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 4.18.0-425.3.1.el8 (RHEL8.7)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 3.10.0-1062.18.1.el7 (RHEL7.7)
11 3.10.0-1127.19.1.el7 (RHEL7.8)
12 3.10.0-1160.80.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.18.0-348.23.1.el8 (RHEL8.5)
19 4.18.0-372.32.1.el8 (RHEL8.6)
20 4.4.120-92.70 (SLES12 SP2)
21 4.4.180-94.100 (SLES12 SP3)
22 4.4.0-131 (Ubuntu 16.04)
23 4.15.0-32 (Ubuntu 18.04)
24 5.4.0-48 (Ubuntu 20.04)
25 vanilla linux 5.4.0 (ZFS + ldiskfs)
26 vanilla linux 5.4.21 (ZFS + ldiskfs)
27 vanilla linux 5.4.136 (ZFS + ldiskfs)
28 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
29 * Client primary kernels built and tested during release cycle:
30 5.14.0-70.30.1.el9 (RHEL9.0)
31 4.18.0-425.3.1.el8 (RHEL8.7)
32 5.4.0-37 (Ubuntu 20.04)
33 5.14.21-150400.24.28 (SLES15 SP4)
34 * Other clients known to build on these kernels at some point (others may also work):
35 3.10.0-862.14.4.el7 (RHEL7.5)
36 3.10.0-957.27.2.el7 (RHEL7.6)
37 4.14.0-49.13.1.el7a (RHEL7.5)
38 4.14.0-115.2.2.el7a (RHEL7.6)
39 3.10.0-1062.18.1.el7 (RHEL7.7)
40 3.10.0-1127.19.1.el7 (RHEL7.8)
41 3.10.0-1160.80.1.el7 (RHEL7.9)
42 4.18.0-80.11.2.el8 (RHEL8.0)
43 4.18.0-147.8.1.el8 (RHEL8.1)
44 4.18.0-193.28.1.el8 (RHEL8.2)
45 4.18.0-240.22.1.el8 (RHEL8.3)
46 4.18.0-305.25.1.el8 (RHEL8.4)
47 4.18.0-348.23.1.el8 (RHEL8.5)
48 4.18.0-372.32.1.el8 (RHEL8.6)
49 4.4.120-92.70 (SLES12 SP2)
50 4.4.180-94.100 (SLES12 SP3)
51 4.12.14-95.48 (SLES12 SP4)
52 4.12.14-122.133 (SLES12 SP5)
53 4.12.14-197.75 (SLES15 SP1)
54 5.3.18-24.96 (SLES15 SP2)
55 5.3.18-150300.59.93 (SLES15 SP3)
56 4.4.0-131 (Ubuntu 16.04)
57 4.15.0-48 (Ubuntu 18.04)
58 5.8.0-53 (Ubuntu 20.04.2 HWE)
59 5.11.0-31 (Ubuntu 20.04.3 HWE)
60 5.11.0 (vanilla kernel.org)
61 5.10.0-60.56.0.84.oe2203 (openEuler 22.03 LTS)
62 * Recommended e2fsprogs version: 1.46.5.wc1 or newer
63 * Recommended ZFS version: 2.1.5
64 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
65 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
66 stack overflow. For more information, please refer to bugzilla 17630.
67 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
68 the re-exporting nodes to be mounted with "32bitapi" mount option
70 --------------------------------------------------------------------------------
75 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
76 for currently supported client and server kernel versions.
77 * Server primary kernels built and tested during release cycle:
78 4.18.0-348.2.1.el8 (RHEL8.5)
79 * Other server kernels known to build and work at some point (others may also work):
80 3.10.0-862.14.4.el7 (RHEL7.5)
81 3.10.0-957.27.2.el7 (RHEL7.6)
82 3.10.0-1062.18.1.el7 (RHEL7.7)
83 3.10.0-1127.19.1.el7 (RHEL7.8)
84 3.10.0-1160.49.1.el7 (RHEL7.9)
85 4.18.0-80.11.2.el8 (RHEL8.0)
86 4.18.0-147.8.1.el8 (RHEL8.1)
87 4.18.0-193.28.1.el8 (RHEL8.2)
88 4.18.0-240.22.1.el8 (RHEL8.3)
89 4.18.0-305.25.1.el8 (RHEL8.4)
90 4.4.120-92.70 (SLES12 SP2)
91 4.4.180-94.100 (SLES12 SP3)
92 4.4.0-131 (Ubuntu 16.04)
93 4.15.0-32 (Ubuntu 18.04)
94 5.4.0-48 (Ubuntu 20.04)
95 vanilla linux 5.4.0 (ZFS + ldiskfs)
96 vanilla linux 5.4.21 (ZFS + ldiskfs)
97 vanilla linux 5.4.136 (ZFS + ldiskfs)
98 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
99 * Client primary kernels built and tested during release cycle:
100 4.18.0-348.2.1.el8 (RHEL8.5)
101 5.4.0-37 (Ubuntu 20.04)
102 5.3.18-59.27 (SLES15 SP3)
103 * Other clients known to build on these kernels at some point (others may also work):
104 3.10.0-862.14.4.el7 (RHEL7.5)
105 3.10.0-957.27.2.el7 (RHEL7.6)
106 4.14.0-49.13.1.el7a (RHEL7.5)
107 4.14.0-115.2.2.el7a (RHEL7.6)
108 3.10.0-1062.18.1.el7 (RHEL7.7)
109 3.10.0-1127.19.1.el7 (RHEL7.8)
110 3.10.0-1160.49.1.el7 (RHEL7.9)
111 4.18.0-80.11.2.el8 (RHEL8.0)
112 4.18.0-147.8.1.el8 (RHEL8.1)
113 4.18.0-193.28.1.el8 (RHEL8.2)
114 4.18.0-240.22.1.el8 (RHEL8.3)
115 4.18.0-305.25.1.el8 (RHEL8.4)
116 4.4.120-92.70 (SLES12 SP2)
117 4.4.180-94.100 (SLES12 SP3)
118 4.12.14-95.48 (SLES12 SP4)
119 4.12.14-122.91 (SLES12 SP5)
120 4.12.14-197.75 (SLES15 SP1)
121 5.3.18-24.96 (SLES15 SP2)
122 4.4.0-131 (Ubuntu 16.04)
123 4.15.0-48 (Ubuntu 18.04)
124 5.8.0-53 (Ubuntu 20.04.2 HWE)
125 5.11.0-31 (Ubuntu 20.04.3 HWE)
126 5.11.0 (vanilla kernel.org)
127 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
128 * Recommended ZFS version: 2.0.7
129 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
130 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
131 stack overflow. For more information, please refer to bugzilla 17630.
132 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
133 the re-exporting nodes to be mounted with "32bitapi" mount option
135 --------------------------------------------------------------------------------
139 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
140 for currently supported client and server kernel versions.
141 * Server primary kernels built and tested during release cycle:
142 3.10.0-1062.18.1.el7 (RHEL7.7)
143 3.10.0-1127.19.1.el7 (RHEL7.8)
144 3.10.0-1160.6.1.el7 (RHEL7.9)
145 * Other server kernels known to build and work at some point (others may also work):
146 3.10.0-862.14.4.el7 (RHEL7.5)
147 3.10.0-957.27.2.el7 (RHEL7.6)
148 4.18.0-80.11.2.el8 (RHEL8.0)
149 4.18.0-147.8.1.el8 (RHEL8.1)
150 4.18.0-193.28.1.el8 (RHEL8.2)
151 4.18.0-240.1.1.el8 (RHEL8.3)
152 4.4.120-92.70 (SLES12 SP2)
153 4.4.180-94.100 (SLES12 SP3)
154 4.4.0-131 (Ubuntu 16.04)
155 4.15.0-32 (Ubuntu 18.04)
156 5.4.0-48 (Ubuntu 20.04)
157 vanilla linux 5.4.0 (ZFS + ldiskfs)
158 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
159 * Client primary kernels built and tested during release cycle:
160 3.10.0-1062.18.1.el7 (RHEL7.7)
161 3.10.0-1127.19.1.el7 (RHEL7.8)
162 3.10.0-1160.6.1.el7 (RHEL7.9)
163 4.12.14-95.48 (SLES12 SP4)
164 4.12.14-122.20 (SLES12 SP5)
165 4.12.14-197.75 (SLES15 SP1)
166 5.3.18-24.24 (SLES15 SP2)
167 4.15.0-48 (Ubuntu 18.04)
168 * Other clients known to build on these kernels at some point (others may also work):
169 3.10.0-862.14.4.el7 (RHEL7.5)
170 3.10.0-957.27.2.el7 (RHEL7.6)
171 4.14.0-49.13.1.el7a (RHEL7.5)
172 4.14.0-115.2.2.el7a (RHEL7.6)
173 4.18.0-80.11.2.el8 (RHEL8.0)
174 4.18.0-147.8.1.el8 (RHEL8.1)
175 4.18.0-193.28.1.el8 (RHEL8.2)
176 4.18.0-240.1.1.el8 (RHEL8.3)
177 4.4.120-92.70 (SLES12 SP2)
178 4.4.180-94.100 (SLES12 SP3)
179 4.4.0-131 (Ubuntu 16.04)
180 5.4.0-37 (Ubuntu 20.04)
181 5.4.0 (vanilla kernel.org)
182 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
183 * Recommended ZFS version: 2.0.0
184 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
185 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
186 stack overflow. For more information, please refer to bugzilla 17630.
187 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
188 the re-exporting nodes to be mounted with "32bitapi" mount option
190 --------------------------------------------------------------------------------
194 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
195 for currently supported client and server kernel versions.
196 * Server primary kernels built and tested during release cycle:
197 3.10.0-1062.1.1.el7 (RHEL7.7)
198 * Other server kernels known to build and work at some point (others may also work):
199 3.10.0-862.14.4.el7 (RHEL7.5)
200 3.10.0-957.27.2.el7 (RHEL7.6)
201 4.18.0-80.7.1.el8 (RHEL8.0)
202 4.4.120-92.70 (SLES12 SP2)
203 4.4.180-94.100 (SLES12 SP3)
204 4.4.0-131 (Ubuntu 16.04)
205 4.15.0-32 (Ubuntu 18.04)
206 vanilla linux 4.6.7 (ZFS only)
207 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
208 * Client primary unpatched kernels built and tested during release cycle:
209 3.10.0-1062.1.1.el7 (RHEL7.7)
210 4.12.14-95.29 (SLES12 SP4)
211 4.15.0-48 (Ubuntu 18.04)
212 * Other clients known to build on unpatched kernels at some point (others may also work):
213 3.10.0-862.14.4.el7 (RHEL7.5)
214 3.10.0-957.27.2.el7 (RHEL7.6)
215 4.14.0-49.13.1.el7a (RHEL7.5)
216 4.14.0-115.2.2.el7a (RHEL7.6)
217 4.18.0-80.7.1.el8 (RHEL8.0)
218 4.4.120-92.70 (SLES12 SP2)
219 4.4.180-94.100 (SLES12 SP3)
220 4.4.0-131 (Ubuntu 16.04)
221 vanilla linux 4.15.0 (ZFS only)
222 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
223 * Recommended ZFS / SPL version: 0.7.13
224 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
225 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
226 stack overflow. For more information, please refer to bugzilla 17630.
