995 Frequency : always for specific kernels and striping counts
3997 Description: client may get "Matching packet too big" without ACL support
3998 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3999 when trying to access files in certain configurations. The
4000 clients should in fact be denied when mounting because they do
4001 not understand ACLs.
4004 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4006 Description: many clients connecting with IO in progress causes connect timeouts
4007 Details : Avoid synchronous journal commits to avoid delays caused by many
4008 clients connecting/disconnecting when bulk IO is in progress.
4009 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4010 OST_IO_PORTAL to avoid delays behind potentially many pending
4014 Frequency : occasionally with multiple writers to a single file
4016 Description: shared writes to file may result in wrong size reported by stat()
4017 Details : Allow growing of kms when extent lock is cancelled
4020 Frequency : always with random mmap IO to multi-striped file
4022 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4023 Details : Only if the hole is at the end of OST object so that kms is too
4024 small. Fix is to increase kms accordingly in ll_nopage.
4027 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4029 Description: writes to a missing object would leak memory on the OST
4030 Details : If there is an inconsistency between the MDS and OST filesystems,
4031 such that the MDS references an object that doesn't exist, writes
4032 to that object will leak memory due to incorrect cleanup in the
4033 error handling path, eventually running out of memory on the OST.
4038 Description: Creating too long symlink causes lustre errors
4039 Details : Check symlink and name lengths before sending requests to MDS.
4042 Frequency : only if flock is enabled (not on by default)
4044 Description: posix locks not released on fd closure on 2.6.9+
4045 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4046 kernels, this caused such locks not to be released on fd close and
4047 then assertions on fs unmount about still used locks.
4050 Frequency : MDS failover only, very rarely
4052 Description: clients may get ASSERTION(granted_lock != NULL)
4053 Details : When request was taking a long time, and a client was resending
4054 a getattr by name lock request. The were multiple lock requests
4055 with the same client lock handle and
4056 mds_getattr_name->fixup_handle_for_resent_request found one of the
4057 lock handles but later failed with ASSERTION(granted_lock != NULL).
4062 Description: handle->h_buffer_credits > 0, assertion failure
4063 Details : h_buffer_credits is zero after truncate, causing assertion
4064 failure. This patch extends the transaction or creates a new
4068 Frequency : NFS re-export or patchless client
4069 Bugzilla : 11179, 10796
4070 Description: Crash on NFS re-export node (__d_move)
4071 Details : We do not want to hash the dentry if we don't have a lock.
4072 But if this dentry is later used in d_move, we'd hit uninitialised
4073 list head d_hash, so we just do this to init d_hash field but
4074 leave dentry unhashed.
4077 Frequency : NFS re-export or patchless client
4079 Description: NFS exports has problem with symbolic link
4080 Details : lustre client didn't properly install dentry when re-exported
4081 to NFS or running patchless client.
4084 Frequency : NFS re-export or patchless client
4086 Description: Various nfs/patchless fixes.
4087 Details : fixes reuse disconected alias for lookup process - this fixes
4088 warning "find_exported_dentry: npd != pd",
4089 fix permission error with open files at nfs.
4090 fix apply umask when do revalidate.
4093 Frequency : occasional
4095 Description: Crash on NFS re-export node
4096 Details : calling clear_page() on the wrong pointer triggered oops in
4097 generic_mapping_read().
4100 Frequency : rarely, using O_DIRECT IO
4102 Description: unaligned directio crashes client with LASSERT
4103 Details : check for unaligned buffers before trying any requests.
4106 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4108 Description: stale data returned from RAID cache
4109 Details : If only a small amount of IO is done to the RAID device before
4110 reading it again it is possible to get stale data from the RAID
4111 cache instead of reading it from disk.
4114 Frequency : always for sles10 kernel
4116 Description: sles10 support
4117 Details : ll_follow_link: compile fixes and using of nd_set_link
4118 under newer kernels.
4121 Frequency : depends on arch, kernel and compiler version, always on sles10
4124 Description: recursive or deep enough symlinks cause stack overflow
4125 Details : getting rid of large stack-allocated variable in
4129 Frequency : depends on hardware
4131 Description: lustre write performance loss in the SLES10 kernel
4132 Details : the performance loss is caused by using of write barriers in the
4133 ext3 code. The SLES10 kernel turns barrier support on by
4134 default. The fix is to undo that change for ldiskfs.
4136 ------------------------------------------------------------------------------
4138 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4140 * Support for kernels:
4141 2.6.9-42.0.3EL (RHEL 4)
4142 2.6.5-7.276 (SLES 9)
4143 2.4.21-47.0.1.EL (RHEL 3)
4144 2.6.12.6 vanilla (kernel.org)
4148 Frequency : quota enabled and large files being deleted
4150 Description: releasing more than 4GB of quota at once hangs OST
4151 Details : If a user deletes more than 4GB of files on a single OST it
4152 will cause the OST to spin in an infinite loop. Release
4153 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4158 Description: statfs data retrieved from /proc may be stale or zero
4159 Details : When reading per-device statfs data from /proc, in the
4160 {kbytes,files}_{total,free,avail} files, it may appear
4161 as zero or be out of date.
4164 Frequency : systems with MD RAID1 external journal devices
4166 Description: lconf's call to blkid is confused by RAID1 journal devices
4167 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4168 of returning all block devices that match the journal UUID.
4171 Frequency : always, for aggregate stripe size over 4GB
4173 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4174 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4175 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4178 Frequency : always if "lfs find" used on a local file/directory
4180 Description: "lfs find" segfaults if used on a local file/directory
4181 Details : The case where a directory component was not specified wasn't
4182 handled correctly. Handle this properly.
4185 Frequency : always on ppc64
4187 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4188 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4189 perform endianness conversion.
4192 Frequency : rarely (truncate to non-zero file size after write under load)
4193 Bugzilla : 10730, 10687
4194 Description: Files padded with zeros to next 4K multiple
4195 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4196 it is possible that files that are truncated to a non-zero size
4197 immediately after being written are filled with zero bytes beyond
4198 the truncated size. No file data is lost.
4200 Severity : enhancement
4202 Description: Allow recovery/failover for liblustre clients.
4203 Details : liblustre clients were unaware of failover configurations until
4206 Severity : enhancement
4208 Description: user file locks should fail when not mounting with flock option
4209 Details : Set up an error-returning stub in ll_file_operations.lock field
4210 to prevent incorrect behaviour when client is mounted without
4211 flock option. Also, set up properly f_op->flock field for
4215 Frequency : always on ia64
4217 Description: "lfs df" loops on printing out MDS statfs information
4218 Details : The obd_ioctl_data was not initialized and in some systems
4219 this caused a failure during the ioctl that did not return
4220 an error. Initialize the struct and return an error on failure.
4223 Frequency : SLES 9 only
4225 Description: Error of copying files with lustre special EAs as root
4226 Details : Client side always return success for setxattr call for lustre
4227 special xattr (currently only "trusted.lov").
4230 Frequency : rarely on clusters with both ia64+i386 clients
4232 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4233 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4234 clients are doing concurrent writes to the same file it is
4235 possible that the ia64 clients may LASSERT because the OST
4236 extent locks are not PAGE_SIZE aligned. Ensure that grown
4237 locks are always aligned on the request boundary.
4240 Frequency : specific use, occasional
4242 Description: Overwriting in use executable truncates on-disk binary image
4243 Details : If one node attempts to overwrite an executable in use by
4244 another node, we now correctly return ETXTBSY instead of
4245 truncating the file.
4247 Severity : enhancement
4249 Description: Async OSC create to avoid the blocking unnecessarily.
4250 Details : If an OST has no remaining object, system will block on the creating
4251 when need to create a new object on this OST. Now, ways use
4252 pre-created objects when available, instead of blocking on an
4253 empty osc while others are not empty. If we must block, we block
4254 for the shortest possible period of time.
4259 Description: chmod on Lustre root is propagated to other clients
4260 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4261 invalid by the follow_mount time.
4264 Frequency : liblustre clients only
4266 Description: Race in 'instant cancel' lock handling could lead to such locks
4267 never to be granted in case of SMP MDS
4268 Details : Do not destroy not yet granted but cbpending locks in
4272 Frequency : replay/resend of open
4274 Description: non null lock assetion failure in mds_intent_policy
4275 Details : Trying to replay/resend lockless open requests resulted in
4276 mds_open() returning 0 with no lock. Now it sets a flag if
4277 a lock is going to be returned.
4279 Severity : enhancement
4281 Description: Checksum enhancements
4282 Details : New checksum enhancements allow for resending RPCs that failed
4285 Severity : enhancement
4287 Description: Tunables on number of dirty pages in cacche
4288 Details : Allow to set limit on number of dirty pages cached.
4293 Description: client crash on unmount - lock still has references
4294 Details : In some error handling cases it was possible to leak a lock
4295 reference on a client while accessing a file. This was not
4296 harmful to the client during operation, but would cause the
4297 client to crash when the filesystem is unmounted.
4300 Frequency : specific case, rare
4302 Description: ETXTBSY on mds though file not in use by client
4303 Details : ETXTBSY is no longer incorrectly returned when attempting to
4304 chmod or chown a directory that the user previously tried to
4305 execute or a currently-executing binary.
4308 Frequency : extremely rare except on liblustre-based clients
4310 Description: Lustre space not freed when files are deleted
4311 Details : Clean up open-unlinked files after client eviction. Previously
4312 the unlink was skipped and the files remained as orphans.
4317 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4318 Details : In some cases it was possible to send a blocking callback to
4319 a client doing a glimpse, even though that client didn't get
4320 a lock granted. When the glimpse lock is cancelled on the OST
4321 the freed lock is left on the waiting list and corrupted the list.
4324 Frequency : all core dumps
4326 Description: Broke core dumps to lustre
4327 Details : Negative dentry may be unhashed if parent does not have UPDATE
4328 lock, but some callers, e.g. do_coredump, expect dentry to be
4329 hashed after successful create, hash it in ll_create_it.
4331 ------------------------------------------------------------------------------
4333 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4335 * Support for kernels:
4336 2.6.9-42.0.2.EL (RHEL 4)
4337 2.6.5-7.276 (SLES 9)
4338 2.4.21-40.EL (RHEL 3)
4339 2.6.12.6 vanilla (kernel.org)
4343 Frequency : always on RHEL 3
4345 Description: Number of open files grows over time
4346 Details : The number of open files grows over time, whether or not
4347 Lustre is started. This was due to a filp leak introduced
4348 by one of our kernel patches.
4350 ------------------------------------------------------------------------------
4352 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4354 * Support for kernels:
4355 2.6.9-42.EL (RHEL 4)
4356 2.6.5-7.276 (SLES 9)
4357 2.4.21-40.EL (RHEL 3)
4358 2.6.12.6 vanilla (kernel.org)
4363 Bugzilla : 5719, 9635, 9792, 9684
4364 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4365 Details : If a server is under heavy load and cannot reply to new
4366 connection requests before the client resends the (re)connect,
4367 the connection handling code can behave badly if two service
4368 threads are concurrently handing separate (re)connections from
4369 the same client. Add better locking to the connection handling
4370 code, and ensure that only a single connection will be processed
4371 for a given client UUID, even if the lock is dropped.
4373 Severity : enhancement
4375 Description: add TCP zero-copy support to kernel
4376 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4377 sends if the hardware supports scatter-gather and checksumming.
4378 This allows socklnd to do client-write and server-read more
4379 efficiently and reduce CPU utilization from skbuf copying.
4382 Frequency : only if NFS exporting from client
4384 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4385 Details : If setting attributes on a file created under NFS that had
4386 never been opened it would be possible to oops the client
4387 if the file had no objects.
4390 Frequency : always for liblustre
4392 Description: liblustre client does MDS+OSTs setattr RPC for each write
4393 Details : When doing a write from a liblustre client, the client
4394 incorrectly issued an RPC to the MDS and each OST the file was
4395 striped over in order to update the timestamps. When writing
4396 with small chunks and many clients this could overwhelm the MDS
4397 with RPCs. In all cases it would slow down the write because
4398 these RPCs are unnecessary.
4400 Severity : enhancement
4402 Description: allow number of MDS service threads to be changed at module load
4403 Details : It is now possible to change the number of MDS service threads
4404 running. Adding "options mds mds_num_threads={N}" to the MDS's
4405 /etc/modprobe.conf will set the number of threads for the next
4406 time Lustre is restarted (assuming the "mds" module is also
4407 reloaded at that time). The default number of threads will
4408 stay the same, 32 for most systems.
4413 Description: OST crash if filesystem is unformatted or corrupt
4414 Details : If an OST is started on a device that has never been formatted
4415 or if the filesystem is corrupt and cannot even mount then the
4416 error handling cleanup routines would dereference a NULL pointer.
4421 Description: NULL pointer deref in llap_from_page.
4422 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4423 incorrect) mapping if the page was truncated/reclaimed while it was
4424 searched for. Check for this condition and skip such pages when
4425 doing readahead. Introduce extra check to llap_from_page() to
4426 verify page->mapping->host is non-NULL (so page is not anonymous).
4429 Frequency : Sometimes when using sys_sendfile
4431 Description: "page not covered by a lock" warnings from ll_readpage
4432 Details : sendfile called ll_readpage without right page locks present.
4433 Now we introduced ll_file_sendfile that does necessary locking
4434 around call to generic_file_sendfile() much like we do in
4438 Frequency : with certain MDS communication failures at client mount time
4440 Description: NULL pointer deref after failed client mount
4441 Details : a client connection request may delayed by the network layer
4442 and not be sent until after the PTLRPC layer has timed out the
4443 request. If the client fails the mount immediately it will try
4444 to clean up before the network times out the request. Add a
4445 reference from the request import to the obd device and delay
4446 the cleanup until the network drops the request.
4449 Frequency : occasionally during client (re)connect
4451 Description: assertion failure during client (re)connect
4452 Details : processing a client connection request may be delayed by the
4453 client or server longer than the client connect timeout. This
4454 causes the client to resend the connection request. If the
4455 original connection request is replied in this interval, the
4456 client may trip an assertion failure in ptlrpc_connect_interpret()
4457 which thought it would be the only running connect process.
4460 Frequency : only with obd_echo servers and clients that are rebooted
4462 Description: kernel BUG accessing uninitialized data structure
4463 Details : When running an obd_echo server it did not start the ping_evictor
4464 thread, and when a client was evicted an uninitialized data
4465 structure was accessed. Start the ping_evictor in the RPC
4466 service startup instead of the OBD startup.
4468 Severity : enhancement
4469 Bugzilla : 10193 (patchless)
4470 Description: Remove dependency on various unexported kernel interfaces.
4471 Details : No longer need reparent_to_init, exit_mm, exit_files,
4472 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4475 Frequency : rare (only users of deprecated and unsupported LDAP config)
4477 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4478 Details : LDAP apparently contains 'lustreName' attributes instead of
4479 'name'. A simple remapping of the name is sufficient.
4482 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4484 Description: waiting_locks_callback trips kernel BUG if client is evicted
4485 Details : Running with the dump_on_timeout debug flag turned on makes
4486 it possible that the waiting_locks_callback() can try to dump
4487 the Lustre kernel debug logs from an interrupt handler. Defer
4488 this log dumping to the expired_lock_main() thread.
4490 Severity : enhancement
4492 Description: Support NFS exporting on 2.6 kernels.
4493 Details : Implement non-rawops metadata methods for NFS server to use without
4494 changing NFS server code.
4497 Frequency : very rare (synthetic metadata workload only)
4499 Description: two racing renames might cause an MDS thread to deadlock
4500 Details : Running the "racer" program may cause one MDS thread to rename
4501 a file from being the source of a rename to being the target of
4502 a rename at exactly the same time that another thread is doing
4503 so, and the second thread has already enqueued these locks after
4504 doing a lookup of the target and is trying to relock them in
4505 order. Ensure that we don't try to re-lock the same resource.
4508 Frequency : only very large systems with liblustre clients
4510 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4511 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4512 all OSTs in parallel. This allows the eviction of stale liblustre
4513 clients to proceed much faster than if they were done in series,
4514 and also offers similar improvements for other set_info RPCs.
4519 Description: excessive CPU usage during initial read phase on client
4520 Details : During the initial read phase on a client, it would agressively
4521 retry readahead on the file, consuming too much CPU and impacting
4522 performance (since 1.4.5.8). Improve the readahead algorithm
4523 to avoid this, and also improve some other common cases (read
4524 of small files in particular, where "small" is files smaller than
4525 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4530 Description: MDS crash when receiving packet with unknown intent.
4531 Details : Do not LBUG in unknown intent case, just return -EFAULT
4533 Severity : enhancement
4534 Bugzilla : 9293, 9385
4535 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4536 Details : Do not serialize getattr (non-intent version) and statfs.
4539 Frequency : occasional, when OST network is overloaded/intermittent
4541 Description: client evicted by OST after bulk IO timeout
4542 Details : If a client sends a bulk IO request (read or write) the OST
4543 may evict the client if it is unresposive to its data GET/PUT
4544 request. This is incorrect if the network is overloaded (takes
4545 too long to transfer the RPC data) or dropped the OST GET/PUT
4546 request. There is no need to evict the client at all, since
4547 the pinger and/or lock callbacks will handle this, and the
4548 client can restart the bulk request.
4551 Frequency : Always when mmapping file with no objects
4553 Description: client crashes when mmapping file with no objects
4554 Details : Check that we actually have objects in a file before doing any
4555 operations on objects in ll_vm_open, ll_vm_close and
4561 Description: Request leak when working with deleted CWD
4562 Details : Introduce advanced request refcount tracking for requests
4563 referenced from lustre intent.
4565 Severity : Enhancement
4567 Description: Cache open file handles on client.
4568 Details : MDS now will return special lock along with openhandle, if
4569 requested and client is allowed to hold openhandle, even if unused,
4570 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4571 closing files for every read/write openration.
4573 Severity : Enhancement
4575 Description: Cache open negative dentries on client when possible.
4576 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4577 negative dentries on lock revocation.
4582 Description: Remounting a client read-only wasn't possible with a zconf mount
4583 Details : It wasn't possible to remount a client read-only with llmount.
4585 Severity : enhancement
4586 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4587 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4588 MPICH is included to add Lustre-specific ADIO interfaces.
4589 This is based closely on the UFS ADIO layer and only differs
4590 in file creation, in order to allow the OST striping to be set.
4591 This is user-contributed code and not supported by CFS.
4596 Description: extended inode attributes (immutable, append-only) work improperly
4597 when 2.4 and 2.6 kernels are used on client/server or vice versa
4598 Details : Introduce kernel-independent values for these flags.
4600 Severity : enhancement
4603 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4604 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4605 granularity. Now, Fractional MB granularity is be supported,
4606 this is very useful for low memory system.
4608 Severity : enhancement
4610 Description: Getattr by fid
4611 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4612 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4613 have locks and avoids getattr rpc requests for it on every stat.
4616 Frequency : Always, for filesystems larger than 2TB
4618 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4619 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4620 filesystems larger than 2TB to be mounted with mballoc enabled.
4623 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4625 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4626 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4627 filesystems larger than 2TB is mounted then the kernel will
4628 silently corrupt the start of the filesystem. CONFIG_LBD is
4629 enabled for all CFS-supported kernels, but the possibility of
4630 this happening with a modified kernel config exists.
4632 Severity : enhancement
4634 Description: add client O_DIRECT support for 2.6 kernels
4635 Details : It is now possible to do O_DIRECT reads and writes to files
4636 in the Lustre client mountpoint on 2.6 kernel clients.
4638 Severity : enhancement
4640 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4641 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4642 parallel, not waiting for response from every OST before sending
4643 a rpc to the next OST.
4648 Description: setattr vs write race when updating file timestamps
4649 Details : Client processes that update a file timestamp into the past
4650 right after writing to the file (e.g. tar) it is possible that
4651 the updated file modification time can be reset to the current
4652 time due to a race between processing the setattr and write RPC.
4654 Severity : enhancement
4656 Description: Bring 'lfs find' closer in line with regular Linux find.
4657 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4658 -print0 options and obtains all the needed info through the lustre
4661 Severity : enhancement
4663 Description: support up to 1024 configured devices on one node
4664 Details : change obd_dev array from statically allocated to dynamically
4665 allocated structs as they are first used to reduce memory usage
4670 Description: Flush dirty partially truncated pages during truncate
4671 Details : Immediatelly flush partially truncated pages in filter_setattr,
4672 this way we completely avoid having any pages in page cache on OST
4673 and can retire ugly workarounds during writes to flush such pages.
4678 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4679 Details : protect inode from truncation within vfs_unlink() context
4680 just take a reference before calling vfs_unlink() and release it
4681 when parent's i_sem is free.
4686 Description: last_id value checked outside lock on OST caused LASSERT failure
4687 Details : If there were multiple MDS->OST object precreate requests in
4688 flight, it was possible that the OST's last object id was checked
4689 outside a lock and incorrectly tripped an assertion. Move checks
4690 inside locks, and discard old precreate requests.
4693 Frequency : always, if extents are used on OSTs
4695 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4696 Details : OSTs using the extents format would not zero the high 16 bits of
4697 the index physical block number. This is not a problem for any
4698 OST filesystems smaller than 16TB, and no kernels support ext3
4699 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4700 new/modified files) and can be fixed for existing filesystems
4701 with e2fsprogs-1.39-cfs1.
4706 Description: import connection selection may be incorrect if timer wraps
4707 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4708 import connections to be ignored if the 32-bit jiffies counter
4709 wraps. Use a 64-bit jiffies counter.
4712 Frequency : very large clusters immediately after boot
4714 Description: LNET request buffers exhausted under heavy short-term load
4715 Details : If a large number of client requests are generated on a service
4716 that has previously never seen so many requests it is possible
4717 that the request buffer growth cannot keep up with the spike in
4718 demand. Instead of dropping incoming requests, they are held in
4719 the LND until the RPC service can accept more requests.
4722 Frequency : Sometimes during replay
4724 Description: Assertion failure in ll_local_open after replay.
4725 Details : If replay happened on an open request reply before we were able
4726 to set replay handler, reply will become not swabbed tripping the
4727 assertion in ll_local_open. Now we set the handler right after
4728 recognising of open request
4731 Frequency : very rare
4733 Description: kernel reports "badness in vsnprintf"
4734 Details : Reading from the "recovery_status" /proc file in small chunks
4735 may cause a negative length in lprocfs_obd_rd_recovery_status()
4736 call to vsnprintf() (which is otherwise harmless). Exit early
4737 if there is no more space in the output buffer.
4739 Severity : enhancement
4741 Description: clear OBD RPC statistics by writing to them
4742 Details : It is now possible to clear the OBD RPC statistics by writing
4743 to the "stats" file.
4748 Description: Client mtime is not the same on different clients after utimes
4749 Details : In some cases, the client was using the utimes() syscall on
4750 a file cached on another node. The clients now validate the
4751 ctime from the MDS + OSTs to determine which one is right.
4756 Description: Inability to activate failout mode
4757 Details : lconf script incorrectly assumed that in pythong string's numeric
4758 value is used in comparisons.
4761 Frequency : always with multiple stripes per file
4763 Description: Inefficient object allocation for mutli-stripe files
4764 Details : When selecting which OSTs to stripe files over, for files with
4765 a stripe count that divides evenly into the number of OSTs,
4766 the MDS is always picking the same starting OST for each file.
4767 Return the OST selection heuristic to the original design.
4772 Description: mount failures may take full timeout to return an error
4773 Details : Under some heavy load conditions it is possible that a
4774 failed mount can wait for the full obd_timeout interval,
4775 possibly several minutes, before reporting an error.
4776 Instead return an error as soon as the status is known.
4778 Frequency : quota enabled and large files being deleted
4780 Description: releasing more than 4GB of quota at once hangs OST
4781 Details : If a user deletes more than 4GB of files on a single OST it
4782 will cause the OST to spin in an infinite loop. Release
4783 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4788 Description: statfs data retrieved from /proc may be stale or zero
4789 Details : When reading per-device statfs data from /proc, in the
4790 {kbytes,files}_{total,free,avail} files, it may appear
4791 as zero or be out of date.
4794 Frequency : systems with MD RAID1 external journal devices
4796 Description: lconf's call to blkid is confused by RAID1 journal devices
4797 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4798 of returning all block devices that match the journal UUID.
4801 Frequency : always, for aggregate stripe size over 4GB
4803 Description: assertion fails when trying to use 4GB stripe size
4804 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4805 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4808 Frequency : always on ppc64
4810 Description: the first write on an ext3 filesystem with mballoc got stuck
4811 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4812 perform endianness conversion.