227 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
228 the re-exporting nodes to be mounted with "32bitapi" mount option
229 * Note for Ubuntu 14.04.5:
230 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
231 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
233 --------------------------------------------------------------------------------
237 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
238 for currently supported client and server kernel versions.
239 * Server known to build on the following kernels (others may also work):
240 2.6.32-696.18.7.el6 (RHEL6.9)
241 3.10.0-862.14.4.el7 (RHEL7.5)
242 3.10.0-957.el7 (RHEL7.6)
243 4.4.120-92.70 (SLES12 SP2)
244 4.4.162-94.69 (SLES12 SP3)
245 4.4.0-131 (Ubuntu 16.04)
246 4.15.0-32 (Ubuntu 18.04)
247 vanilla linux 4.6.7 (ZFS only)
248 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
249 * Client known to build on unpatched kernels (others may also work):
250 2.6.32-696.18.7.el6 (RHEL6.9)
251 2.6.32-754.6.3.el6 (RHEL6.10)
252 3.10.0-862.14.4.el7 (RHEL7.5)
253 3.10.0-957.el7 (RHEL7.6)
254 4.14.0-49.13.1.el7a (RHEL7.5)
255 4.14.0-115.2.2.el7a (RHEL7.6)
256 4.4.120-92.70 (SLES12 SP2)
257 4.4.162-94.69 (SLES12 SP3)
258 4.4.0-131 (Ubuntu 16.04)
259 4.15.0-32 (Ubuntu 18.04)
260 vanilla linux 4.15.0 (ZFS only)
261 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
262 * Recommended ZFS / SPL version: 0.7.9
263 * Tested with ZFS / SPL version: 0.7.9
264 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
265 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
266 stack overflow. For more information, please refer to bugzilla 17630.
267 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
268 the re-exporting nodes to be mounted with "32bitapi" mount option
269 * Note for Ubuntu 14.04.5:
270 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
271 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
273 --------------------------------------------------------------------------------
275 2018-04-03 Intel Corporation
277 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
278 for currently supported client and server kernel versions.
279 * Server known to build on patched kernels:
280 2.6.32-431.29.2.el6 (RHEL6.5)
281 2.6.32-504.30.3.el6 (RHEL6.6)
282 2.6.32-573.26.1.el6 (RHEL6.7)
283 2.6.32-642.15.1.el6 (RHEL6.8)
284 2.6.32-696.18.7.el6 (RHEL6.9)
285 3.10.0-693.21.1.el7 (RHEL7.4)
286 3.0.101-0.47.71 (SLES11 SP3)
287 3.0.101-107 (SLES11 SP4)
288 3.12.74-60.64.40 (SLES12 SP1)
289 4.4.114-92.64 (SLES12 SP2)
290 4.4.114-94.11 (SLES12 SP3)
291 3.13.0-101 (Ubuntu 14.04, ZFS only)
292 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
293 4.4.0-116 (Ubuntu 16.04)
294 vanilla linux 4.6.7 (ZFS only)
295 * Client known to build on unpatched kernels:
296 2.6.32-431.29.2.el6 (RHEL6.5)
297 2.6.32-504.30.3.el6 (RHEL6.6)
298 2.6.32-573.26.1.el6 (RHEL6.7)
299 2.6.32-642.15.1.el6 (RHEL6.8)
300 2.6.32-696.18.7.el6 (RHEL6.9)
301 3.10.0-693.21.1.el7 (RHEL7.4)
302 3.0.101-0.47.71 (SLES11 SP3)
303 3.0.101-107 (SLES11 SP4)
304 3.12.74-60.64.40 (SLES12 SP1)
305 4.4.114-92.64 (SLES12 SP2)
306 4.4.114-94.11 (SLES12 SP3)
307 3.13.0-101 (Ubuntu 14.04)
308 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
309 4.4.0-116 (Ubuntu 16.04)
311 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
312 * Recommended ZFS / SPL version: 0.7.6
313 * Tested with ZFS / SPL version: 0.7.6
314 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
315 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
316 stack overflow. For more information, please refer to bugzilla 17630.
317 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
318 the re-exporting nodes to be mounted with "32bitapi" mount option
319 * Note for Ubuntu 14.04.5:
320 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
321 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
323 --------------------------------------------------------------------------------
325 2017-07-05 Intel Corporation
327 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
328 for currently supported client and server kernel versions.
329 * Server known to build on patched kernels:
330 2.6.32-431.29.2.el6 (RHEL6.5)
331 2.6.32-504.30.3.el6 (RHEL6.6)
332 2.6.32-573.26.1.el6 (RHEL6.7)
333 2.6.32-642.15.1.el6 (RHEL6.8)
334 2.6.32-696.3.1.el6 (RHEL6.9)
335 3.10.0-514.21.1.el7 (RHEL7.3)
336 3.0.101-0.47.71 (SLES11 SP3)
337 3.0.101-100 (SLES11 SP4)
338 3.12.74-60.64.40 (SLES12 SP1)
339 4.4.59-92.17 (SLES12 SP2)
340 vanilla linux 4.6.7 (ZFS only)
341 * Client known to build on unpatched kernels:
342 2.6.32-431.29.2.el6 (RHEL6.5)
343 2.6.32-504.30.3.el6 (RHEL6.6)
344 2.6.32-573.26.1.el6 (RHEL6.7)
345 2.6.32-642.15.1.el6 (RHEL6.8)
346 2.6.32-696.3.1.el6 (RHEL6.9)
347 3.10.0-514.21.1.el7 (RHEL7.3)
348 3.0.101-0.47.71 (SLES11 SP3)
349 3.0.101-100 (SLES11 SP4)
350 3.12.74-60.64.40 (SLES12 SP1)
351 4.4.59-92.17 (SLES12 SP2)
353 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
354 * Recommended ZFS / SPL version: 0.6.5.9
355 * Tested with ZFS / SPL version: 0.6.5.9
356 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
357 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
358 stack overflow. For more information, please refer to bugzilla 17630.
359 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
360 the re-exporting nodes to be mounted with "32bitapi" mount option
362 --------------------------------------------------------------------------------
364 12-07-2016 Intel Corporation
366 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
367 for currently supported client and server kernel versions.
368 * Server known to build on patched kernels:
369 2.6.32-431.29.2.el6 (RHEL6.5)
370 2.6.32-504.30.3.el6 (RHEL6.6)
371 2.6.32-573.26.1.el6 (RHEL6.7)
372 2.6.32-642.11.1.el6 (RHEL6.8)
373 3.10.0-514.el7 (RHEL7.3)
374 3.0.101-0.47.71 (SLES11 SP3)
375 3.0.101-80 (SLES11 SP4)
376 3.12.62-60.64.8 (SLES12 SP1)
377 vanilla linux 4.5.7 (ZFS only)
378 * Client known to build on unpatched kernels:
379 2.6.32-431.29.2.el6 (RHEL6.5)
380 2.6.32-504.30.3.el6 (RHEL6.6)
381 2.6.32-573.26.1.el6 (RHEL6.7)
382 2.6.32-642.11.1.el6 (RHEL6.8)
383 3.10.0-514.el7 (RHEL7.3)
384 3.0.101-0.47.71 (SLES11 SP3)
385 3.0.101-80 (SLES11 SP4)
386 3.12.62-60.64.8 (SLES12 SP1)
388 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
389 * Recommended ZFS / SPL version: 0.6.5.8
390 * Tested with ZFS / SPL version: 0.6.5.8
391 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
392 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
393 stack overflow. For more information, please refer to bugzilla 17630.
394 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
395 the re-exporting nodes to be mounted with "32bitapi" mount option
397 --------------------------------------------------------------------------------
399 02-29-2016 Intel Corporation
401 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
402 for currently supported client and server kernel versions.
403 * Server known to build on patched kernels:
404 2.6.32-431.29.2.el6 (RHEL6.5)
405 2.6.32-504.30.3.el6 (RHEL6.6)
406 2.6.32-573.12.1.el6 (RHEL6.7)
407 3.10.0-327.3.1.el7 (RHEL7.2)
408 3.0.101-0.47.71 (SLES11 SP3)
409 3.0.101-68 (SLES11 SP4)
410 vanilla linux 4.2.1 (ZFS only)
411 * Client known to build on unpatched kernels:
412 2.6.32-431.29.2.el6 (RHEL6.5)
413 2.6.32-504.30.3.el6 (RHEL6.6)
414 2.6.32-573.12.1.el6 (RHEL6.7)
415 3.10.0-327.3.1.el7 (RHEL7.2)
416 3.0.101-0.47.71 (SLES11 SP3)
417 3.0.101-68 (SLES11 SP4)
420 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
421 * Recommended ZFS / SPL version: 0.6.4.2
422 * Tested with ZFS / SPL version: 0.6.4.2
423 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
424 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
425 stack overflow. For more information, please refer to bugzilla 17630.
426 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
427 the re-exporting nodes to be mounted with "32bitapi" mount option
429 --------------------------------------------------------------------------------
431 03-10-2015 Intel Corporation
433 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
434 for currently supported client and server kernel versions.
435 * Server known to build on patched kernels:
436 2.6.32-431.29.2.el6 (RHEL6.5)
437 2.6.32-504.8.1.el6 (RHEL6.6)
438 3.0.101-0.46 (SLES11 SP3)
439 * Client known to build on unpatched kernels:
440 2.6.32-431.29.2.el6 (RHEL6.5)
441 2.6.32-504.8.1.el6 (RHEL6.6)
442 3.10.0-123.20.1.el7 (RHEL7)
443 3.0.101-0.46 (SLES11 SP3)
444 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
445 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
446 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
447 stack overflow. For more information, please refer to bugzilla 17630.
448 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
449 the re-exporting nodes to be mounted with "32bitapi" mount option
451 Severity : enhancement
453 Description: control OST-index in IDIF via ROCOMPAT flag.
454 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
455 last_rcvd file. For new formatted OST device, it will be auto set;
456 for the case of upgrading from old OST device, you can enable it
457 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
458 enabled, for new created OST-object, its IDIF-in-LMA will contain
459 the OST-index; for the existing OST-object, the OSD will convert
460 old format IDIF as new format IDIF with OST-index stored in the
461 LMA EA when accessing such OST-object or via OI scrub. Once such
462 flag is enabled, it cannot be reverted back, so the system cannot
463 be downgraded to the orignal incompatible version.
465 --------------------------------------------------------------------------------
467 07-30-2014 Intel Corporation
469 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
470 for currently supported client and server kernel versions.
471 * Server known to build on patched kernels:
472 2.6.32-431.20.3.el6 (RHEL6.5)
473 3.0.101-0.31 (SLES11 SP3)
474 * Client known to build on unpatched kernels:
475 2.6.32-431.20.3.el6 (RHEL6.5)
476 3.0.101-0.31 (SLES11 SP3)
479 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
480 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
481 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
482 stack overflow. For more information, please refer to bugzilla 17630.
483 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
484 the re-exporting nodes to be mounted with "32bitapi" mount option
486 --------------------------------------------------------------------------------
488 10-31-2013 Intel Corporation
490 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
491 * Server builds for kernels:
492 2.6.32-358.18.1.el6 (RHEL6.4)
493 * Client builds for unpatched kernels:
494 2.6.32-358.18.1.el6 (RHEL6.4)
495 3.0.92-0.8 (SLES11 SP3)
497 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
498 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
499 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
500 stack overflow. For more information, please refer to bugzilla 17630.
501 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
502 the re-exporting nodes to be mounted with "32bitapi" mount option
504 --------------------------------------------------------------------------------
506 05-30-2013 Intel Corporation
508 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
509 * Server builds for kernels:
510 2.6.32-358.6.2.el6 (RHEL6.4)
511 * Client builds for unpatched kernels:
512 2.6.32-358.6.2.el6 (RHEL6.4)
513 3.0.74-0.6.6 (SLES11 SP2)
515 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
516 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
517 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
518 stack overflow. For more information, please refer to bugzilla 17630.
519 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
520 the re-exporting nodes to be mounted with "32bitapi" mount option
522 --------------------------------------------------------------------------------
524 9-30-2012 Whamcloud, Inc.
526 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
527 * Server support for kernels:
528 2.6.32-279.5.1.el6 (RHEL6)
529 * Client support for unpatched kernels:
530 2.6.32-279.5.1.el6 (RHEL6)
532 2.6.18-238.19.1.el5 (RHEL5)
533 2.6.16.60-0.69.1 (SLES10)
534 * Recommended e2fsprogs version: 1.41.90.wc4
535 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
536 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
537 stack overflow. For more information, please refer to bugzilla 17630.
538 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
539 the re-exporting nodes to be mounted with "32bitapi" mount option
541 Severity : enhancement
543 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
545 Severity : enhancement
547 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
549 Severity : enhancement
551 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
553 Severity : enhancement
555 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
557 Severity : enhancement
559 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
561 --------------------------------------------------------------------------------
563 04-30-2012 Whamcloud, Inc.
565 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
566 * Server support for kernels:
567 2.6.32-220.4.2.el6 (RHEL6)
568 * Client support for unpatched kernels:
569 2.6.32-220.4.2.el6 (RHEL6)
571 2.6.18-238.19.1.el5 (RHEL5)
572 2.6.16.60-0.69.1 (SLES10)
573 * Recommended e2fsprogs version: 1.41.90.wc4
574 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
575 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
576 stack overflow. For more information, please refer to bugzilla 17630.
577 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
578 the re-exporting nodes to be mounted with "32bitapi" mount option
580 Severity : enhancement
582 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
584 Severity : enhancement
586 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
588 Severity : enhancement
590 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
592 Severity : enhancement
594 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
596 --------------------------------------------------------------------------------
598 09-30-2011 Whamcloud, Inc.
600 * Server support for kernels:
601 2.6.18-238.19.1.el5 (RHEL5)
602 2.6.32-131.6.1.el6 (RHEL6)
603 * Client support for unpatched kernels:
604 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
605 2.6.18-238.19.1.el5 (RHEL5)
606 2.6.32-131.6.1.el6 (RHEL6)
607 2.6.29.4-167.fc11 (FC11)
608 2.6.16.60-0.69.1 (SLES10)
610 * Recommended e2fsprogs version: 1.41.90.wc3
611 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
612 Lustre filesystem with 4K stack may cause a stack overflow. For more
613 information, please refer to bugzilla 17630.
614 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
615 mounted with "32bitapi" mount option
617 Severity : enhancement
618 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
620 Severity : enhancement
621 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
623 Severity : enhancement
624 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
627 Frequency : only with SLES10
629 Description: Use OFED "KMP" provided by Novell
630 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
631 Lustre is now built against this package. That means you need to
632 install the ofed-kmp package from Novell for the patchless client.
633 Note that the ofed-kmp that Novell ships may not exactly match the
634 kernel version but should still be compatible.
636 Severity : enhancement
638 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
639 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
641 --------------------------------------------------------------------------------
643 2010-07-15 Oracle, Inc.
645 * Server support for kernels:
646 2.6.18-164.11.1.el5 (RHEL 5)
647 2.6.18-164.11.1.0.1.el5 (OEL 5)
648 * Client support for unpatched kernels:
649 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
650 2.6.18-164.11.1.el5 (RHEL 5),
651 2.6.18-164.11.1.0.1.el5 (OEL 5)
652 2.6.16.60-0.42.8 (SLES 10),
654 2.6.29.4-167.fc11 (FC11)
655 * Recommended e2fsprogs version: 1.41.10-sun2
657 Severity : enhancement
659 Description: Distribute lfs_migrate script with lustre
660 Details : The lfs_migrate mentioned in the manual is now included in the
661 lustre tree in lustre/scripts and will distributed in lustre RPMs.
663 Severity : enhancement
665 Description: Failover nid list added to import proc file.
667 Severity : enhancement
669 Description: "som_preview" mount option required to use SOM
670 Details : The clients will need to use som_preview mount option
674 Description: Lustre does not do 1MB IOs to HW RAID
675 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
677 Severity : enhancement
679 Description: lfs quota output is not very convenient for awk/sed-parsing
680 Details : Some positions in lfs quota output table could be empty or
681 non-empty which made it hard to parse it with scripts, now
682 a dash is put instead of space where there is not supposed
687 Description: Update to OFED 1.5.1
688 Details : OFED 1.5.1 will be used in environments that don't use
689 vendor supplied OFED.
693 Description: Remove lnet. prefix from lctl params display
694 Details : Also renamed lnet.memused to lnet_memused to
699 Description: Use conf_param -d to remove permanant settings.
703 Description: Add list_param -R
704 Details : lctl list_param now supports recursive mode.
707 Bugzilla : 21829, 22312
708 Description: Fix llobdstat.
709 Details : 1. Now updating the write_bytes counter
710 2. Port script fixes from 1.8.
714 Description: decrease the usage of memory on clients.
715 Details : 1. On clients, recycle dentries and inodes unused.
716 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
721 Description: llobdstat fix and enhancement
722 Details : add a counter to set a limit to how many samples will be returned
723 fix a wildcard in the path to limit to obdfilter stats only
727 Description: lfs check only allowed for root
728 Details : Code cleanups and added a regresion test of non-root use of
733 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
734 Details : Fix hr->hr_index overload.
738 Description: Fix FIEMAP functionality.
742 Description: FC11 patchless client support added
746 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
747 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
751 Description: New acc-sm script and various test-framework enhancments.
755 Description: Previously landed patch was reverted.
759 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
763 Description: Add ldiskfs maxdirsize mount option.
767 Description: Fix mkfs.lustre support for 16TB LUNs.
771 Description: Print mmp_check_interval
772 Details : print mmp_check_interval and make it possible to abort
773 mount operation in case it takes too long.
777 Description: Add cascading_rw to lustre mpi tests.
781 Description: Store SOM data in EA.
782 Details : Also several SOM related cleanups.
786 Description: Add refcount to md_open_data
790 Description: Use refcount on lov_request_set to prevent memory corruption.
794 Description: Statahead: use parent dir's i_mutex for synchronization.
798 Description: loadgen improvements.
802 Description: Enable peer health detection in ksocklnd and ko2iblnd.
806 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
810 Description: Fix race on obd_devcie:md_stats
814 Description: Fix SLES 11 client builds.
815 Details : Resolved uninitalized pointer gcc warnings.
819 Description: Allow non-root access for "lfs check".
820 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
824 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
826 Severity : enhancement
828 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
829 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
833 Description: Sanity test_27p fails.
834 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
835 Sanity tests 27 were cleaned up and improved.
839 Description: Unify req format on client/servers
840 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
841 buffers and swabbers approach..
846 Description: drop join file stuff.
848 Severity : enhancement
850 Description: Add 2.6.27 and SLES11 (patchless client) support
854 Description: Fix in ptlrpc_expire_one_request() to print the signed time
857 Severity : enhancement
859 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
862 Severity : enhancement
864 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
866 Severity : enhancement
868 Description: correctly handle big reply message.
869 Details : send LNet event if reply is bigger than buffer and adjust this buffer
874 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
876 Severity : enhancement
878 Description: Add support for OEL5.
882 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
884 Severity : enhancement
886 Description: "weak-modules" support
887 Details : Implement "weak-modules" support which enables kernel modules
888 to be used with any kernel that implements the same kABI. In
889 order to achieve this modules are now installed in
890 /lib/modules/$(uname -r)/updates/kernel on all distributions.
894 Description: client could not reconnect to OST because of an active request.
895 Details : The client now retries to reconnect to the same server, if a
896 connect request failed with EBUSY or -EAGAIN. Test case 35b
897 conf-sanity.sh has been created.
901 Description: Don't allow make backward step on assignin osc next id.
902 Details : race between allocation next id and ll_sync thread can be cause
903 of set wrong osc next id and can be kill valid ost objects.
907 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
911 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
915 Description: Error handling in osc_statfs_interpret() has been improved.
916 Details : Check in osc_statfs_interpret() for EBADR.
920 Description: Do not send statfs() requests to OSTs disabled by administrator.
921 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
923 Severity : enhancement
925 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
927 Severity : enhancement
929 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
931 Severity : enhancement
933 Description: Enhancement for lfs(1) command to use numeric uid/gid.
937 Description: Improved cleanup procedure for conf-sanity test_46a.
941 Description: Do not start bulk transfer, if client tried to reconnect.
942 Details : Do not start bulk transfer, if client tried to reconnect,
943 but failed due to active rpcs. Check for exp_abort_active_req
944 has been implemented.
948 Description: Prevent inconsistences between linux and lustre mount structures.
949 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
950 Make the sleep interruptible.
952 Severity : enhancement
954 Description: provide server to client comms path
955 Details : server to client communications path via new LDLM_SET_INFO rpc
959 Description: truncate starts GFP_FS allocation under transaction causing deadlock
960 Details : ldiskfs_truncate calls grab_cache_page which may start page
961 allocation under an open transaction. This may lead to
962 calling prune_icache with consequent lustre reentrance.
966 Description: send correctly lsm on open replay
967 Details : MDS is trust to LSM size on replay open, but client can set wrong size
972 Description: enable client interface failover
973 Details : When a child reconnects from another NID, properly update export
974 nid hash position and ldlm reverse import.