4814 ------------------------------------------------------------------------------
4816 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4818 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4819 INTEROPERATE with older versions automatically. Please read the
4820 user documentation before upgrading any part of a live system.
4821 * WARNING: Lustre networking configuration changes are required with
4822 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4825 * Support for kernels:
4826 2.6.9-22.0.2.EL (RHEL 4)
4827 2.6.5-7.244 (SLES 9)
4828 2.6.12.6 vanilla (kernel.org)
4831 Severity : enhancement
4832 Bugzilla : 7981/8208
4833 Description: Introduced Lustre Networking (LNET)
4834 Details : LNET is new networking infrastructure for Lustre, it includes
4835 a reorganized network configuration mode (see the user
4836 documentation for full details) as well as support for routing
4837 between different network fabrics. Lustre Networking Devices
4838 (LNDs) for the supported network fabrics have also been
4839 created for this new infrastructure.
4841 Severity : enhancement
4842 Description: Introduced Access control lists
4843 Details : clients can set ACLs on files and directories in order to have
4844 more fine-grained permissions than the standard Unix UGO+RWX.
4845 The MDS must be started with the "-o acl" mount option.
4847 Severity : enhancement
4848 Description: Introduced filesystem quotas
4849 Details : Administrators may now establish per-user quotas on the
4852 Severity : enhancement
4854 Description: Configuration change for the XT3
4855 The PTLLND is now used to run Lustre over Portals on the XT3
4856 The configure option(s) --with-cray-portals are no longer used.
4857 Rather --with-portals=<path-to-portals-includes> is used to
4858 enable building on the XT3. In addition to enable XT3 specific
4859 features the option --enable-cray-xt3 must be used.
4864 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4865 Details : A race between allocating a new llcd and re-getting the llcd_lock
4866 allowed another thread to grab newly-allocated llcd.
4868 Severity : enhancement
4870 Description: 2.6 OST async journal commit and locking fix to improve performance
4871 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4872 2.6 kernels are now async as they already were in 2.4 kernels so
4873 that they can commit concurrently with the network bulk transfer.
4874 For block-allocated files the filter allocation semaphore is held
4875 to avoid filesystem fragmentation during allocation. BKL lock
4876 removed for 2.6 xattr operations where it is no longer needed.
4881 Description: lconf incorrectly determined whether two IP networks could talk
4882 Details : In some more complicated routing and multiple-network
4883 configurations, lconf will avoid trying to make a network
4884 connection to a disjoint part of the IP space. It was doing the
4885 math incorrectly for one set of cases.
4890 Description: Fix for potential infinite loop processing records in an llog.
4891 Details : If an llog record is corrupted/zeroed, it is possible to loop
4892 forever in llog_process(). Validate the llog record length
4893 and skip the remainder of the block on error.
4896 Frequency : occasional (liblustre only)
4898 Description: liblustre could not open files whose last component is a symlink
4899 Details : sysio_path_walk() would incorrectly pass the open intent to
4900 intermediate path components.
4903 Frequency : rare (liblustre only with non-standard tuning)
4904 Bugzilla : 7201 (7350)
4905 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4906 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4907 after it has been released, only do this for referenced locks
4909 Severity : enhancement
4911 Description: specifying an (invalid) directory default stripe_size of -1
4912 would reset the directory default striping
4913 Details : stripe_size -1 was used internally to signal directory stripe
4914 removal, now use "all default" to signal dir stripe removal
4915 as a directory striping of "all default" is not useful
4918 Frequency : common for large clusters running liblustre clients
4920 Description: doing an ls when liblustre clients are running is slow
4921 Details : sending a glimpse AST to a liblustre client waits for every AST
4922 to time out, as liblustre clients will not respond. Since they
4923 cannot cache data we refresh the OST lock LVB from disk instead.
4925 Severity : enhancement
4927 Description: doing an ls at the same time as file IO can be slow
4928 Details : enqueue and other "small" requests can be blocked behind many
4929 large IO requests. Create a new OST IO portal for non-IO
4930 requests so they can be processed faster.
4933 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4935 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4936 created file with mode 000
4937 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4938 allow an ftruncate() as a non-root user to a file with mode 000.
4939 Lustre now respects this flag to disable mode checks when
4940 truncating a file owned by the user
4943 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4946 Description: Revoking locks from clients that went dead or catatonic might take
4948 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4949 cancellation of such locks instant on servers without waiting for
4950 any reply from clients. Clients drops these locks when cancel
4951 notification from server is received without replying.
4954 Frequency : liblustre-only, when liblustre client dies or becomes busy
4956 Description: Doing ls on Linux clients can take a long time with active
4958 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4959 granting such locks to it in the first place if possible. Locks
4960 are taken by proxy on the OST during the read or write and
4961 dropped immediately afterward. Add connect flags handling, do
4962 not grant locks to liblustre clients for glimpse ASTs.
4964 Severity : enhancement
4966 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4967 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4968 sequential reads and behaves badly with applications doing random
4969 reads. Improve it by reading ahead at least read region, and
4970 avoiding excessive large RPC for small reads.
4972 Severity : enhancement
4974 Description: Creating more than 1000 files for a single job may cause a load
4975 imbalance on the OSTs if there are also a large number of OSTs.
4976 Details : qos_prep_create() uses an OST index reseed value that is an
4977 even multiple of the number of available OSTs so that if the
4978 reseed happens in the middle of the object allocation it will
4979 still utilize the OSTs as uniformly as possible.
4984 Description: OST or MDS may oops in ping_evictor_main()
4985 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4986 but doesn't restart at beginning of obd_exports_timed list
4989 Severity : enhancement
4991 Description: improve by-nid export eviction on the MDS and OST
4992 Details : allow multiple exports with the same NID to be evicted at one
4993 time without re-searching the exports list.
4996 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4998 Description: MDS may oops in groups_free()
4999 Details : in rare race conditions a newly allocated group_info struct is
5000 freed again, and this can be NULL. The 2.4 compatibility code
5001 for groups_free() checked for a NULL pointer, but 2.6 did not.
5004 Frequency : common for liblustre clients doing little filesystem IO
5005 Bugzilla : 9352, 7313
5006 Description: server may evict liblustre clients accessing contended locks
5007 Details : if a client is granted a lock or receives a completion AST
5008 with a blocking AST already set it would not reply to the AST
5009 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5012 Frequency : lfs setstripe, only systems with more than 160 OSTs
5014 Description: unable to set striping with a starting offset beyond OST 160
5015 Details : llapi_create_file() incorrectly limited the starting stripe
5016 index to the maximum single-file stripe count.
5019 Frequency : LDAP users only
5021 Description: lconf did not handle in-kernel recovery with LDAP properly
5022 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5024 Severity : enhancement
5026 Description: bind OST threads to NUMA nodes to improve performance
5027 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5028 node and do their allocations there to localize memory access
5030 Severity : enhancement
5032 Description: llmount can determine client NID directly from Myrinet (GM)
5033 Details : the client NID code from gmnalnid was moved directly into
5034 llmount, removing the need to use this or specifying the
5035 client NID explicitly when mounting GM clients with zeroconf
5038 Frequency : if client is started with down MDS
5040 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5041 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5042 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5043 ptlrpc_import_delayed_req() because we didn't check if the
5044 request was interrupted, and we also didn't break out of the
5045 event loop if there was no timeout
5050 Description: data loss during non-page-aligned writes to a single file from
5051 both multiple nodes and multiple threads on one node at same time
5052 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5053 inconsistency led to false short-reads, that were cached and later
5054 used by ->prepare_write() to fill in partially written page,
5055 leading to data loss.
5058 Frequency : always, if lconf --abort_recovery used
5060 Description: lconf --abort_recovery fails with 'Operation not supported'
5061 Details : lconf was attempting to abort recovery on the MDT device and not
5064 Severity : enhancement
5066 Description: remove cleanup logs
5067 Details : replace lconf-generated cleanup logs with lustre internal
5068 cleanup routines. Eliminates the need for client-cleanup and
5071 Severity : enhancement
5073 Description: add support for EAs (user and system) on lustre filesystems
5074 Details : it is now possible to store extended attributes in the Lustre
5075 client filesystem, and with the user_xattr mount option it
5076 is possible to allow users to store EAs on their files also
5078 Severity : enhancement
5080 Description: Add possibility (config option) to show minimal available OST free
5082 Details : When compiled with --enable-mindf configure option, statfs(2)
5083 (and so, df) will return least minimal free space available from
5084 all OSTs as amount of free space on FS, instead of summary of
5085 free spaces of all OSTs.
5087 Severity : enhancement
5089 Description: do not expand extent locks acquired on OST-side
5090 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5091 by server: they are not cached anyway.
5094 Frequency : when mmap is used/binaries executed from Lustre
5096 Description: Unmmap pages before throwing them away from read cache.
5097 Details : llap_shrink cache now attempts to unmap pages before discarding
5098 them (if unmapping failed - do not discard). SLES9 kernel has
5099 extra checks that trigger if this unmapping is not done first.
5104 Description: lconf didn't resolve symlinks before checking to see whether a
5105 given mountpoint was already in use
5108 Frequency : when migrating failover services
5109 Bugzilla : 6395, 9514
5110 Description: When migrating a subset of services from a node (e.g. failback
5111 from a failover service node) the remaining services would
5112 time out and evict clients.
5113 Details : lconf --force (implied by --failover) sets the global obd_timeout
5114 to 5 seconds in order to quickly disconnect, but this caused
5115 other RPCs to time out too quickly. Do not change the global
5116 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5118 Severity : enhancement
5119 Frequency : if MDS is started with down OST
5120 Bugzilla : 9439,5706
5121 Description: Allow startup/shutdown of an MDS without depending on the
5122 availability of the OSTs.
5123 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5124 Add appropriate locking and lov-osc refcounts for safe
5125 cleaning. Add osc abort_inflight calls in case the
5126 synchronize never started.
5129 Frequency : occasional (Cray XT3 only)
5131 Description: root not authorized to access files in CRAY_PORTALS environment
5132 Details : The client process capabilities were not honoured on the MDS in
5133 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5134 been accessed by an authorized user then root was able to access
5135 the file on the local client also. The root user capabilities
5136 are now allowed on the MDS, as this environment has secure UID.
5139 Frequency : occasional
5141 Description: ldiskfs "too long searching" message happens too often
5142 Details : A debugging message (otherwise harmless) prints too often on
5143 the OST console. This has been reduced to only happen when
5144 there are fragmentation problems on the filesystem.
5149 Description: Division by zero in statfs when all OSCs are inactive
5150 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5151 lov_statfs divides by value returned by lov_get_stripecnt().
5155 Bugzilla : 9489, 3273
5156 Description: First write from each client to each OST was only 4kB in size,
5157 to initialize client writeback cache, which caused sub-optimal
5158 RPCs and poor layout on disk for the first written file.
5159 Details : Clients now request an initial cache grant at (re)connect time
5160 and so that they can start streaming writes to the cache right
5161 away and always do full-sized RPCs if there is enough data.
5162 If the OST is rebooted the client also re-establishes its grant
5163 so that client cached writes will be honoured under the grant.
5168 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5169 Details : Now I/O RPCs go to different portal number and (presumably) fast
5170 lock requests (and glimses) and other RPCs get their own service
5171 threads pool that should be able to service those RPCs
5174 Severity : enhancement
5176 Description: Ability to exchange lustre version between client and servers and
5177 issue warnings at client side if client is too old. Also for
5178 liblustre clients there is ability to refuse connection of too old
5180 Details : New 'version' field is added to connect data structure that is
5181 filled with version info. That info is later checked by server and
5185 Frequency : rare, liblustre only.
5186 Bugzilla : 9296, 9581
5187 Description: Two simultaneous writes from liblustre at offset within same page
5188 might proceed at the same time overwriting eachother with stale
5190 Details : I/O lock withing llu_file_prwv was released too early, before data
5191 actually was hitting the wire. Extended lock-holding time until
5192 server acknowledges receiving data.
5195 Frequency : extremely rare. Never observed in practice.
5197 Description: avoid generating lustre_handle cookie of 0.
5198 Details : class_handle_hash() generates handle cookies by incrementing
5199 global counter, and can hit 0 occasionaly (this is unlikely, but
5200 not impossible, because initial value of cookie counter is
5201 selected randonly). Value of 0 is used as a sentinel meaning
5202 "unassigned handle" --- avoid it. Also coalesce two critical
5203 sections in this function into one.