976 Severity : enhancement
978 Description: Communicate OST degraded/readonly state via statfs to MDS
979 Details : Flags in the statfs returned from OSTs indicate whether the
980 OST is in a degraded RAID state, or if the filesystem has
981 turned read-only after a filesystem error is detected.
986 Description: don't panic if EPROTO was hit when reading symlink
987 Details : correctly handling request reference in error cases.
989 Severity : enhancement
991 Description: Add LustreNetLink, a kernel-userspace communcation path.
993 Severity : enhancement
995 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1000 Description: access to llog context before init.
1001 Details : move handling CATALOGS file at osc layer and forbid access to llog
1002 context before init.
1005 Frequency : rare, only when client is writing to OST on local node
1007 Description: Avoid deadlock for local client writes
1008 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1009 memory freeing context. This allows OST threads to set the
1010 PF_MEMALLOC flag on task structures in order to allocate memory
1011 from reserved pools and complete IO.
1012 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1013 so that the OST threads generate memory pressure and allow
1014 inactive pages to be reclaimed.
1016 Severity : enhancement
1018 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1023 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1024 Details : move ldlm namespace creation in setup phase to avoid grab
1025 _lprocfs_lock with cli_sem held.
1029 Description: Temporarily disable grant shrink.
1030 Details : Disable the feature for debugging.
1033 Frequency : only when formatting test filesystems
1035 Description: Unable to run several mkfs.lustre on loop devices at the same time
1036 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1037 formatting. The solution is to proper handle the error.
1039 Severity : enhancement
1041 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1043 Severity : enhancement
1045 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1050 Description: resolve race between obd_disconnect and class_disconnect_exports
1051 Details : if obd_disconnect will be called to already disconnected export he
1052 forget release one reference and osc module can't unloaded.
1054 Severity : enhancement
1056 Description: Allow tuning service thread via /proc
1057 Details : For each service a new
1058 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1059 created that can be used to set min/max thread counts, and get the
1060 current number of running threads.
1065 Descriptoin: don't return error if have partially created objects for file.
1066 Details : lov_update_create_set uses set->set_success as index for created
1067 objects, so if some requests will be failed, they will have hole
1068 at end of array and qos_shrink_lsm for allocate correct lsm.
1070 Severity : enhancement
1072 Description: Update OFED support to 1.4.1
1077 Description: Reduce small size read RPC
1078 Details : Set read-ahead limit for every file and only do read-ahead when
1079 available read-ahead pages are bigger than 1M to avoid small size
1082 Severity : enhancement
1084 Description: move AT tunable parameters for more consistent usage
1085 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1087 Severity : enhancement
1089 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1090 Details : allow skip disconnected ost for send statfs request and hide error
1096 Description: fix racy locking of mballoc block bitmaps causing BUG
1097 Details : The locking of the mballoc buddy bitmap and the in-memory
1098 block bitmap was using two different spin locks in some
1099 cases. This made it possible to incorrectly access the
1100 mballoc bitmap while another process was modifying it,
1101 causing a sanity assertion to fail. While no on-disk corruption
1102 was reported, there was some risk of this happening.
1105 Frequency : rare, on llog test 6
1107 Descriptoin: don't allow connect to already connected import
1108 Details : allowing connect to already connected import is hide connecting problem.
1113 Descriptoin: allow kill process which wait statahead result
1114 Details : for some reasons 'ls' can stick in waiting result from statahead,
1115 in this case need way for kill this process.
1117 Severity : enhancement
1119 Description: Add state history info file, enhance import info file
1120 Details : Track import connection state changes in a new osc/mdc proc file;
1121 add overview-type data to the osc/mdc import proc file.
1123 Severity : enhancement
1125 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1127 Severity : enhancement
1129 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1134 Descriptoin: OST grant too much space to client even there are not enough space.
1135 Details : Client will shrink its grant cache to OST if there are no write
1136 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1137 this grant cache if there are already not enough avaible space
1138 (left_space < total_clients * 32M).
1141 Frequency : start MDS on uncleanly shutdowned MDS device
1143 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1144 Details : stay in waiting mds<>ost recovery finished produce random bugs
1145 due race between two ll_sync thread for one lov target. send
1146 ACTIVATE event only if connect realy finished and import have
1150 Frequency : rare, connect and disconnect target at same time
1152 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1153 Details : don't call obd_disconnect under lov_lock. this long time
1154 operation and can block ptlrpcd which answer to connect request.
1159 Descriptoin: don't lose wakeup for imp_recovery_waitq
1160 Details : recover_import_no_retry or invalidate_import and import_close can
1161 both sleep on imp_recovery_waitq, but we was send only one wakeup
1165 Frequency : always with long access acl
1167 Descriptoin: mds can't pack reply with long acl.
1168 Details : mds don't control size of acl but they limited by reint/getattr
1171 Severity : enhancement
1173 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1175 Severity : enhancement
1177 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1180 Frequency : start MDS on uncleanly shutdowned MDS device
1182 Descriptoin: aborting recovery hang on MDS
1183 Details : don't throttle destroy RPCs for the MDT.
1186 Frequency : on remount
1188 Description: external journal device not working after the remount
1189 Details : clear dev_rdonly flag for external journal devices in
1195 Description: shutdown vs evict race
1196 Details : client_disconnect_export vs connect request race.
1197 if client will evicted at this time - we start invalidate
1198 thread without referece to import and import can be freed
1202 Frequency : rare, need acl's on inode.
1204 Description: client can't handle ost additional correctly
1205 Details : if ost was added after client connected to mds client can have
1206 hit lnet_try_match_md ... to big messages to wide striped files.
1207 in this case need teach client to handle config events about add
1208 lov target and update client max ea size at that event.
1210 Severity : enhancement
1212 Description: Changelogs
1213 Details : Changelogs are a lightweight mechanism to track filesystem
1214 metadata and namespace changes. The changelog is recorded
1215 permanently on the MDTs, and is periodically "consumed" / purged
1216 when records are no longer needed.
1218 Severity : enhancement
1220 Description: compact fld format with extents
1221 Details : Store range of seq rather than every seq in FLD. Seq
1222 controller update FLD rather than clients. In Case of CMD, mdt0
1223 has FLD, all other metadata server act as non persistent proxy
1224 for FLD queries and cache fld entries in fld cache.
1229 Description: don't skip ost target if they assigned to file
1230 Details : Drop slow OSCs if we can, but not for requested start idx.
1231 This means "if OSC is slow and it is not the requested
1232 start OST, then it can be skipped, otherwise skip it only
1233 if it is inactive/recovering/out-of-space.
1237 Description: more cleanup in mds_lov
1238 Details : not send LOV EA under replay, we can't know about they size at this
1239 time. Don't allow client connect to mds before any ost connected,
1240 for avoid problems with LOV EA size and returning EIO to client.
1242 Severity : enhancement
1244 Description: Interoperability at server side (Disk interoperability)
1246 Severity : enhancement
1248 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1250 Severity : enhancement
1252 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1254 Severity : enhancement
1256 Description: New client IO stack (CLIO).
1258 Severity : enhancement
1260 Description: Commit on sharing. Eliminate inter-client dependencies between
1261 uncommitted transactions by doing transaction commits.
1262 Thereby clients may recovery independently.
1265 Frequency : Create a symlink file with a very long name
1267 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1268 Details : If there is no extra space in the request for early cancels,
1269 ldlm_req_handles_avail() returns 0 instead of a negative value.
1271 Severity : enhancement
1273 Description: Add /proc entry for import status
1274 Details : The mdc, osc, and mgc import directories now have
1275 an import directory that contains useful import data for debugging
1276 connection problems.
1278 Severity : enhancement
1280 Description: Re-disable certain /proc logging
1281 Details : Enable and disable client's offset_stats, extents_stats and
1282 extents_stats_per_process stats logging on the fly.
1285 Frequency : Only on FC kernels 2.6.22+
1287 Description: oops in statahead
1288 Details : Do not drop reference count for the dentry from VFS when lookup,
1289 VFS will do that by itself.
1291 Severity : enhancement
1293 Description: Generic /proc file permissions
1294 Details : Set /Proc file permissions in a more generic way to enable non-
1295 root users operate on some /proc files.
1299 Description: Hitting mdc_commit_close() ASSERTION
1300 Details : Properly handle request reference release in
1301 ll_release_openhandle().
1305 Description: quota recovery deadlock during mds failover
1306 Details : This patch includes att18982, att18236, att18237 in bz14840.
1308 1. fix osts hang when mds does failover with quotaon
1309 2. prevent watchdog storm when osts threads wait for the
1314 Frequency : only patchless client
1315 Description: add workaround for race between add/remove dentry from hash
1317 Severity : enhancement
1319 Description: Allow OST glimpses to return PW locks
1323 Description: LBUG when llog conf file is full
1324 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1328 Description: Prevent import from entering FULL state when server in recovery
1332 Description: service mount cannot take device name with ":"
1333 Details : Only when device name contains ":/" will mount treat it as
1339 Description: replace ptlrpcd with the statahead thread to interpret the async
1340 statahead RPC callback
1344 Frequency : on recovery
1345 Description: I/O failures after umount during fail back
1346 Details : if client reconnected to restarted server we need join to recovery
1347 instead of find server handler is changed and process self
1348 eviction with cancel all locks.
1350 Severity : enhancement
1352 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1354 Severity : enhancement
1356 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1358 Severity : enhancement
1360 Description: Upcall on Lustre log has been dumped
1361 Details : Allow for a user mode script to be called once a Lustre log has
1362 been dumped. It passes the filename of the dumped log to the
1363 script, the location of the script can be specified via
1364 /proc/sys/lnet/debug_log_upcall.
1369 Description: avoid idr_remove called for id which is not allocated.
1370 Details : Move assigment s_dev for clustered nfs to end of initialization,
1371 to avoid problem with error handling.
1376 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1377 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1378 into CONN_USED_HASH and this prodice warning when put connection
1379 again in unused hash.
1382 Severity : enhancement
1384 Description: File striping can now be set to use an arbitrary pool of OSTs.
1386 Severity : enhancement
1388 Description: Export bytes_read/bytes_write count on OSC/OST.
1392 Description: Early reply size mismatch, MGC loses connection
1393 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1394 the connect flags are properly negotiated.
1399 Description: quotas are not honored with O_DIRECT
1400 Details : all writes with the flag O_DIRECT will use grants which leads to
1401 this problem. Now using OBD_BRW_SYNC to guard this.
1405 Description: add quota statistics
1406 Details : 1. sort out quota proc entries and proc code.
1407 2. add quota statistics
1409 Severity : enhancement
1411 Description: enable quota support for HEAD.
1415 Description: Properly propagate oinfo flags from lov to osc for statfs
1416 Details : restore missing copy oi_flags to lov requests.
1418 Severity : enhancement
1420 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1421 : l_getgroups(8), lst(8), routerstat(8)
1422 Details : included man pages for llobdstat(8), llstat(8),
1423 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1425 Severity : enhancement
1427 Description: configure's --enable-quota should check the
1428 : kernel .config for CONFIG_QUOTA
1429 Details : configure is terminated if --enable-quota is passed but
1430 : no quota support is in kernel
1434 Description: Remove portals compatibility
1435 Details : Remove portals compatibility, not interoperable with releases
1440 Description: Resolve device initialization race
1441 Details : Prevent proc handler from accessing devices added to the
1442 obd_devs array but yet be intialized.
1444 Severity : enhancement
1446 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1448 Severity : enhancement
1450 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1455 Description: Using wrong pointer in osc_brw_prep_request
1456 Details : Access to array[-1] can produce panic if kernel compiled with
1457 CONFIG_PAGE_ALLOC enabled
1461 Description: Client runs out of low memory
1462 Details : Consider only lowmem when counting initial number of llap pages
1466 Description: *optional* service tags registration
1467 Details : if the "service tags" package is installed on a Lustre node
1468 When the filesystem is mounted, a local-node service tag will
1469 be created. See http://inventory.sun.com/ for more information
1470 about the Service Tags asset management system.
1474 Description: Kernel BUG tries to release flock
1475 Details : Lustre does not destroy flock lock before last reference goes
1476 away. So always drop flock locks when client is evicted and
1477 perform unlock regardless of successfulness of speaking to MDS.
1481 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1485 Description: sanity test 65a fails if stripecount of -1 is set
1486 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1491 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1492 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1495 Severity : enhancement
1497 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1501 Description: do not process already freed flock
1502 Details : flock can possibly be freed by another thread before it reaches
1503 to ldlm_flock_completion_ast.
1507 Description: LBUG during stress test
1508 Details : Need properly lock accesses the flock deadlock detection list.
1512 Description: oops in page fault handler
1513 Details : kernel page fault handler can return two special 'pages' in error
1514 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1518 Description: timeout with invalidate import.
1519 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1520 should be handled by ptlrpcd. This produce long age waiting and
1521 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1523 Severity : enhancement
1525 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1529 Description: enable MGS and MDT services start separately
1530 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1531 a co-located MGS without starting the MGS, which is a complement
1532 to 'nosvc' mount option.
1536 Frequency : after recovery
1537 Description: precreate to many object's after del orphan.
1538 Details : del orphan st in oscc last_id == next_id and this triger growing
1539 count of precreated objects. Set flag LOW to skip increase count
1540 of precreated objects.
1544 Frequency : rare, on clear nid stats
1545 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1546 Details : when clean nid stats sometimes try destroy live entry,
1547 and this produce panic in free.
1551 Description: Stack overflow during MDS log replay
1552 ease stack pressure by using a thread dealing llog_process.
1556 Description: wait until IO finished before start new when do lock cancel.
1557 Details : VM protocol want old IO finished before start new, in this case
1558 need wait until PG_writeback is cleared until check dirty flag and
1559 call writepages in lock cancel callback.
1561 Severity : enhancement
1563 Description: using special macro for print time and cleanup in includes.
1567 Description: mds_mfd_close() ASSERTION(rc == 0)
1568 Details : In mds_mfd_close(), we need protect inode's writecount change
1569 within its orphan write semaphore to prevent possible races.
1573 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1577 Frequency : rare, on shutdown ost
1578 Description: don't hit live lock with umount ost.
1579 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1580 use shrink_dcache_sb instead.
1584 Description: don't panic with use echo client
1585 Details : echo client pass NULL as client nid pointer and this produce null
1586 pointer dereference.
1590 Description: fix build on ppc32
1591 Details : compile code with -m64 flag produce wrong object file for ppc32.
1595 Description: add message levels for liblustreapi
1599 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1600 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1601 will result in an -EAGAIN and in the exit call path we call
1602 alloc_rr() which will with increasing aggressiveness attempt to
1603 aquire precreated objects on the minimum number of required OSCs.
1607 Description: Use old size assignment to avoid deadlock
1608 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1609 the scheduling while holding a spinlock. We do not need locking
1610 for size in ll_update_inode() because size is only updated from
1611 the MDS for directories or files without objects, so there is no
1612 other place to do the update, and concurrent access to such inodes
1613 are protected by the inode lock.
1617 Description: resolve "_IOWR redefined" build error on SLES10
1621 Description: dump the memory debugging after all modules are unloaded to
1622 suppress false negative in conf_sanity test 39
1624 Severity : enhancement
1626 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1631 Description: fix SLES kernel versioning
1632 Details : the kernel version for our SLES 10 kernel did not include a "-"
1633 before the "smp" at the end. while this was not a problem in
1634 general, it did mean that software trying to use the kernel
1635 version to try to detect a vendor specific kernel would fail.
1636 this was most evident by the OFED build scripts.
1640 Description: Don't update lov_desc members until making sure they are valid
1641 Details : When updating lov_desc members via proc fs, need fix their
1642 validities before doing the real update.
1646 Description: don't put request into delay list while invalidate in flight.
1647 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1648 import in flight. this produce timeout for invalidate and sometimes
1649 can cause stale data.
1653 Frequency : on ppc only
1654 Description: not convert ost objects for directory because it's not exist.
1655 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1659 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1661 Severity : enhancement
1663 Description: Update to support 2.6.22.14 vanilla kernel.
1667 Frequency : rare, on recovery
1668 Description: read procfs can produce deadlock in some situation
1669 Details : Holding lprocfs lock which send rpc can produce block for destroy
1670 obd objects and this also block reconnect with -EALREADY. This
1671 isn't fix all lprocfs bugs - but make it rare.
1673 Severity : enhancement
1675 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1678 Frequency : frequent on X2 node
1680 Description: mdc_set_open_replay_data LBUG
1681 Details : Set replay data for requests that are eligible for replay.
1685 Description: lustre_mgs: operation 101 on unconnected MGS
1686 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1687 MGC, and late on MGC cannot successfully connect to MGS and a lot
1688 of the error messages complaining that MGS is not connected.
1692 Frequency : on network error
1693 Description: panic with double free request if network error
1694 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1695 but it's true only for synchronus enqueue, for async enqueue
1696 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1699 Severity : enhancement
1701 Description: client-side metadata stat-ahead during readdir(directory readahead)
1702 Details : perform client-side metadata stat-ahead when the client detects
1703 readdir and sequential stat of dir entries therein
1706 Frequency : on start mds
1708 Description: Implement get_info(last_id) in obdfilter.
1711 Frequency : occasional
1713 Description: Correctly check stale fid, not start epoch if ost not support SOM
1714 Details : open with flag O_CREATE need set old fid in op_fid3 because
1715 op_fid2 was overwritten with new generated fid, but mds can answer
1716 with one of these two fids and both is not stale. Setattr
1717 incorrectly started an epoch and assume will be called
1718 done_writing, but without SOM done_writing ever being called.
1721 Frequency : rare, depends on device drivers and load
1723 Description: MDS or OSS nodes crash due to stack overflow
1724 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1725 In some cases, in conjunction with device drivers that use a lot
1726 of stack the MDS (or possibly OSS) service threads could overflow
1727 the stack. One change which was identified to consume additional
1728 stack has been reworked to avoid the extra stack usage.
1731 Frequency : occasional
1733 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1734 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1735 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1736 oscc_has_objects and see OSC has no objests and call
1737 oscc_internal_create to resend create request.
1739 Severity : enhancement
1741 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1743 Severity : enhancement
1745 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1748 Frequency : very rare
1750 Description: Fix replay if there is an un-replied request and open
1751 Details : In some cases, older replay request will revert the
1752 mcd->mcd_last_xid on MDS which is used to record the client's
1753 latest sent request.
1755 Severity : enhancement
1757 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1759 Severity : enhancement
1761 Description: Add rhel5 support to HEAD.
1763 Serverity : enhancement
1765 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1770 Description: Don't allow skipping OSTs if index has been specified.
1771 Details : Don't allow skipping OSTs if index has been specified, make
1772 locking in internal create lots better.
1776 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1777 Details : access to bitfield in structure is always rounded to long
1778 and this produce problem with not atomic change any bit.
1782 Description: Lustre make rpms failed.
1783 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1784 builds Lustre rpms from tarball.
1787 Frequency : rare on shutdown ost
1789 Description: If llog cancel was not send before clean_exports phase, this can
1790 produce deadlock in llog code.
1791 Details : If llog thread has last reference to obd and call class_import_put
1792 this produce deadlock because llog_cleanup_commit_master wait when
1793 last llog_commit_thread exited, but this never success because was
1794 called from llog_commit_thread.
1798 Description: allow userland application know is lost one of stripes.
1799 Details : fill lvb_blocks with error code on ost and return it to
1800 application if error flag found.
1804 Description: NULL lov_tgts causing MDS oops
1805 Details : more safe checks for NULL lov_tgts for avoid oops.
1807 Severity : enhancement
1809 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1813 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1814 Details : If an OST is down the MDS will hang indefinitely in
1815 obd_statfs_rqset() waiting for the statfs data. While for
1816 MDS QOS usage of statfs, it should not stuck in waiting.
1818 Severity : enhancement
1820 Description: remote_acl support
1821 Details : Support ACL-based permission check for remote user.
1822 Support setfacl/getfacl for remote user with the utils
1823 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1824 the system "{s,g}etfacl" utils.
1826 Severity : enhancement
1828 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1830 Severity : enhancement
1832 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1836 Description: make mgs_setparam() handle fsname containing dash
1837 Details : fsname containing a dash does not work with lctl conf_param
1839 Severity : enhancement
1841 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1843 Severity : enhancement
1845 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1849 Frequency : rare, at shutdown
1850 Description: access already free / zero obd_namespace.
1851 Details : if client_disconnect_export was called without force flag set,
1852 and exist connect request in flight, this can produce access to
1853 NULL pointer (or already free pointer) when connect_interpret
1854 store ocd flags in obd_namespace.
1858 Frequency : only at startup
1859 Description: not alloc memory with spinlock held.
1860 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1861 if any spinlock held.
1863 Severity : enhancement
1865 Description: make lustre randomly fail allocating memory
1866 Details : Make lustre randomly failed allocating memory for testing purpose.
1868 Severity : enhancement
1870 Description: lost problems with lov objid file
1871 Details : Fixes some scability and access to not inited memory problems
1872 in work with lov objdid file.
1877 Description: lfs find does not continue on file error
1878 Details : Continue other files processing when a file/dir is absent.
1882 Description: Inconsistent usage of lustre_pack_reply()
1883 Details : Standardize the usage of lustre_pack_reply() such that it
1884 always generate a CERROR on failure.
1887 Frequency : occasional
1889 Description: MDS hang or stay in waiting lock
1890 Details : If client receive lock with CBPENDING flag ldlm need send lock
1891 cancel as separate rpc, to avoid situation when cancel request
1892 can't processed due all i/o threads stay in wait lock.