5205 Severity : enhancement
5207 Description: allow liblustre clients to delegate truncate locking to OST
5208 Details : To avoid overhead of locking, liblustre client instructs OST to
5209 take extent lock in ost_punch() on client's behalf. New connection
5210 flag is added to handle backward compatibility.
5212 Severity : enhancement
5213 Bugzilla : 4928, 7341, 9758
5214 Description: allow number of OST service threads to be specified
5215 Details : a module parameter allows the number of OST service threads
5216 to be specified via "options ost ost_num_threads={N}" in the
5217 OSS's /etc/modules.conf or /etc/modprobe.conf.
5221 Bugzilla : 6146, 9635, 9895
5222 Description: servers crash with bad pointer in target_handle_connect()
5223 Details : In rare cases when a client is reconnecting it was possible that
5224 the connection request was the last reference for that export.
5225 We would temporarily drop the export reference and get a new
5226 one, but this may have been the last reference and the export
5227 was just destroyed. Get new reference before dropping old one.
5229 Severity : enhancement
5230 Frequency : if client is started with failover MDS
5232 Description: Allow multiple MDS hostnames in the mount command
5233 Details : Try to read the configuration from all specified MDS
5234 hostnames during a client mount in case the "primary"
5237 Severity : enhancement
5239 Description: Stop sending data to evicted clients as soon as possible.
5240 Details : Check if the client we are about to send or are sending data to
5241 was evicted already. (Check is done every second of waiting,
5242 for which l_wait_event interface was extended to allow checking
5243 of exit condition at specified intervals).
5246 Frequency : rare, normally only when NFS exporting is done from client
5248 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5250 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5253 Frequency : always, if srand() is called before liblustre initialization
5255 Description: Liblustre uses system PRNG disturbing its usage by user application
5256 Details : Introduce internal to lustre fast and high-quality PRNG for
5257 lustre usage and make liblustre and some other places in generic
5258 lustre code to use it.
5260 Severity : enhancement
5261 Bugzilla : 9477, 9557, 9870
5262 Description: Verify that the MDS configuration logs are updated when xml is
5263 Details : Check if the .xml configuration logs are newer than the config
5264 logs stored on the MDS and report an error if this is the case.
5265 Request --write-conf, or allow starting with --old_conf.
5267 Severity : enhancement
5269 Description: Handle symlinks in the path when checking if Lustre is mounted.
5270 Details : Resolve intermediate symlinks when checking if a client has
5271 mounted a filesystem to avoid duplicate client mounts.
5276 Description: lconf can hit an error exception but still return success.
5277 Details : The lconf command catches the Command error exception at the top
5278 level script context and will exit with the associated exit
5279 status, but doesn't ensure that this exit status is non-zero.
5284 Description: failure of ptlrpc thread startup can cause oops
5285 Details : Starting a ptlrpc service thread can fail if there are a large
5286 number of threads or the server memory is very fragmented.
5287 Handle this without oopsing.
5290 Frequency : always, only if liblustre and non-default acceptor port was used
5292 Description: liblustre cannot connect to servers with non-default acceptor port
5293 Details : tcpnal_set_default_params() was not called and was therefore
5294 ignoring the environment varaible TCPNAL_PORT, as well as other
5295 TCPNAL_ environment variables
5300 Description: two objects could be created on the same OST for a single file
5301 Details : If an OST is down, in some cases it was possible to create two
5302 objects on a single OST for a single file. No problems other
5303 than potential performance impact and spurious error messages.
5307 Bugzilla : 5681, 9562
5308 Description: Client may oops in ll_unhash_aliases
5309 Details : Client dcache may become inconsistent in race condition.
5310 In some cases "getcwd" can fail if the current directory is
5316 Description: Inode refcounting problems in NFS export code
5317 Details : link_raw functions used to call d_instantiate without obtaining
5318 extra inode reference first.
5322 Bugzilla : 9942, 9903
5323 Description: Referencing freed requests leading to crash, memleaks with NFS.
5324 Details : We used to require that call to ll_revalidate_it was always
5325 followed by ll_lookup_it. Also with revalidate_special() it is
5326 possible to call ll_revalidate_it() twice for the same dentry
5327 even if first occurence returned success. This fix changes semantic
5328 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5329 reference on a request referred from the intent.
5330 ll_intent_release() then releases such a request.
5333 Frequency : rare, normally benchmark loads only
5335 Description: unlinked inodes were kept in memory on the client
5336 Details : If a client is repeatedly creating and unlinking files it
5337 can accumulate a lot of stale inodes in the inode slab cache.
5338 If there is no other client load running this can cause the
5339 client node to run out of memory. Instead flush old inodes
5340 from client cache that have the same inode number as a new inode.
5343 Frequency : SLES9 2.6.5 kernel and long filenames only
5344 Bugzilla : 9969, 10379
5345 Description: utime reports stale NFS file handle
5346 Details : SLES9 uses out-of-dentry names in some cases, which confused
5347 the lustre dentry revalidation. Change it to always use the
5351 Frequency : rare, unless heavy write-truncate concurrency is continuous
5352 Bugzilla : 4180, 6984, 7171, 9963, 9331
5353 Description: OST becomes very slow and/or deadlocked during object unlink
5354 Details : filter_destroy() was holding onto the parent directory lock
5355 while truncating+unlinking objects. For very large objects this
5356 may block other threads for a long time and slow overall OST
5357 responsiveness. It may also be possible to get a lock ordering
5358 deadlock in this case, or run out of journal credits because of
5359 the combined truncate+unlink. Solution is to do object truncate
5360 first in one transaction without parent lock, and then do the
5361 final unlink in a new transaction with the parent lock. This
5362 reduces the lock hold time dramatically.
5365 Frequency : rare, 2.4 kernels only
5367 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5368 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5369 before unmounting their filesystem in order to avoid deadlock.
5370 The kernel_locked() function in 2.4 kernels only checks whether
5371 the kernel lock is held, not whether it is this process that is
5372 holding it as 2.6 kernels do.
5377 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5378 Details : The client ptlrpc code may be trying to reconnect to a down
5379 server before a previous connection attempt has timed out.
5380 Increase the reconnect interval to be longer than the connection
5381 timeout interval to avoid sending duplicate connections to
5385 Frequency : echo_client brw_test command
5387 Description: fix echo_client to work with OST preallocated code
5388 Details : OST preallocation code (5137) didn't take echo_client IO path
5389 into account: echo_client calls filter methods outside of any
5390 OST thread and, hence, there is no per-thread preallocated
5391 pages and buffers to use. Solution: hijack pga pages for IO. As
5392 a byproduct, this avoids unnecessary data copying.
5396 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5397 Description: Client can oops in mdc_commit_close() after open replay
5398 Details : It was possible for the MDS to return an open request with no
5399 transaction number in mds_finish_transno() if the client was
5400 evicted, but without actually returning an error. Clients
5401 would later try to replay that open and may trip an assertion
5402 Simplify the client close codepath, and always return an error
5403 from the MDS in case the open is not successful.
5406 Frequency : rare, 2.6 OSTs only
5408 Description: OST may deadlock under high load on fragmented files
5409 Details : If there was a heavy load and highly-fragmented OST filesystems
5410 it was possible to have all the OST threads deadlock waiting on
5411 allocation of biovecs, because the biovecs were not released
5412 until the entire RPC IO was completed. Instead, release biovecs
5413 as soon as they are complete to ensure forward IO progress.
5415 Severity : enhancement
5417 Description: Support for specifying external journal device at mount
5418 Details : If an OST or MDS device is formatted with an external journal
5419 device, this device major/minor is stored in the ext3 superblock
5420 and may not be valid for failover. Allow detecting and
5421 specifying the external journal at mount time.
5426 Description: Mounting an MDS with pending unlinked files may cause oops
5427 Details : target_finish_recovery() calls mds_postrecov() which returned
5428 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5429 considers this an error and immediately begins cleaning up the
5430 lov, just after starting the mds_lov process
5432 Severity : enhancement
5434 Description: Implement 'lfs df' to report actual free space on per-OST basis
5435 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5436 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5437 usage of objects; '-h' to report in human readable format.
5439 ------------------------------------------------------------------------------
5441 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5448 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5449 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5450 at startup during orphan unlinking. Instead, skip buddy update
5451 before setup, as it will be regenerated after recovery is complete.
5456 Description: If an OST is inactive, its locks might reference stale inodes.
5457 Details : lov_change_cbdata() must iterate over all namespaces, even if
5458 they are inactive to clear inode references from the lock.
5460 Severity : enhancement
5461 Frequency : occasional, if non-standard max_dirty_mb used
5463 Description: Client will block write RPCs if not enough grant
5464 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5465 then the client will block writes while waiting for another RPC
5466 to complete instead of consuming its dirty limit. With change
5467 we get improved performance when max_dirty_mb is small.
5469 Severity : enhancement
5470 Bugzilla : 3389, 6253
5471 Description: Add support for supplementary groups on the MDS.
5472 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5473 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5474 lookups for user supplementary groups into a cache.
5478 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5479 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5480 for liblustre O_EXCL is enforced by MDS.
5484 Description: readdir never returns NULL in liblustre.
5485 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5486 dentry in struct dirent.
5490 Frequency : liblustre only, depends on application IO pattern
5491 Description: liblustre clients evicted if not contacting servers
5492 Details : Don't put liblustre clients into the ping_evictor list, so
5493 they will not be evicted by the pinger ever.
5495 Severity : enhancement
5497 Description: Add ability to evict clients by NID from MDS.
5498 Details : By echoing "nid:$NID" string into
5499 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5500 $NID would be instantly evicted from this MDS and from all active
5501 OSTs connected to it.
5505 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5506 Details : lookup_it_finish() used to query file size from OSTs that was not
5511 Description: service threads change working directory to that of init
5512 Details : Starting lustre service threads may pin the working directory
5513 of the parent thread, making that filesystem busy. Threads
5514 now change to the working directory of init to avoid this.
5518 Frequency : during shutdown only
5519 Description: shutdown with a failed MDS or OST can cause unmount to hang
5520 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5525 Frequency : 2.6 only
5526 Description: chmod/chown may include an extra supplementary group
5527 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5528 supplementary group and if none is specified this is used.
5531 Bugzilla : 5479 (6816)
5533 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5534 Details : If lookup is in progress on a file that is unlinked we might try
5535 to revalidate the inode and fail in revalidate after lookup is
5536 complete and ll_file_open() enqueues the open again but
5537 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5540 Frequency : always, if lconf --abort_recovery used
5542 Description: lconf --abort_recovery fails with 'Operation not supported'
5543 Details : lconf was attempting to abort recovery on the MDT device and not
5546 ------------------------------------------------------------------------------
5548 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5553 Frequency : rare (only unsupported configurations with a node running as an
5555 Bugzilla : 6514, 5137
5556 Description: Mounting a Lustre file system on a node running as an OST could
5558 Details : OSTs now preallocates memory needed to write out data at
5559 startup, instead of when needed, to avoid having to
5560 allocate memory in possibly low memory situations.
5561 Specifically, if the file system is mounted on on OST,
5562 memory pressure could force it to try to write out data,
5563 which it needed to allocate memory to do. Due to the low
5564 memory, it would be unable to do so and the node would
5565 become unresponsive.
5567 Severity : enhancement
5569 Description: Addition of lconf --service command line option
5570 Details : lconf now accepts a '--service <arg>' option, which is
5571 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5573 Severity : enhancement
5575 Description: Failover mode is now the default for OSTs.
5576 Details : By default, OSTs will now run in failover mode. To return to
5577 the old behaviour, add '--failout' to the lmc line for OSTs.
5579 Severity : enhancement
5581 Description: Health checks are now provided for MDS and OSTs
5582 Details : Additional detailed health check information on MSD and OSTs
5583 is now provided through the procfs health_check value.
5586 Frequency : occasional, depends on IO load
5588 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5589 after numerous create/delete cycles
5590 Details : The ext3 inode allocation policy would not allocate new inodes
5591 very well on the OSTs because there are no new directories
5592 being created. Instead we look for groups with free space if
5593 the parent directories are nearly full.
5598 Description: Network or server problems during mount may cause partially
5599 mounted clients instead of returning an error.