1896 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1900 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1904 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1908 Description: Update to RHEL4 latest kernel.
1910 Severity : enhancement
1912 Description: Build SLES10 patchless client fails
1913 Details : The configure was broken by run ./configure with
1914 --with-linux-obj=.... argument for patchless client. When the
1915 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1916 can't search header adequately. Use absolute path such as
1917 -I($LINUX)/include instead.
1921 Description: interrupt oig_wait produce painc on resend.
1922 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1923 private set, and this produce situation when rq_ptlrpcd_data not
1924 copyed to new allocated request and triggered LBUG on assert
1925 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1926 ptlrpcd set if request is changed and can be safety changed to use
1931 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1932 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1933 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1934 these macros and replace them with their difinitions instead.
1938 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1939 Details : Update which_patch & target file for SLES10 latest kernel.
1942 Description: Remove client patches from SLES 10 kernel.
1943 Details : This causes SLES 10 clients to behave as patchless clients
1944 even on a Lustre-patched (server) kernel.
1946 Severity : enhancement
1948 Description: self-adjustable client's lru lists
1949 Details : use adaptive algorithm for managing client cached locks lru
1950 lists according to current server load, other client's work
1951 pattern, memory activities, etc. Both, server and client
1952 side namespaces provide number of proc tunables for controlling
1955 Severity : enhancement
1957 Description: light-weight GSS support
1958 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1959 and reduce performance overhead.
1961 Severity : enhancement
1963 Description: Linux keyring support
1964 Details : Support using service of Linux keyring for Lustre GSS internal
1965 context refresh/cache mechanism.
1969 Description: Fix errors in lfs documentation
1970 Details : Fixes man pages
1974 Description: don't use GFP_* in generic Lustre code.
1975 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1979 Description: obdclass is limited by single OBD_ALLOC(idarray)
1980 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1984 Description: warnings with build patchless client with vanila 2.6.19 and up
1985 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1986 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1990 Frequency : Only for SLES
1991 Description: sanity_quota fail test_1
1992 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1993 file, which makes TSTID[2] inunique.
1997 Frequency : Only for SLES9
1998 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1999 Details : Update target/ChangeLog/which_patch .
2003 Frequency : Only for test_10 in sanity-quota.sh
2004 Description: a bug in quota test code
2005 Details : checking a test flag in a wrong place
2009 Frequency : Only for test_13 in sanity-quota.sh
2010 Description: a bug in quota test script
2011 Details : shouldn't check overfull allocated quota
2015 Description: jbd statistics
2016 Details : Port older jbd statistics patch for sles10
2018 Severity : enhancement
2020 Description: Add printing inode info into message about error in writepage.
2025 Description: reply_lock_interpret crash due to race with it and lock cancel.
2026 Details : Do not replay locks that are being cancelled. Do not reference
2027 locks by their address during replay, just by their handle.
2031 Frequency : When flocks are used.
2032 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2033 Details : Flock locks might destroy just granted lock if it could be merged
2034 with another existing flock, this is done in completion handler,
2035 so teach ldlm_cli_enquque_fini that this is a valid case for
2041 Description: Oops in read and write path when failing to allocate lock.
2042 Details : Check if lock allocation failed and return error back.
2046 Description: lstripe command fails for valid OST index
2047 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2048 instead of lov->desc.ld_active_tgt_count.
2052 Description: sanity.sh failed test 103
2053 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2058 Description: df doesn't work properly if diskfs blocksize != 4K
2059 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2063 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2064 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2066 Severity : enhancement
2068 Description: Adaptive timeouts
2069 Details : RPC timeouts adapt to changing server load and network
2070 conditions to reduce resend attempts and improve recovery time.
2074 Description: llapi_file_create() does not allow some changes
2075 Details : add llapi_file_open() that allows specifying the mode and
2076 open flags, and also returns an open file handle.
2080 Description: merge and cleanup kernel patches.
2081 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2085 Description: Add journal checksum support.(Kernel part)
2086 Details : The journal checksum feature adds two new flags i.e
2087 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2088 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2089 indicates that the commit block contains the checksum for
2090 the blocks described by the descriptor blocks. Now commit
2091 record can be sent to disk without waiting for descriptor
2092 blocks to be written to disk. This behavior is controlled
2093 using JBD2_FEATURE_ASYNC_COMMIT flag.
2097 Description: OSS needs mutliple precreate threads
2098 Details : Add ability to start more than one create thread per OSS.
2102 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2103 Details : execute lfs setstripe on client
2107 Description: mds_obd_create error creating tmp object
2108 Details : When the user sets quota on root, llog will be affected and can't
2109 create files and write files.
2112 Frequency : Always on ia64 patchless client, and possibly others.
2114 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2115 Details : This allows the patchless client to be loaded on architectures
2116 without this export.
2120 Description: RedHat Update kernel for RHEL5
2121 Details : Add the kernel config file for RHEL5.
2125 Description: RedHat Update kernel for RHEL5
2126 Details : Modify the kernel config file more closer RHEL5.
2130 Description: Build failure against Centos5 (RHEL5)
2131 Details : Define PAGE_SIZE when it isn't present.
2135 Description: test_7 on sanity-quota.sh can't be used on separate machines
2136 Details : add facet to handle it
2140 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2141 Details : with NFS, the anon dentry's parent was set to itself in
2142 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2143 corresponding dentry other than use rec->ur_name.
2145 Severity : enhancement
2147 Description: lfs setstripe enhancement
2148 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2152 Description: enable data checksumming by default
2153 Details : enable checksum by default, allow --disable-checksum
2154 configure option and "-o nochecksum" mount option. Checksums
2155 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2159 Description: lustre support for RHEL5
2160 Details : Add support for RHEL5.
2164 Description: Client eviction due to failover config
2165 Details : after a connection loss, the lustre client should attempt to
2166 reconnect to the last active server first before trying the
2167 other potential connections.
2171 Description: when mds and osts use different quota unit(32bit and 64bit),
2172 quota will be released repeatly.
2173 Details : void sending multiple quota reqs to mds, which will keep the
2174 status between the reqs.
2178 Description: osts not allocated evenly to files
2179 Details : change the condition to increase offset_idx
2182 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2184 Description: Data corruption for OSTs that are formatted larger than 2TB
2186 Details : When generating the bio request for lustre file writes the
2187 sector number would overflow a temporary variable before being
2188 used for the IO. The data reads correctly from Lustre (which
2189 will overflow in a similar manner) but other file data or
2190 filesystem metadata may be corrupted in some cases.
2194 Description: Tune the kernel for good SCSI performance.
2195 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2196 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2201 Description: rewrite ext2-derived code in obdclass/uuid.c
2202 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2203 from scratch preserving functionality.
2207 Description: "lfs find -obd UUID" prints directories
2208 Details : "lfs find -obd UUID" will return all directory names instead
2209 of just file names. It is incorrect because the directories
2210 do not reside on the OSTs.
2214 Description: use i_size_read and i_size_write in 2.6 port
2215 Details : replace inode->i_size access with i_size_read/write()
2218 Frequency : when removing large files
2220 Description: scheduling issue during removal of large Lustre files
2221 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2222 It causes scheduling issues when removing large files (17TB in the
2226 Frequency : only with liblustre clients on XT3
2228 Description: evictions taking too long
2229 Details : allow llrd to evict clients directly on OSTs
2232 Frequency : only on ppc
2234 Description: /proc/fs/lustre/devices broken on ppc
2235 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2236 We should make sure the type of 'index' is loff_t and then cast
2237 explicitly as needed below. Do not assign an explicitly cast
2241 Frequency : only for rhel5
2243 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2244 Details : Modify the target file & which_kernel.
2246 Severity : enhancement
2248 Description: omit set fsid for export NFS
2249 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2250 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2254 Frequency : Always, for kernels after 2.6.16
2255 Description: Fix warning idr_remove called for id=.. which is not allocated.
2256 Details : Last kernels save old s_dev before kill super and not allow
2257 to restore from callback - restore it before call kill_anon_super.
2261 Description: buffer overruns could theoretically occur
2262 Details : llapi_semantic_traverse() modifies the "path" argument by
2263 appending values to the end of the origin string, and a
2264 overrun may occur. Adding buffer overrun check in liblustreapi.
2268 Description: Fix error on 'ls .' at the top of the Lustre mount.
2269 Details : Don't revalidate dentry if it is a root dentry.
2273 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2274 Details : Modify vm-tunables-rhel4.patch.
2278 Description: Kernel config for 2.6.18-vanilla.
2279 Details : Modify targets/2.6-vanilla.target.in.
2280 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2281 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2282 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2283 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2287 Description: improve handling recoverable errors
2288 Details : if request processig with error which can be recoverable on server
2289 request should be resend, otherwise page released from cache and
2294 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2295 Details : Modify target file & which_patch.
2296 A flaw was found in the IA32 system call emulation provided
2297 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2298 value could be stored in the %RAX register, which could trigger an
2299 out-of-bounds system call table access. An untrusted local user
2300 could exploit this flaw to run code in the kernel
2301 (ie a root privilege escalation). (CVE-2007-4573).
2305 Description: O_DIRECT bypasses client statistics.
2306 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2307 read_ahead_stats were updated. copied the stats section from
2308 osc_send_oap_rpc() into async_internal().
2312 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2316 Description: change order of libsysio includes
2317 Details : '#include sysio.h' should always come before '#include xtio.h'
2321 Description: handle "serious error: objid * already exists" more gracefully
2322 Details : If LAST_ID value on disk is smaller than the objects existing in
2323 the O/0/d* directories, it indicates disk corruption and causes an
2324 LBUG(). If the object is 0-length, then we should use the existing
2325 object. This will help to avoid a full fsck in most cases.
2327 Severity : enhancement
2329 Description: adapt the lustre_config script to support the upgrade case
2330 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2331 server targets to 1.6 in parallel.
2335 Description: To avoid grant space > avaible space when the disk is almost
2336 full. Without this patch you might see the error "grant XXXX >
2337 available" or some LBUG about grant, when the disk is almost
2339 Details : In filter_check_grant, for non_grant cache write, we should
2340 check the left space by if (*left > ungranted + bytes), instead
2341 of (*left > ungranted), because only we are sure the left space
2342 is enough for another "bytes", then the ungrant space should be
2343 increase. In client, we should update cl_avail_grant only there
2344 is OBD_MD_FLGRANT in the reply.
2348 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2349 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2350 described in CVE-2007-4573. This problem could allow untrusted
2351 local users to gain root access.
2354 Frequency : when using O_DIRECT and quotas
2356 Description: Incorrect file ownership on O_DIRECT output files
2357 Details : block usage reported by 'lfs quota' does not take into account
2358 files that have been written with O_DIRECT.
2363 Description: touch file failed when fs is not full
2364 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2365 otherwise we can get ENOSP while fs is not full.
2369 Description: parallel lock callbacks
2370 Details : Instead of sending blocking and completion callbacks as separated
2371 requests, adding them to a set and sending in parallel.
2374 Frequency : only for Cray XT3
2375 Bugzilla : 12829/13455
2376 Description: Changing primary group doesn't change the group lustre assigns to
2378 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2379 overridden with the primary group provided by the group upcall,
2380 whereas the supplied fsgid can be trusted if it is in the list of
2381 supplementary groups returned by the group upcall.
2383 Severity : enhancement
2385 Description: Allow masking D_WARNING, D_ERROR messages from console
2386 Details : Console messages can now be disabled via lnet.printk.
2390 Description: User code with malformed file open parameter crashes client node
2391 Details : Before packing join_file req, all the related reference should be
2392 checked carefully in case some malformed flags cause fake
2393 join_file req on client.
2397 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2398 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2399 mode is equal to the granted mode. However, it should also check
2400 that LDLM_FL_CP_REQD is not set.
2404 Description: Heavy nfs access might result in deadlocks
2405 Details : After ELC code landed, it is now improper to enqueue any mds
2406 locks under och_sem, because enqueue might want to decide to
2407 cancel open locks for same inode we are holding och_sem for.
2411 Description: Client eviction while running blogbench
2412 Details : A lot of unlink operations with concurrent I/O can lead to a
2413 deadlock causing evictions. To address the problem, the number of
2414 oustanding OST_DESTROY requests is now throttled to
2415 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2416 callbacks are processed in priority.
2420 Description: enable ACLs on MDS by default
2421 Details : ACLs must be enabled on MDS by default.
2424 Frequency : PPC/PPC64 only
2426 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2427 Details : fix duplicated definitions between asm-ppc64/types.h and
2428 lustre_types.h on PPC.
2431 Frequency : PPC/PPC64 only
2433 Description: asm-ppc/segment.h does not exist
2434 Details : fix compile issue on PPC.
2438 Description: better handle error messages in extents code
2441 Frequency : RHEL4 only
2443 Description: mkfs is very slow on IA64/RHEL4
2444 Details : A performance regression has been discovered in the MPT Fusion
2445 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2446 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2447 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2449 Severity : enhancement
2451 Description: SNMP support enhancement
2452 Details : Adding total number of sampled request for an MDS node in snmp
2455 Severity : enhancement
2457 Description: Optimize ldlm waiting list processing for PR extent locks
2458 Details : When processing waiting list for read extent lock and meeting read
2459 lock that is same or wider to it that is not contended, skip
2460 processing rest of the list and immediatelly return current
2461 status of conflictness, since we are guaranteed there are no
2462 conflicting locks in the rest of the list.
2466 Description: Time out and refuse to reconnect
2467 Details : When the failover node is the primary node, it is possible
2468 to have two identical connections in imp_conn_list. We must
2469 compare not conn's pointers but NIDs, otherwise we can defeat
2470 connection throttling.
2474 Description: port llog fixes from b1_6 into HEAD
2475 Details : Port llog reference couting and some llog cleanups from b1_6
2476 (bug 10800) into HEAD, for protect from panic and access to already
2477 free llog structures.
2481 Description: Detect stride IO mode in read-ahead
2482 Details : When a client does stride read, read-ahead should detect that and
2483 read-ahead pages according to the detected stride pattern.
2487 Description: data checksumming impacts single node performance
2488 Details : add support for several checksum algorithm. Currently, only CRC32
2489 and Adler-32 are supported. The checksum type can be changed on
2490 the fly via /proc/fs/lustre/osc/*/checksum_type.
2494 Description: use adler32 for page checksums
2495 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2500 Description: build for x2 fails
2501 Details : fix compile issue on Cray systems.
2505 Description: Properly match for duplicate locks
2506 Details : Due to different lock order from skiplists code, we need to
2507 traverse entire list for now
2510 Frequency : only on PPC/SLES10
2512 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2513 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2514 be defined. Add a hack in mkfs_lustre.c to work around this compile
2519 Description: LASSERT on MDS when client holding flock lock dies
2520 Details : ldlm pool logic depends on number of granted locks equal to
2521 number of released locks which is not true for flock locks, so
2522 just exclude such locks from consideration.
2526 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2527 Details : Use fsfilt_sync() for both the whole filesystem sync and
2528 individual file sync to eliminate dangerous inode locking
2529 with I_LOCK that can lead to a deadlock.
2533 Description: performance in 1.6.3
2534 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2535 sure that 1MB requests are not fragmented by the block layer.
2537 Severity : enhancement
2539 Description: organize the server-side client stats on per-nid basis
2540 Details : Change the structure of stats under obdfilter and mds to
2548 The "uuid"s file would list the uuids of _active_ exports.
2549 And the clear entry is to clear all stats and stale nids.
2551 Severity : enhancement
2553 Description: eliminate client locks in face of contention
2554 Details : file contention detection and lockless i/o implementation
2555 for contended files.
2559 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2562 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2564 Description: MDS may lose file striping (and hence file data) in some cases
2565 Details : If there are additional extended attributes stored on the MDS,
2566 in particular ACLs, SELinux, or user attributes (if user_xattr
2567 is specified for the client mount options) then there is a risk
2568 of attribute loss. Additionally, the Lustre file striping
2569 needs to be larger than default (e.g. striped over all OSTs),
2570 and an additional attribute must be stored initially in the
2571 inode and then increase in size enough to be moved to the
2572 external attribute block (e.g. ACL growing in size) for file
2577 Description: skiplist implementation simplification
2578 Details : skiplists are used to group compatible locks on granted list
2579 that was implemented as tracking first and last lock of each
2580 lock group the patch changes that to using doubly linked lists
2584 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2585 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2586 because LASSERT on both the data supplied by a client, and the
2587 data on disk is dangerous and incorrect.
2589 Severity : enhancement
2591 Description: Slow truncate/writes to huge files at high offsets.
2592 Details : Directly associate cached pages to lock that protect those pages,
2593 this allows us to quickly find what pages to write and remove
2594 once lock callback is received.
2598 Description: more ldlm soft lockups
2599 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2600 starve other threads from the resource lock for a long time in
2601 case of long waiting queue, so change the debug level from
2602 D_OTHER to the less frequently used D_INFO.
2604 Severity : enhancement
2606 Description: add -gid, -group, -uid, -user options to lfs find
2610 Description: Hung threads in invalidate_inode_pages2_range
2611 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2612 once one is completed. As a result, some RPCs are stuck in the
2613 queue and are never sent.
2617 Description: filter threads hungs on waiting journal commit
2618 Details : Cleanup filter group llog code, then only filter group llog will
2619 be only created in the MDS/OST syncing process.
2623 Description: Procfs and llog threads access destoryed import sometimes.
2624 Details : Sync the import destoryed process with procfs and llog threads by
2625 the import refcount and semaphore.
2627 Severity : enhancement
2629 Description: openlock cache of b1_6 port to HEAD
2634 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2635 Details : If insertion of an extent fails, then discard the inode
2636 preallocation and free data blocks else it can lead to duplicate
2641 Description: don't always update ctime in ext3_xattr_set_handle()
2642 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2643 In some cases the ctime should not be updated, for example for
2644 2.0->1.8 compatibility it is necessary to delete an xattr and it
2645 should not update the ctime.
2649 Bugzilla : 15713/16362
2650 Description: Assertion in iopen_connect_dentry in 1.6.3
2651 Details : looking up an inode via iopen with the wrong generation number can
2652 populate the dcache with a disconneced dentry while the inode
2653 number is in the process of being reallocated. This causes an
2654 assertion failure in iopen since the inode's dentry list contains
2655 both a connected and disconnected dentry.
2659 Description: assertion failure in ldlm_handle2lock()
2660 Details : fix a race between class_handle_unhash() and class_handle2object()
2661 introduced in lustre 1.6.5 by bug 13622.
2666 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2667 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2672 Description: oops in sd_iostats_seq_show()
2673 Details : unloading/reloading the scsi low level driver triggers a kernel
2674 bug when trying to access the sd iostat file.
2679 Description: Kernel panics during QLogic driver reload
2680 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2681 patch, causing memory corruption.
2686 Description: journal_dev option does not work in b1_6
2687 Details : pass mount option during pre-mount.
2689 Severity : enhancement
2691 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2692 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2693 extent information of a file. It can be used to map logical blocks
2694 in a file to physical blocks in the block device.
2698 Description: LDLM soft lockups - improvement
2699 Details : It is be possible to send the lock handle along with each read
2700 or write request because the client is already doing a lock match
2701 itself so there isn't any reason the OST should have to re-do that
2705 Frequency : only on Cray X2
2707 Description: X2 build failures
2708 Details : fix build failures on Cray X2.
2712 Description: xid & resent requests
2713 Details : Initialize RPC XID from clock at startup (randomly if clock is
2716 Severity : enhancement
2718 Description: Add lustre_start utility to start or stop multiple Lustre servers
2723 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2724 Details : In case of memory pressure, list_del() can be called twice on
2725 req->rq_history_list, causing a kernel oops.
2729 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2730 Details : fix stack overflow in the distributed lock manager by defering
2731 export eviction after a failed AST to the elt thread instead of
2732 handling it in the dlm interpret routine.
2736 Description: Convert some comments to new format.
2737 Details : Update documenting comments to match doxygen conventions.
2741 Description: Grammar fixes.
2742 Details : A couple of trivial spelling fixes.
2746 Description: OSD_COUNTERS-mandatory
2747 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2748 to simplify the code.
2752 Description: simplify cmm_device freeing logic.
2753 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2757 Description: Add lockdep support to dt_object_operations locking interface.
2758 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2759 parameter indicating lock ordering. Update mdd code to use new
2764 Description: Introduce failloc constants for lockless IO tests.
2765 Details : Add two new failloc constants to test lockless IO. Only one of
2766 them in implemented---another is checked in yet to be landed
2771 Description: Add lockdep support for inode mutex.
2772 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2773 in the situations where Lustre has to lock more than one inode
2778 Description: Add optional invariants checking support.
2779 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2780 configure switch. This macro is to be used for consistency and
2781 sanity checks that are too expensive to be left in `production'
2786 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2787 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2792 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2793 Details : Use spin_lock_nested() in (the only) situation where more than
2794 one ldlm_lock is locked simultaneously. Also, fix possible
2795 dead-lock in ldlm_lock_change_resource() by enforcing particular
2800 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2801 Details : Instead of specifying each ldlm_lock call-back through separate
2802 parameter, wrap them into struct ldlm_callback_suite.
2806 Description: Kill join_lru obd method and its callers.
2807 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2808 in memory---a requirement implemented through ->o_join_lru() obd
2809 method in HEAD. Remove this method and its users.