5600 Details : The config llog parsing code may overwrite the error return
5601 code during mount error handling, returning success instead
5607 Description: MDS can fail to allocate large reply buffers
5608 Details : After long uptimes the MDS can fail to allocate large reply
5609 buffers (e.g. zconf client mount config records) due to memory
5610 fragmentation or consumption by the buffer cache. Preallocate
5611 some large reply buffers so that these replies can be sent even
5612 under memory pressure.
5616 Frequency : rare (liblustre)
5617 Description: fsx running with liblustre complained that using truncate() to
5618 extend the file doesn't work. This patch corrects that issue.
5619 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5620 ATTR_SIZE and lsm use in llu_setattr_raw.
5624 Frequency : rare, only 2.6 kernels
5625 Description: Unusual file access patterns on the MDS may result in inode
5626 data being lost in very rare circumstances.
5627 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5628 code in the 2.6 kernel caused the inode and/or EA data not to
5629 be read from disk, causing single-file corruption.
5633 Frequency : rare, only 2.6 filesystems using extents
5634 Description: Heavy concurrent write and delete load may cause data corruption.
5635 Details : It was possible under high-load situations to have an extent
5636 metadata block in the block device cache from a just-unlinked
5637 file overwrite a newly-allocated data block. We now unmap any
5638 metadata buffers that alias just-allocated data blocks.
5642 Frequency : filesystems with default stripe_count larger than 77
5643 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5644 Details : lconf specifies an inode size of 4096 bytes when the default
5645 stripe_count is larger than 77. This conflicts with the default
5646 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5647 this case to avoid pinning too much memory for large EAs.
5649 ------------------------------------------------------------------------------
5651 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5656 Frequency : rare (extremely heavy IO load with hundreds of clients)
5658 Description: Client is evicted, gets IO error writing to file
5659 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5660 caused clients to be evicted for AST timeouts. The fixes in
5661 bug 5192 mean we no longer need to have such short AST timeouts
5662 so ldlm_timeout has been increased.
5665 Frequency : occasional during --force or --failover shutdown under load
5666 Bugzilla : 5949, 4834
5667 Description: Server oops/LBUG if stopped with --force or --failover under load
5668 Details : a collection of import/export refcount and cleanup ordering
5669 issues fixed for safer force cleanup
5672 Frequency : only filesystems larger than 120 OSTs
5673 Bugzilla : 5990, 6223
5674 Description: lfs getstripe would oops on a very large filesystem
5675 Details : lov_getconfig used kfree on vmalloc'd memory
5678 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5679 Bugzilla : 6242, 6243
5680 Description: reading from files that had been truncated to a non-zero size
5681 but never opened returned no data
5682 Details : ll_file_read() reads zeros from no-object files to EOF
5687 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5688 Details : There's little harm in aborting MDS/OSS recovery and letting it
5689 try again, so I removed the LASSERT and return an error instead.
5691 Severity : enhancement
5693 Description: New debugging infrastructure for tracking down data corruption
5694 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5695 (b) cover more of the client-side code path, and (c) try to narrow
5696 down where problems occurred
5700 Bugzilla : 3819, 4364, 4397, 6313
5701 Description: Racing close and eviction MDS could cause assertion in mds_close
5702 Details : It was possible to get multiple mfd references during close and
5703 client eviction, leading to one thread referencing a freed mfd.
5705 Severity: : enhancement
5706 Bugzilla : 3262, 6359
5707 Description: Attempts to reconnect to servers are now more aggressive.
5708 Details : This builds on the enhanced upcall-less recovery that was added
5709 in 1.4.2. When trying to reconnect to servers, clients will
5710 now try each server in the failover group every 10 seconds. By
5711 default, clients would previously try one server every 25 seconds.
5716 Description: After recovery, certain operations trigger a failed
5717 assertion on a client.
5718 Details : Failing over an mds, using lconf -d --failover, while a
5719 client was doing a readdir() call would cause the client to
5720 LBUG after recovery completed and the readdir() was resent.
5722 Severity : enhancement
5724 Description: Default groups are now added by lconf
5725 Details : You can now run lconf --group <servicename> without having to
5726 manually add groups with lmc.
5729 Frequency : occasional
5731 Description: Nodes with an elan id of 0 trigger a failed assertion
5734 Frequency : always when accessing e.g. tty/console device nodes
5736 Description: tty and some other devices nodes cannot be used on lustre
5737 Details : file's private_data field is used by device data and lustre
5738 values in there got lost. New field was added to struct file to
5739 store fs-specific private data.
5742 Frequency : when exporting Lustre via NFS
5744 Description: NFSD failed occasionally when looking up a path component
5745 Details : NFSD is looking up ".." which was broken in ext3 directories
5746 that had grown large enough to become hashed.
5749 Frequency : Clusters with multiple interfaces not on the same subnet
5751 Description: Nodes will repeatedly try to reconnect to an interface which it
5752 cannot reach and report an error to the log.
5753 Details : Extra peer list entries will be created by lconf with some peers
5754 unreachable. lconf now validates the peer before adding it.
5757 Frequency : Only if a default stripe is set on the filesystem root.
5759 Description: Setting a default stripe on the filesystem root prevented the
5760 filesystem from being remounted.
5761 Details : The client was sending extra request flags in the root getattr
5762 request and did not allocate a reply buffer for the dir EA.
5765 Frequency : occasional, higher if lots of files are accessed by one client
5766 Bugzilla : 6159, 6097
5767 Description: Client trips assertion regarding lsm mismatch/magic
5768 Details : While revalidating inodes the VFS looks up inodes with ifind()
5769 and in rare cases can find an inode that is being freed.
5770 The ll_test_inode() code will free the lsm during ifind()
5771 when it finds an existing inode and then the VFS later attaches
5772 this free lsm to a new inode.
5776 Bugzilla : 6422, 7030
5777 Description: MDS deadlock between mkdir and client eviction
5778 Details : Creating a new file via mkdir or mknod (starting a transaction
5779 and getting the ns lock) can deadlock with client eviction
5780 (gets ns lock and trying to finish a synchronous transaction).
5783 Frequency : occasional
5784 Description: While starting a server, the fsfilt_ext3 module could not be
5786 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5787 kernels. Previously, lconf would still use the ext3 name
5788 when trying to load modules. Now, it will correctly use
5789 ext3 on 2.4 and ldiskfs on 2.6.
5791 Severity : enhancement
5792 Description: The default stripe count has been changed to 1
5793 Details : The interpretation of the default stripe count (0, to lfs
5794 or lmc) has been changed to mean striping across a single
5795 OST, rather than all available. For general usage we have
5796 found a stripe count of 1 or 2 works best.
5798 Severity : enhancement
5799 Description: Add support for compiling against Cray portals.
5800 Details : Conditional compiling for some areas that are different
5804 Frequency : occasional
5805 Bugzilla : 6409, 6834
5806 Description: Creating files with an explicit stripe count may lead to
5807 a failed assertion on the MDS
5808 Details : If some OSTs are full or unavailable, creating files may
5809 trigger a failed assertion on the MDS. Now, Lustre will
5810 try to use other servers or return an error to the
5814 Frequency : occasional
5816 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5817 caused lock timeout during readahead (since 1.4.2).
5818 Details : Processes doing readahead might match a lock that hasn't been
5819 granted yet if there are overlapping and conflicting lock
5820 requests. The readahead process waits on ungranted lock
5821 (original lock is CBPENDING), while OST waits for that process
5822 to cancel CBPENDING read lock and eventually evicts client.
5824 Severity : enhancement
5826 Description: Initial enabling of flock support for clients
5827 Details : Implements fcntl advisory locking and file status functions.
5828 This feature is provided as an optional mount flag (default
5829 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5830 locking are implemented yet, and those that are are not guaranteed
5831 to be completely correct in production environments.
5832 mount -t lustre -o [flock|noflock] ...
5835 Frequency : occasional
5837 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5838 assertion in the ext3 JBD (journaling) layer.
5839 Details : The b_committed_data struct is protected by the big kernel lock
5840 in 2.4 kernels, serializing journal_commit_transaction() and
5841 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5842 access to this struct. In 2.6 kernels there is finer grained
5843 locking to improve SMP performance of the JBD layer.
5847 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5852 Description: merge and cleanup kernel patches.
5855 -----------------------------------------------------------------------------
5857 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5859 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5860 versions of Lustre. Please update all servers and clients to
5861 version 1.4.2 or later at the same time. You must also run
5862 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5863 to update the configuration logs.
5865 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5866 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5867 - swab llog records of type '0' so we get proper header size/idx (5861)
5868 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5869 - fix rename of one directory over another leaking an inode (5953)
5870 - avoid SetPageDirty on 2.6 (5981)
5871 - don't re-add just-being-destroyed locks to the waiting list (5653)
5872 - when creating new directories, inherit the parent's custom
5873 striping settings if present parent (3048)
5874 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5875 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5876 - don't hold client locks on temporary worklist from l_lru (5666)
5877 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5878 - automatically evict dead clients (5921)
5879 - Update file size properly in create+truncate+fstat case (6196)
5880 - Do not unhash mountpoint dentries, do not allow removal of
5882 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5883 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5884 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5885 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5886 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5887 - Correctly handle reads of files with no objects (6243)
5888 - lctl recover will also mark a device active if deactivate used (5933)
5890 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5891 - allow --write-conf on an MDS with different nettype than client (5619)
5892 - don't write config llogs to MDS for mounts not from that MDS (5617)
5893 - lconf should create multiple TCP connections from a client (5201)
5894 - init scripts are now turned off by default; run chkconfig --on
5895 lustre and chkconfig --on lustrefs to use them
5896 - upcalls are no longer needed for clients to recover to failover
5898 - add --abort-recovery option to lconf to abort recovery on device
5900 - add support for an arbitrary number of OSTs (3026)
5901 - Quota support protocol changes.
5902 - forward compatibility changes to wire structs (6007)
5903 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5904 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5905 - improved "lustre status" script
5906 - initialize blocksize for non-regular files (6062)
5907 - added --disable-server and --disable-client configure options (5782)
5908 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5909 - Vanilla 2.4.29 support
5910 - increase maximum number of obd devices to 520 (6242)
5911 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5912 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5913 SLES 9 2.6.5 kernels
5914 - limit stripes per file to 160 (the maximum EA size) (6093)
5916 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5919 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5920 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5921 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5922 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5923 - small ext3 extents cleanups and fixes (5733)
5924 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5925 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5926 - avoid needless client->OST connect, fix handle mismatch (5317)
5927 - fix DLM error path that led to out-of-sync client, long delays (5779)
5928 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5929 - fix several locking issues related to i_size (5492,5624,5654,5672)
5930 - don't move pending lock onto export if it is already evicted (5683)
5931 - fix kernel oops when creating .foo in unlinked directory (5548)
5932 - fix deadlock in obdfilter statistics vs. object create (5811)
5933 - use time_{before,after} to avoid timer jiffies wrap (5882)
5934 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5935 - Do not leak request if server was not able to process it (5154)
5936 - If mds_open unable to find parent dir, make that negative lookup(5154)
5937 - don't create new directories with extent-mapping (5909, 5936)
5939 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5940 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5941 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5942 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5943 - support for RHEL4 kernel on i686 (5773)
5944 - provide error messages when incompatible logs are encountered (5898)
5946 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5947 * version 1.4.0.10 (1.4.1 release candidate 1)
5949 - don't keep a lock reference when lock is not granted (4238)
5950 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5951 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5952 - drop import inflight refcount on signal_completed_replay error (5255)
5953 - unlock page after async write error during send (3677)
5954 - handle missing objects in filter_preprw_read properly (5265)
5955 - no transno return for symlink open, don't save no-trasno open (3440)
5956 - don't try to complete elan receive that already failed (4012)
5957 - free RPC server reply state on error (5406)
5958 - clean up thread from ptlrpc_start_thread() on error (5160)
5959 - readahead could read extra page into cache that wasn't ejected (5388)
5960 - prevent races in class_attach/setup/cleanup/detach (5260)
5961 - don't dereference de->d_inode after l_dput of de (5458)
5962 - use "int" for stripe value returned from lock_to_stripe (5544)
5963 - mballoc allocation and error-checking fixes in 2.6 (5504)
5964 - block device patches to fix I/O request sizes in 2.6 (5482)
5965 - look up hostnames for IB nals (5602)
5966 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5967 - don't start multiple acceptors for the same port (5277)
5968 - fix incorrect LASSERT in mds_getattr_name (5635)
5969 - export a proc file for general "ping" checking (5628)
5970 - fix "lfs check" to not block when the MDS is down (5628)
5972 - service request history (4965)
5973 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5974 - create an "evict_client" /proc entry on OSTs, like the MDS has
5975 - fix mount usage message, return errors per mount(8) (5168)
5976 - change grep [] to grep "[]" in tests so they work in more UMLs
5977 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5978 - remove ancient LOV_MAGIC_V0 EA support (5047)
5979 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5980 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5981 - allow readahead allocations to fail when low on memory (5383)
5982 - mmap locking landed again, after considerable improvement (2828)
5983 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5985 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5988 - send OST transaction number in read/write reply to free req (4966)
5989 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5990 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5991 - the watchdog thread now runs as interruptible (5246)
5992 - flock/lockf fixes (but it's still disabled, pending 5135)
5993 - don't use EXT3 constants in llite code (5094)
5994 - memory shortage at startup could cause assertion (5176)
5996 - reorganization of lov code
5997 - single portals codebase
5999 - add extents/mballoc support (5025)
6000 - direct I/O reads in the obdfilter (4048)
6001 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6003 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6006 - send OST transaction number in read/write reply to free req (4966)
6007 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6008 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6009 - the watchdog thread now runs as interruptible (5246)
6010 - handle missing objects in filter_preprw_read properly (5265)
6011 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6012 - drop import inflight refcount on signal_completed_replay error (5255)
6013 - unlock page after async write error during send (3677)
6014 - return original error code on reconstructed replies (3761)
6015 - no transno return for symlink open, don't save no-trasno open (3440)
6017 - add pid to ldlm debugging output (4922)
6018 - bump the watchdog timeouts -- we can't handle 30sec yet
6019 - extra debugging for orphan dentry/inode bug (5259)
6021 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6024 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6025 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6026 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6027 - fix NULL dereference in /proc/sys/portals/routes (4827)
6028 - allow failed mdc_close() operations to be interrupted (4561)
6029 - stop precreate on OST before MDS would time out on it (4778)
6030 - don't send partial-page writes before EOF from client (4410)
6031 - discard client grant for sub-page writes on large-page clients (4520)
6032 - don't free dentries not owned by NFS code, check generation (4806)
6033 - fix lsm leak if mds_create_objects() fails (4801)
6034 - limit debug_daemon file size, always print CERROR messages (4789)
6035 - use transno after validating reply (3892)
6036 - process timed out requests if import state changes (3754)
6037 - update mtime on OST during writes, return in glimpse (4829)
6038 - add mkfsoptions to LDAP (4679)
6039 - use ->max_readahead method instead of zapping global ra (5039)
6040 - don't interrupt __l_wait_event() during strace
6042 - add software watchdogs to catch hung threads quickly (4941)
6043 - make lustrefs init script start after nfs is mounted
6044 - fix CWARN/ERROR duplication (4930)
6045 - return async write errors to application if possible (2248)
6046 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6047 - print NAL number in %x format (4645)
6048 - update barely-supported suse-2.4.21-171 series (4842)
6049 - support for sles 9 %post scripts
6050 - support for building 2.6 kernel-source packages
6051 - support for sles km_* packages
6053 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6056 - ignore -ENOENT errors in osc_destroy (3639)
6057 - notify osc create thread that OSC is being cleaned up (4600)
6058 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6059 - reconstruct ost_handle() like mds_handle() (4657)
6060 - create a new thread to do import eviction to avoid deadlock (3969)
6061 - let lconf resolve symlinked-to devices (4629)
6062 - don't unlink "objects" from directory with default EA (4554)
6063 - hold socknal file ref over connect in case target is down (4394)
6064 - allow more than 32000 subdirectories in a single directory (3244)
6065 - fix blocks count for O_DIRECT writes (3751)
6066 - OST returns ENOSPC from object create when no space left (4539)
6067 - don't send truncate RPC if file size isn't changing (4410)
6068 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6069 - bind to privileged port in socknal and tcpnal (3689)
6071 - rate limit CERROR/CWARN console message to avoid overload (4519)
6072 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6073 - basic mmap support (3918)
6074 - kernel patch series update from b1_4 (4711)
6076 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6079 - avoid crash during MDS cleanup with OST shut down (2775)
6080 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6081 - don't use bad inodes on the MDS (3744)
6082 - dynamic object preallocation to improve recovery speed (4236)
6083 - don't hold spinlock over lock dumping or change debug flags (4401)
6084 - don't zero obd_dev when it is force cleaned (3651)
6085 - print grants to console if they go negative (4431)
6086 - "lctl deactivate" will stop automatic recovery attempts (3406)
6087 - look for existing locks in ldlm_handle_enqueue() (3764)
6088 - don't resolve lock handle twice in recovery avoiding race (4401)
6089 - revalidate should check working dir is a directory (4134)
6091 - don't always mark "slow" obdfilter messages as errors (4418)
6093 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6096 - don't close LustreDB during write_conf until it is done (3860)
6097 - fix typo in lconf for_each_profile (3821)
6098 - allow dumping logs from multiple threads at one time (3820)
6099 - don't allow multiple threads in OSC recovery (3812)
6100 - fix debug_size parameters (3864)
6101 - fix mds_postrecov to initialize import for llog ctxt (3121)
6102 - replace config semaphore with spinlock (3306)
6103 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6104 - don't allow enqueue to complete on a destroyed export (3822)
6105 - down write_lock before checking llog header bitmap (3825)
6106 - recover from lock replay timeout (3764)
6107 - up llog sem before sending rpc (3652)
6108 - reduce ns lock hold times when setting kms (3267)
6109 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6110 - fix NULL deref and obd_dev leak on setup error (3312)
6111 - replace some LBUG about llog ops with error handling (3841)
6112 - don't match INVALID dentries from d_lookup and spin (3784)
6113 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6114 - fix invalid assertion in ptlrpc_set_wait (3880)
6116 - add libwrap support for the TCP acceptor (3996)
6117 - add /proc/sys/portals/routes for non-root route listing (3994)
6118 - allow setting MDS UUID in .xml (2580)
6119 - print the stack of a process that LBUGs (4228)
6121 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6124 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6125 - make sure to unset replay flag from failed open requests (3440)
6126 - if default stripe count is 0, use OST count for inode size (3636)
6127 - update parent mtime/ctime on client for create/unlink (2611)
6128 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6129 - free recovery state on server during a forced cleanup (3571)
6130 - unregister_reply for resent reqs (3063)
6131 - loop back devices mounting and status check on 2.6 (3563)
6132 - fix resource-creation race that can provoke i_size == 0 (3513)
6133 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6134 - more debugging for page-accounting assertion (3746)
6135 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6136 - avoid deadlock after precreation failure (3758)
6137 - fix race and lock order deadlock in orphan handling (3450, 3750)
6138 - add validity checks when grabbing inodes from l_ast_data (3599)
6140 - add /proc/.../recovery_status to obdfilter (3428)
6141 - lightweight CDEBUG infrastructure, debug daemon (3668)
6142 - change default OSC RPC parameters to be better on small clusters
6143 - turn off OST read cache for files smaller than 32MB
6144 - install man pages and include them in rpms (3100)
6145 - add new init script for (un)mounting lustre filesystems (2593)
6146 - run chkconfig in %post for init scripts (3701)
6147 - drop scimac NAL (unmaintained)
6149 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6152 - clean kiobufs before and after use (3485)
6153 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6154 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6155 - change init script to fail more clearly if not run as root (1528)
6156 - allow clients to reconnect during replay (1742)
6157 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6158 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6159 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6160 - fix cond_resched (3554)
6161 - only evict unfinished clients after recovery (3515)
6162 - allow bulk resend, prevent data loss (3570)
6163 - dynamic ptlrpc request buffer allocation (2102)
6164 - don't allow unlinking open directory if it isn't empty (2904)
6165 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6166 - remove extraneous obd dereference causing LASSERT failure (3334)
6167 - don't use get_cycles() when creating temp. files on the mds (3156)
6168 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6169 - handle EEXIST for set-stripe, set proper directory name (3336)
6171 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6172 - fix ksocknal_fmb_callback() error messages (2918)
6174 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6177 - don't copy lvb into (possibly NULL) reply on error (2983)
6178 - don't deref dentry after dput, don't free lvb on error (2922)
6179 - use the kms to determine writeback rpc length (2947)
6180 - increment oti_logcookies when osc is inactive (2948)
6181 - update client's i_blocks count via lvb messages (2543)
6182 - handle intent open/close of special files properly (1557)
6183 - mount MDS with errors=remount-ro, like obdfilter (2009)
6184 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6185 - don't use cancelling-locks' kms values (2947)
6186 - use highest lock extent for kms, not last one (2925)
6187 - don't dereference ERR_PTR() dentry in error handling path (3107)
6188 - fix thread race in portals_debug_dumplog() (3122)
6189 - create lprocfs device entries at setup instead of at attach (1519)
6190 - common AST error handler, don't evict client on completion race (3145)
6191 - zero nameidata in detach_mnt in 2.6 (3118)
6192 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6193 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6194 - initialize RPC timeout timer earlier for 2.6 (3219)
6195 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6196 - print nal/nid for unknown nid (3258)
6197 - additional checks for oscc recovery before doing precreate (3284)
6198 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6199 - don't crash in mdc_close for bad permissions on open (3285)
6200 - zero i_rdev for non-device files (3147)
6201 - clear page->private before handing to FS, better assertion (3119)
6202 - tune the read pipeline (3236)
6203 - fix incorrect decref of invalidated dentry (2350)
6204 - provide read-ahead stats and refine rpc in flight stats (3328)
6205 - don't hold journal transaction open across create RPC (3313)
6206 - update atime on MDS at close time (3265)
6207 - close LDAP connection when recovering to avoid server load (3315)
6208 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6209 - don't leak open file on MDS after open resend (3325)
6210 - serialize filter_precreate and filter_destroy_precreated (3329)
6211 - loop device shouldn't call sync_dev() for nul device (3092)
6212 - clear page cache after eviction (2766)
6213 - resynchronize MDS->OST in background (2824)
6214 - refuse to mount the same filesystem twice on same mountpoint (3394)
6215 - allow llmount to create routes for mounting behind routers (3320)
6216 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6217 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6218 - fix rare problem with rename on htree directories (3417)
6220 - allow default OST striping configuration per directory (1414)
6221 - fix compilation for qswnal for 2.6 kernels (3125)
6222 - increase maximum number of MDS request buffers for large systems
6223 - change liblustreapi to be useful for external progs like lfsck (3098)
6224 - increase local configuration timeout for slow disks (3353)
6225 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6227 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6230 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6231 - don't overwrite extent policy data in reply if lock was blocked (2901)
6232 - drop filter export grants atomically with removal from device (2663)
6233 - del obd_self_export from work_list in class_disconnect_exports (2908)
6234 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6235 - swab reply message in mdc_close, other PPC fixes (2464)
6236 - fix destroying of named logs (2325)
6237 - overwrite old logs when running lconf --write_conf (2264)
6238 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6239 - fix race in target_handle_connect (2898)
6240 - mds_reint_create() should take same inode create lock (2926)
6241 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6242 - don't close files for self_export to avoid uninitialized obd (2936)
6243 - allow MDS with the same name as client node (2939)
6244 - hold dentry reference for closed log files for unlink (2325)
6245 - reserve space for all logs during transactions (2059)
6246 - don't evict page beyond end of stripe extent (2925)
6247 - don't oops on a deleted current working directory (2399)
6248 - handle hard links to targets without a parent properly (2517)
6249 - don't dereference NULL lock when racing during eviction (2867)
6250 - don't grow lock extents when lots of conflicting locks (2919)
6252 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6255 - account for cache space usage on clients to avoid data loss (974)
6256 - lfsck support in lustre kernel code (2349)
6257 - reduce journal credits needed for BRW writes (2370)
6258 - orphan handling to avoid losing space on client/server crashes
6259 - ptlrpcd can be blocked, stopping ALL progress (2477)
6260 - use lock value blocks to assist in proper KMS, faster stat (1021)
6261 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6262 - recovery for initial connections (2355)
6263 - fixes for mds_cleanup_orphans (1934)
6264 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6265 - block all file creations until orphan recovery completes (1901)
6266 - client remove rq_connection from request struct (2423)
6267 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6268 - recovery timer race (2670)
6269 - mdc_close recovey bug (2532)
6270 - ptlrpc cleanup bug (2710)
6271 - mds timeout on local locks (2588)
6272 - namespace lock held during RPCs (2431)
6273 - handle interrupted sync write properly (2503)
6274 - don't try to handle a message that hasn't been replied to (2699)
6275 - client assert failure during cleanup after abort recovery (2701)
6276 - leak mdc device after failed mount (2712)
6277 - ptlrpc_check_set allows timedout requests to complete (2714)
6278 - wait for inflight reqs when ptlrpcd finishes (2710)
6279 - make sure unregistered services are removed from the srv_list
6280 - reset bulk XID's when resending them (caught by 1138 test)
6281 - unregister_bulk after timeout
6282 - fix lconf error (2694)
6283 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6284 - readahead locks pages, leaves pending causing memory pressure (2673)
6285 - increase OST request buffers to 4096 on large machines (2729)
6286 - fix up permission of existing directories in simple_mkdir (2661)
6287 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6288 - don't assign transno to errored transactions (2742)
6289 - don't delete objects on OST if given a bogus objid from MDS (2751)
6290 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6291 - if rq_no_resend, then timeout request after recovery (2432)
6292 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6293 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6294 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6295 - MDS recovery shouldn't delete objects that it has given out (2730)
6296 - if enqueue arrives after completion, don't clobber LVB (2819)
6297 - don't unlock pages twice when trigger_group_io returns error (2814)
6298 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6299 - don't write pages to disk if there was an error (1450)
6300 - don't ping imports that have recovery disabled (2676)
6301 - take buffered bytes into account when balancing socknal conn (2817)
6302 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6303 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6304 - remove little-endian swabbing of llog records (1987)
6305 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6306 - retry reposting request buffers if they fail (1191)
6307 - grow extent at grant time to avoid granting a revoked lock (2809)
6308 - lock revoke doesn't evict page if covered by a second lock (2765)
6309 - disable VM readahead to avoid reading outside lock extents (2805)
6311 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6312 - updated kernel patches for hp-2.4.20 kernel (2681)
6314 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6317 - fix truncated write corruption (2366)
6318 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6320 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6321 - ensure len(uuid) < 37 in lmc (1171)
6322 - fix ia64 OOPS in llog_test (2255)
6323 - zero end of page at obdfilter for partial page writes (2648)
6324 - don't leave stale dentries around after renames (bug 2428)
6325 - fix timeouts when evicting a client with a single lock held (2642)
6326 - set deadline for the initial HELLO message to drain (2634)
6327 - print out dotted-quad IP addresses in the socknal (2302)
6329 - additional debugging for MDS client eviction problem (2443)
6330 - fix mkfsoptions support for osts (2603, 2604)
6332 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6335 - add series for the vanilla 2.6.0 kernel
6336 - add series for the vanilla 2.4.24 kernel
6337 - add series for a cray x86/64 UL kernel drop
6338 - fix xattr patches for the vanilla 2.4.19 series
6340 - generate true UUIDs in lmc (1171)
6341 - have portals stack dumping break in UML (2466)
6342 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6343 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6344 - wake callback waiting threads on client eviction (2460)
6345 - Add --ptldebug and --subsystem to lmc (1719)
6346 - update assertion to allow safe interrupt allocation
6347 - set rq_no_resend for cancel requests (2432)
6348 - recalculate ptlrpcd timeout after resend (2494)
6349 - call vfs_rmdir when removing pending directories (2368)
6350 - fix renaming a file to itself (2429)
6351 - lmc creates a default one-stripe lov (2454)
6352 - expand procfs space to handle large clusters (2326)
6353 - increase UML stack to avoid overflow
6354 - update lconf's list of debug and subsystem masks
6355 - fix lfs find --obd (2510)
6356 - /proc tunable for disabling filter read caching (2591)
6357 - stop rpm packages from altering slapd.conf (2301)
6358 - disable nagle in the socknal under 0conf (2578)
6359 - choose mds inode size based on stripe count (2572)
6360 - fix kernel-source rpm problems (2516)
6362 - add --disable-doc to avoid pdf generation (2421)
6363 - update documentation, tests, type-os, comments
6364 - avoid format warnings on ia64
6365 - remove the TOE NAL
6366 - tiny code cleanups by removing unused fields
6368 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6371 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6372 - remove the most blatant lies from BUILDING file (2371)
6373 - change default debug level to reasonable production setting
6374 - reduce client side cache size to reduce cache flush time
6375 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6376 - make TCP zerocopy and pinger support enabled by default (2476)
6377 - sync writes completed after process exits caused crashes (2319)
6378 - maintain correct mount count on the MDS (2356)
6379 - backout 1557, because 2316 wasn't really fixed
6380 - better file I/O statistics gathering in /proc
6381 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6382 - another kernel patch to fix zero-copy TCP function export
6383 - don't take duplicate lock when processing re-sent getattr (2420)
6384 - lctl uses obd_self_export instead of creating new conn (2353)
6385 - MDS/OST recovery case which requires object creation asserted (2425)
6386 - move lfs from /usr/sbin to /usr/bin in packages
6387 - fix race between mds_client_add and mds_client_free (2417)
6388 - use kmalloc instead of slabs in portals (2430)
6389 - don't create duplicate records when a failover MDS is present (2442)
6390 - remove unnecessary mount age check (2332)
6391 - don't remove directory inodes from locks prematurely (2451)
6392 - don't break if MDS service name is the same as hostname (2103)
6393 - fix races in client write RPC generation when cache full (2482)
6395 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6398 - remove now-unused request->rq_obd (278)
6399 - if an allocation fails, print out how much memory we've used (1933)
6400 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6401 - add the "configurable stack size" patch to most series files (1256)
6402 - ability to write large log records, for 100+ OST configs (2306)
6403 - fix NULL deref when filter_prep fails (2314)
6404 - fix operator precedence error in filter_sync
6405 - dynamic allocation of socknal TX descriptors (2315)
6406 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6407 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6408 - ensure that CWARN messages in Portals always get to the syslog
6409 - __init/__exit are not for prototype decls (ldlm_init/exit)
6410 - x86-64 compile warning fixes
6411 - fix gateway LMC keyword conflict (2318)
6412 - fix MDS lock inversions in getattr/reint paths (1844)
6413 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6414 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6415 - DLM race condition prevented some lock evictions (2328)
6416 - ENOMEM detection and retry on socknal sends (2230)
6417 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6418 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6419 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6420 - better error messages when a client is rejected during recovery (1505)
6421 - avoid cancelling locks which were never granted, after failure (2330)
6422 - fix i_sem/journal inversion in mds_client_add (2333)
6423 - fix truncate/getattr lock cycle deadlock (2334)
6424 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6425 - fix two rare exit paths which could leak an l_lock() ref (2321)
6426 - fencepost error in MDS/OST orphan recovery (2226)
6427 - make log record alignment 8 bytes (1988)
6428 - lstripe now fails when requested offset > ost_count (2237)
6429 - ensure that all kernel series have a complete list.h (1607)
6430 - fix crashes in special-file operations (2316)
6431 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6433 - allow configurable automake binary, for testing new versions
6434 - small update to the lfs documentation
6436 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6438 * fix negative export reference count in fsfilt_sync (2312)
6440 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6441 * release candidate 0.9.1
6443 - orphans are moved into the PENDING directory for possible recovery
6444 - replayed opens now open by fid for orphan/rename safety (1042)
6445 - last close of an orphan inode generates a transno (683)
6446 - chdir() and mount() now pin the directory entry (1020)
6447 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6448 - discard very old requests without processing them (1502)
6449 - remove some common, well-understood CERRORs (1505)
6450 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6451 - clear "grant" flags in OST replies until OST grant code lands (1644)
6452 - fix read performance by not clobbering i_blksize on client (1598)
6453 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6454 - make LRU size a /proc tunable, clears locks when reduced (707)
6455 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6456 - server locks take references on exports now (1558)
6457 - build fixes for 2.4.20-rh trees (1663)
6458 - return an error from lov_create if all OSCs are inactive (1751)
6459 - fix import levels when a reconnect happens without a timeout (1597)
6460 - exit early from mds_open if we get a lookup error (1749)
6461 - partial page read at EOF wouldn't wait for disk before sending (1642)
6462 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6463 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6464 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6465 - write/truncate lock inversion (1639)
6466 - Don't auto-load obdclass, portals modules during cleanup (1495)
6467 - fix timestamps from jumping to "now" (1763)
6468 - extra journal assertions (1648)
6469 - add an extra multiunlink test (1771)
6470 - fix read_record/write_record API (1776)
6471 - fix leak of offset_extent, possible incorrect i_size later (1772)
6472 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6473 - Debugging for the kqswnal_get_idle_tx problems (1820)
6474 - Allow recovery to be attempted multiple times (1536)
6475 - Write out MDS last_rcvd file after it is first created (1600)
6476 - Fix tx_descriptor leak in failed transmit situations (1827)
6477 - ext3 journaling fixes for assertion failure after IO error (1871)
6478 - class_export_put() on freed export after completion AST error (1896)
6479 - Fix revalidate looping in VFS (1322)
6480 - Don't access a freed export during MDS_REINT timeout (1521)
6481 - Add open-unlink recovery support on the MDS (1673,1764)
6482 - Return an error if no MDS data was read from last_rcvd (1946)
6483 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6484 - Don't setup a disk device that doesn't match exported UUID (317)
6485 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6486 - avoid committing NULL handle in force close
6487 - local.sh is now a one-stripe LOV configuration
6488 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6489 - fix ext3 htree duplicate directory entry corruption (1516)
6490 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6491 - update to new LOV EA format (2097)
6492 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6494 2003-06-15 Phil Schwan <phil@clusterfs.com>
6497 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6498 - per-import/export recovery handling (958, 931, 959)
6499 - multiple last-rcvd slots, for serving multiple FSes (949)
6500 - connections are again shared between multiple imp/exports (963, 964)
6501 - "umount -f" would hang if any requests needed to be sent (393, 978)
6502 - avoid pinning large req buffer by copying for queued messages (989)
6503 - add "uuid" to "lctl device" command to help upcalls (991)
6504 - "open" RPCs with transnos would confuse recovery counters (1037)
6505 - do proper endian conversion of all wire messages (288, 340, 891)
6506 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6507 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6508 - fix ptlrpc_connection leak in target_handle_connect (1174)
6509 - fix import refcounting bug in OST and MDS cleanup (1134)
6510 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6511 - fix very unlikely obd_types race condition (501)
6512 - remove osc_open hack for echo_client (1187)
6513 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6514 - a failure in read_inode2 leads to deadlock (1139)
6515 - cancel ack-locks as soon as transaction is committed (1072)
6516 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6517 - make sure to commitrw after any preprw to avoid deadlock (1162)
6518 - failing to execute a file in a lustre FS would lock inode (1203)
6519 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6520 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6521 - obd_uuid2tgt would walk off the end of the list (1255)
6522 - on IA64 the osc would give portals incorrect bulk size (1258)
6523 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6524 - fix lock inversion caused by new llite matching code (1282)
6525 - limit the number of dirty pages on a client to 10MB (1286)
6526 - timed out locks were not being corrected cancelled (1289)
6527 - fix O_DIRECT above 4GB on IA-32 (1292)
6528 * major user-visible changes
6529 - fail out/fail over policy now controlled by the upcall (993)
6531 - add OBD_PING to check server availability and failure (954)
6532 - lustre messages are now sent in sending host order (288, 340, 891)
6533 - add eadatalen to MDS getattr reply (340)
6534 - OST read replies may contain second buffer, with per-page status (593)
6536 2003-03-11 Phil Schwan <phil@clusterfs.com>
6539 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6540 - failed open()s could cause deadlock; fixed (867, 869)
6541 - stop cancelling OST locks when files are closed (481)
6542 - overlapping XID spaces caused network corruption (851, 853)
6543 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6544 - setattr_raw added, to do single-RPC, server-side setattrs
6545 - lmc/lconf syntax change for OST UUIDs
6546 - fix crashy race condition between ptlrpc_free_req and osc_close
6547 - don't use request in mdc_enqueue if we hit a timeout (889)
6548 - don't set the inode i_size for regular files from the MDS (896)
6549 - handle out of order completion AST (842)
6550 - don't LBUG if a lock request times out after receiving AST (913)
6551 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6552 - if a bad lock AST arrives, send an error instead of dropping entirely
6553 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6554 - fix leak in bulk IO when only partially completed (899, 900, 926)
6555 - fix O_DIRECT for ia64 (55)
6556 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6557 - C-z after timeout could hang a process forever; fixed (977)
6559 - client-side I/O cache (678, 924, 929, 941, 970)
6561 - READPAGE and SETATTRs which don't take server-side locks get
6564 2003-02-11 Phil Schwan <phil@clusterfs.com>
6567 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6568 - Fix fsfilt_extN_readpage() to read a full page of directory
6569 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6570 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6571 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6572 - Most intent operations were removed, in favour of a new RPC mode
6573 that does a single RPC to the server and bypasses most of the VFS
6574 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6575 - Aggressively cancel local locks on DLM servers
6576 - mds_reint_unlink sends EA to the client if it's the last nlink.
6577 client uses that EA to unlink OST objects.
6578 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6579 - recursive symlinks were fixed (439)
6580 - fixed NULL deref in DEBUG_REQ
6581 - filter_update_lastobjid no longer calls sync, which annoyed extN
6582 - fixed multi-client small-writes to a single file problem (445)
6583 - fixed mtime updates during file writes (607)
6584 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6585 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6586 - fixed Linux OST size reporting problem (444, 656)
6587 - OST now updates object mtime with writes or setattr (607, 619)
6588 - client verifies file size before zeroing page past EOF (445)
6589 - OST now writes last allocated objid to disk with allocation (108)
6590 - LOV on echo now works (409)
6592 - mds_reint_unlink sends a new buffer, with the EA included. this
6593 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6594 set if a regular file was being unlinked, and it was the last link
6595 - use PtlGet from the target for bulk writes (315)
6596 - OST now updates object mtime with writes or setattr (607, 619)
6597 - LDLM now has a grant-time callback to revalidate locked items, if
6599 - Many MDS operations were reorganized to combat race conditions
6601 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6602 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6604 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6607 - Fully reactivate OST imports after reconnection (512, others)
6608 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6609 - More graceful error handling for truncating on dead OST (515)
6610 - Don't error out unless we're actually accessing dead stripes (474)
6611 - Fix garbage sizes when stripes are missing (410)
6612 - LRU counters were broken, causing constant lock purge (433, 432)
6613 - garbage on read from stripes with failed OSTs (441)
6614 - mark OSCs as active before reconnecting during recovery (438)
6615 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6616 - lfind did not preserve OST order in output (443)
6617 - symlinks cause hung clients, incorrect data (439)
6618 - stop dereferencing request after dropping refcount (457)
6619 - don't LASSERT(spin_is_locked) on non-SMP (455)
6620 - fixes for many rename() bugs
6621 - fstat didn't correctly synchronize attributes (399)
6622 - server must handle lock cancellation during blocking AST prep (487)
6623 - bulk descriptors were free()d too soon (511)
6624 - fix paths in lconf, which would load incorrect modules (451, 507)
6625 - fix confusing lconf 'host not found' error message (386)
6626 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6627 - fix race condition in mdc_blocking_ast() for inode access (526)
6628 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6629 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6630 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6631 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6632 - fix lstripe in several ways (kernel side) (527)
6633 - fix request leak in ldlm_cli_enqueue (262)
6634 - incorrect OSC was marked inactive after OST failure
6635 - call mds_fs_cleanup before unmounting filesystem (524)
6636 - fix races between taking ns_lock and ldlm_lock_change_resource
6637 - fix races updating LOV export open file list
6638 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6639 - fix recovery NULL deref in ldlm_cli_cancel_unused
6640 - fix some DLM races by using new hash table for lock handles (419)
6641 - permit the client to specify desired inodes, at replay
6642 - duplicate requests when we queue them for replay reintegration
6643 - fix last_rcvd offset calculation
6644 - sync after each recovered transaction, so we always make progress
6645 - never, not always, ERESTART requests without transnos
6646 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6648 - skip replay if the MDS says that the client is already connected
6649 - don't check for a recovery-enabled export to match lctl's UUID
6650 - don't INC_USE_COUNT for phantom exports
6651 - don't crash when cleaning up phantom exports (567)
6652 - don't double-finish or set replay data for errored mdc_open requests
6653 - abort requests when they time out, so we don't get old replies
6654 - send/receive replies for AST messages again
6655 - if the client says that it doesn't have the lock, cancel it on the
6657 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6658 mark it as destroyed, it will all work out when decref is called
6659 - fix module use counts (22, 581)
6661 - ASTs now expect a reply (server cancels lock on error reply)
6663 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6666 - fix many simultaneous client startup (392)
6667 - fix dentry->d_it clobbering
6668 - credentials weren't being shipped for readdir/getattr operations
6669 - remove invalid assertions triggered during some concurrent MD
6671 - proper Lustre versions added (336, 389)
6672 - fix memory leak for create error case (398)
6673 - fix LOV locking bug that would get cli/srv out of sync
6674 - fix echo client over LOV (409)
6675 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6676 - fix double-O_EXCL intent crash (424)
6677 - avoid sending multiple lock CANCELs (352)
6679 - MDS can do multi-client recovery (modulo bugs in new code)
6681 - many updates, edits, cleanups
6683 2002-11-18 Phil Schwan <phil@clusterfs.com>
6686 - fix null d_it dereference (346)
6687 - fix full OST/dbench hang (333)
6688 - fix permission problem with file removal (286)
6689 - fix removal of OSCs from LOV when they fail
6690 - fix NULL deref during bulk timeout (214)
6691 - fix problems related to multiple filesystems on one MDS (241)
6692 - fixed serious subtle metadata locking bugs
6693 - free locks on clients when inodes are removed due to memory
6695 - fix inode pointer in lock data (285)
6696 - partial support for multiple MDS on a single host (241)
6697 - data locks weren't cancelled at clear_inode time (290, 311)
6698 - intent locks could lead to unbounded lock growth (205)
6699 - added a maximum lock count, an LRU list, and a flusher
6700 - fix multiple rename (365)
6701 - properly abstracted the echo client
6702 - OSC locked 1 byte too many; fixed
6703 - rewrote brw callback code:
6704 - fixed recovery bugs related to LOVs (306)
6705 - fixed too-many-pages-in-one-write crash (191)
6706 - fixed (again) crash in sync_io_timeout (214)
6707 - probably fixed callback-related race (385)
6709 - Add capability to MDS protocol
6710 - LDLM cancellations and callbacks on different portals
6712 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6715 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6716 - defer open object destruction to close time (601981, 138)
6717 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6718 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6719 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6720 - don't call obd_setattr in ll_file_release for destroyed objects
6721 * protocol change to lustre_msg: move |version| and add |flags|
6722 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6723 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6724 * added replay of create, unlink, link and rename operations during
6725 MDS failover; recovery should be much more robust now
6726 * remove failed OSCs from LOVs (only lov_create uses this so far)
6727 * the lustre-HOWTO was brought (more) up to date (582544)
6729 2002-10-23 Phil Schwan <phil@clusterfs.com>
6732 - in-use dentries weren't being reused properly (617851)
6733 - prevent multiple LDLM setup (599178)
6734 - fix LOV size calculations for truncate (617853)
6735 - fix client handling of MDS intent errors (POSIX)
6736 - fix permission bug in lovstripe.c test (624321)
6737 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6738 - truncate past end of file could corrupt data
6739 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6740 - a race in recovery could return ETIMEDOUT to apps (623947)
6741 - building outside the source directory was fixed
6742 * the lustre-HOWTO was brought (more) up to date (582544)
6743 * major progress was made on recovery functionality
6745 2002-10-10 Phil Schwan <phil@clusterfs.com>
6748 - recovery deadlock fix
6749 - rm -rf causes LBUG fix (617817)
6750 - file open by multiple tasks fix (618962)
6751 - directory permissions bugs (602707 and 620007)
6752 - journal_stop fixed with locking (611313)
6753 - O_APPEND failures resolved (618273, perhaps 614459)
6754 - lconf PATH fix (619770)
6755 - IA64 build fix (621450)
6756 - RPC buffer sizes scale with amount of memory
6758 2002-10-01 Phil Schwan <phil@clusterfs.com>
6761 - locks would be cancelled without throwing away data pages,
6762 resulting in inconsistent data (605627)
6763 - inode attributes were not always being refreshed (605627, 612449)
6764 - lconf now continues to cleanup after lctl reports an error
6765 - MDS now enforces user permissions (602707)
6766 - lprocfs cleanup fixed, but not yet enabled (614157)
6767 - fixed infinite server hang, should a client not respond to an AST
6768 - avoid going into recovery if user calls readlink() with a buffer
6769 that's too small (613941)
6770 - AST RPCs no longer require replies (614867) -- this may be changed
6771 - don't crash server if client sends an IOV that's too big (611336)
6772 - fixed lock conversion deadlock (611892)
6773 - fixed the following of symlinks (614622)
6774 * recovery: the server can remove locks from a client that dies, other
6775 clients can make progress
6776 * more extN patch fixes
6777 * compile-time configurable ptlrpc buffer allocations
6779 - collaborative read cache document
6780 - Lustre Lite Performance CDR document-in-progress
6782 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6785 - fix typo in patch-2.4.18
6787 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6790 - clear ptlrpc request each time in handle_incoming_request()
6791 - unlink of files now destroys the object on the OST
6793 2002-09-19 Peter Braam <braam@clusterfs.com>
6795 * add hard link support
6796 * change obdfile creation method
6797 * kernel patch changed
6799 2002-09-19 Peter Braam <braam@clusterfs.com>
6802 - stack overflow bug in extN fixed
6804 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6806 * documentation updates
6807 - add man pages for config tools
6808 - update tests/README to describe testing with new config tools
6809 - finish metadata API descriptions
6810 * bug fixes and cleanups
6811 - statfs workaround for 16TB limit
6812 - LOV stripe allocation improved, can stripe on subset of OSTs
6813 - LOV file size/IO offset was wrong for files > 4GB in size
6814 - object EA data was being dropped, caused files to be unreadable
6815 - memory overflow with non-LOV OST caused memory corruption
6816 - fixed regression tests to work with new config tools, obdfilter
6817 - fixed bug when directory size became larger than 1 block
6818 - fixed bug (for single client case) when PWD was deleted
6819 - invalidate local directory pages when doing intent-based ops
6820 - avoid LDLM oops when lock callback contained bad data
6822 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6824 * documentation updates
6825 * bug fixes and cleanups
6826 - configuration tools
6829 - 64-bit compile warnings
6830 - 64-bit internal statfs data
6832 * test_brw on persistent OST devices
6834 * lprocfs (disabled)
6836 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6838 * documentation updates
6839 * bug fixes and cleanups
6840 * configuration tools
6842 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6845 * many small fixes to 0_5_4
6846 * io/network handling
6847 * thinkos in MDS operations
6849 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6852 * crucial basic fixes to 0.5.3
6853 * IOR, Iozone work over Elan
6856 2002-08-07 Phil Schwan <phil@clusterfs.com>
6857 * version 0_5_3, our first alpha
6858 * we use the new Portals iovs
6859 * documentation updates
6860 * bug fixes and cleanups
6861 * small changes in the DLM wire protocol
6863 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6864 * version 0_5_1 with some initial stability,
6865 * locking on MD and file I/O.
6866 * documentation updates
6867 * several bug fixes since 0.5.0
6868 * small changes in wire protocol
6870 2002-07-18 Phil Schwan <phil@clusterfs.com>
6872 * delivered as Lustre Light Alpha
6873 * fixed a crash after handling invalid MDS requests
6874 * fixed directory pages for architectures with non-4k pages sizes
6876 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6877 * release version v0_4_4
6878 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6880 * Fixes a number of interruption problems with OST operations.
6881 * Update documentation for portals header changes
6882 * Move all wire protocol structs/defines to lustre_idl.h
6883 * Fixes symlink length bug.
6884 * Add tcpdump to repository.
6886 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6887 * release version v0_4_3
6888 * Fixes statfs for inodes on extN.
6889 * Fixes bug in runtests which would delete /etc/hosts.
6890 * Use 64-bit object IDs wherever possible (not into VFS though)
6891 Remove ost_get_info, which is unused by lustre, and out of date.
6893 2002-07-03 Peter Braam <braam@clusterfs.com>
6894 * release version v0_4_2 Fixes a lookup error (type not passed)
6895 * move forward to head of Portals
6896 * move forward to latest Lustre kernel
6898 2002-06-25 Peter Braam <braam@clusterfs.com>
6899 * release version v0_4_1. Hopefully stable on single node use.