2813 Description: Add asynchronous ldlm ENQUEUE completion handler.
2814 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2815 case that stock ldlm_completion_ast() cannot handle as it waits
2816 until lock is granted. Introduce new ldlm_completion_ast_async()
2817 for this. Also comment ldlm_completion_ast().
2821 Description: ldlm_error <-> errno conversion.
2822 Details : Add functions to map (rather arbitrary) between LDLM error codes
2823 and standard errno values. CLIO needs this to prevent LDLM specific
2824 constants from escaping ldlm and osc.
2828 Description: Kill unused ldlm_handle2lock_ns() function.
2829 Details : Kill unused ldlm_handle2lock_ns() function.
2833 Description: Add lu_ref support to ldlm_lock
2834 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2835 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2836 and ldlm_resource. LDLM interface has to be changed, because code that
2837 releases a reference on a lock, has to "know" what reference this is.
2838 In the most frequent case
2840 lock = ldlm_handle2lock(handle);
2842 LDLM_LOCK_PUT(lock);
2844 no changes are required. When any other reference (received _not_ from
2845 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2846 instead of LDLM_LOCK_PUT().
2848 Arguably, changes are pervasive, and interface requires some discipline
2849 for proper use. On the other hand, it was very instrumental in finding
2850 a few leaked lock references.
2854 Description: Add ldlm_lock_addref_try().
2855 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2856 attempts to addref a lock that might be being canceled
2861 Description: Add ldlm_weigh_callback().
2862 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2863 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2864 hard-coded `number of pages' logic.
2868 Description: Add lockdep annotations to llog code.
2869 Details : Use appropriately tagged _nested() locking calls in the places
2870 where llog takes more than one ->lgh_lock lock.
2874 Description: Add loi_kms_set().
2875 Details : Wrap kms updates into a helper function.
2879 Description: Constify instances of struct lsm_operations.
2880 Details : Constify instances of struct lsm_operations.
2884 Description: lu_conf support.
2885 Details : On a server, a file system object is uniquely identified
2886 by a fid, which is sufficient to locate and load all object
2887 state (inode). On a client, on the other hand, more data are
2888 necessary instantiate an object. Change lu_object_find() and
2889 friends to take additional `lu_conf' argument describing object.
2890 Typically this includes layout information.
2894 Description: lu_context fixes.
2895 Details : Introduce new lu_context functions that are needed on the client
2896 side, where some system threads (ptlrpcd) are shared by multiple
2897 modules, and so cannot be stopped during module shutdown.
2901 Description: Add start and stop methods to lu_device_type_operations.
2902 Details : Introduce two new methods in lu_device_type_operations, that are
2903 invoked when first instance of a given type is created and last one
2904 is destroyed respectively. This is need by CLIO.
2908 Description: Add lu_ref support to struct lu_device.
2909 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2910 track leaked references.
2914 Description: Introduce lu_kmem_descr.
2915 Details : lu_kmem_descr and its companion interface allow to create
2916 and destroy a number of kmem caches at once.
2920 Description: Fix lu_object finalization race.
2921 Details : Fix a race between lu_object_find() finding an object and its
2922 concurrent finalization. This race is (most likely) not possible
2923 on the server, but might happen on the client.
2927 Description: Introduce lu_ref interface.
2928 Details : lu_ref is a debugging module allowing to track references to
2929 a given object. It is quite cpu expensive, and has to be
2930 explicitly enabled with --enable-lu_ref. See usage description
2935 Description: Factor lu_site procfs stats into a separate function.
2936 Details : Separate lu_site stats printing code into a separate function
2937 that can be reused on a client.
2941 Description: Constify instances of struct {lu,dt,md}_device_operations.
2942 Details : Constify instances of struct {lu,dt,md}_device_operations.
2946 Description: Introduce struct md_site and move meta-data specific parts of
2947 struct lu_site here.
2948 Details : Move md-specific fields out of struct lu_site into special struct
2949 md_site, so that lu_site can be used on a client.
2953 Description: Kill mdd_lov_destroy().
2954 Details : Remove unused mdd code.
2958 Description: Add st_block checking to multistat.c.
2959 Details : Add st_block checking to multistat.c.
2963 Description: Add lu_ref support to struct obd_device.
2964 Details : Add lu_ref tracking to obd_device.
2968 Description: Kill obd_set_fail_loc().
2969 Details : Remove unused code.
2973 Description: Add special type for ptlrpc_request interpret functions.
2974 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2975 there. Actual usage will be in CLIO.
2979 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2980 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2981 doesn't work with lockdep.
2985 Description: Add rwv.c test program.
2986 Details : New testing program exercising readv(2) and writev(2) (Qian).
2990 Description: Add sendfile.c test program.
2991 Details : New testing program exercising sendfile(2) (Jay).
2995 Description: Ratelimit a message that can be very frequent.
2996 Details : Ratelimit a memory allocation failure message that can
3001 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3002 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3006 Description: Liblustre build fixes.
3007 Details : Liblustre build fixes.
3011 Description: libcfs: add cfs_{need,cond}_resched() interface.
3012 Details : libcfs: add cfs_{need,cond}_resched() definition and
3013 implementations for Linux, NT, and liblustre.
3015 Severity : enhancement
3017 Description: More exported tunables for mballoc
3018 Details : Add support for tunable preallocation window and new tunables for
3019 large/small requests
3023 Description: Detect corruption of block bitmap and checking for preallocations
3024 Details : Checks validity of on-disk block bitmap. Also it does better
3025 checking of number of applied preallocations. When corruption is
3026 found, it turns filesystem readonly to prevent further corruptions.
3030 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3031 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3032 window to 0. If it does hit cache miss, and read-pattern is still
3033 stride-io mode, does not reset the stride window, but also does not
3034 increase the stride window length in this case.
3038 Frequency : only for big-endian servers
3039 Description: Check if system is big-endian while mounting fs with extents feature
3040 Details : Mounting a filesystem with extents feature will fail on big-endian
3041 systems since ext3-based ldiskfs is not supported on big-endian
3042 systems. This can be over-riden with "bigendian_extents" mount option.
3044 Severity : enhancement
3046 Description: The root squash functionality
3047 Details : A security feature, which is to prevent users from being able
3048 to mount lustre on their desktop, run as root, and delete
3049 all of the files in the filesystem. The goal is accomplished by
3050 remapping user id (UID) and group id (GID) of the root user to
3051 a UID and GID specified by the system administartor via Lustre
3052 configuration management server (MGS). The functionality also
3053 allows to specify sets of clients for which the remapping does
3058 Description: Excessive recovery window
3059 Details : With AT enabled, the recovery window can be excessively long (6000+
3060 seconds). To address this problem, we no longer use
3061 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3062 timeout no longer depends on the service time, it is set to
3063 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3064 time via pb_service_time.
3068 Description: Watchdog triggered on MDS failover
3069 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3070 the OSTs know that the MDS "UUID" can be reused for the same export
3071 from a different NID, so we do not need to wait for the export to be
3075 Frequency : rare, only if using MMP with Linux RAID
3077 Description: MMP doesn't work with Linux RAID
3078 Details : While using HA for Lustre servers with Linux RAID, it is possible
3079 that MMP will not detect multiple mounts. To make this work we
3080 need to unplug the device queue in RAID when the MMP block is being
3081 written. Also while reading the MMP block, we should read it from
3082 disk and not the cached one.
3084 Severity : enhancement
3086 Description: open file using fid
3087 Details : A file can be opened using just its fid, like
3088 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3091 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3094 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3095 Details : add proper locking for creation of jbd_4k slab cache
3099 Description: MMP check in ext3_remount() fails without displaying any error
3100 Details : When multiple mount protection fails during remount, proper error
3103 Severity : enhancement
3105 Description: Allow stripe size to be up to 4G-64k
3106 Details : Fix math logic to allow large stripe sizes.
3110 Description: add check for >8TB ldiskfs filesystems
3111 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3112 Don't allow >8TB ldiskfs filesystems to be mounted without
3113 force_over_8tb mount option
3118 Descriptoin: OSS complains "too many blocks" and crashes
3119 Details : In filter_commitrw_write():
3120 1. lquota_chkquota();
3121 2. obdo_from_inode();
3122 3. lquota_pending_commit();
3123 "2." has changed uid and gid in this case, then caused
3124 lquota_chkquota() and lquota_pending_commit() to use
3125 different uid and gid. That is the root of the bug.
3129 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3130 Details : Detect and print out the 'stripe count = -1' case properly.
3134 Description: Out or order replies might be lost on replay
3135 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3136 request with tid smaller than one currently being added, add it
3137 to the start, not end of the list.
3139 --------------------------------------------------------------------------------
3141 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3143 * Support for kernels:
3144 2.6.5-7.283 (SLES 9),
3145 2.6.9-55.EL (RHEL 4),
3146 2.6.16.46-0.14 (SLES 10),
3147 2.6.18.8 vanilla (kernel.org)
3148 * Client support for unpatched kernels:
3149 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3150 2.6.16 - 2.6.22 vanilla (kernel.org)
3151 * Due to recently discovered recovery problems, we do not recommend
3152 using patchless RHEL 4 clients with this or any earlier release.
3153 * Recommended e2fsprogs version: 1.39.cfs8
3154 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3155 * Starting with this release, the ldiskfs backing filesystem required
3156 by Lustre is now in its own package, lustre-ldiskfs. This package
3157 should be installed. It is versioned separately from Lustre and
3158 may be released separately in future.
3162 Description: block reactivating mgc import until all deactivates complete
3163 Details : Fix race when failing back MDT/MGS to itself (testing)
3166 Frequency : at statup only
3168 Description: mds_lov_synchronize race leads to various problems
3169 Details : simultaneous MDT->OST connections at startup can cause the
3170 sync to abort, leaving the OSC in a bad state.
3172 Severity : enhancement
3174 Description: add optional extra BUILD_VERSION info
3175 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3176 to override the lustre version.
3179 Frequency : 2.6.18 servers only
3181 Description: ll_kern_mount() doesn't release the module reference
3182 Details : The ldiskfs module reference count never drops down to 0
3183 because ll_kern_mount() doesn't release the module reference.
3188 Description: server LBUG when using old ost_num_threads parameter
3189 Details : Accept the old ost_num_threads parameter but warn that it
3190 is deprecated, and fix an off-by-one error that caused an LBUG.
3195 Description: Transient SCSI error results in persistent IO issue
3196 Details : iobuf->dr_error is not reinitialized to 0 between two
3200 Frequency : sometimes when underlying device returns I/O errors
3202 Description: OSTs not going read-only during write failures
3203 Details : OSTs are not remounted read-only when the journal commit threads
3204 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3205 instead of the ext3 wrappers.
3209 Description: poor connect scaling with increasing client count
3210 Details : Don't run filter_grant_sanity_check for more than 100 exports
3211 to improve scaling for large numbers of clients.
3214 Frequency : SLES10 only
3216 Description: sanity-quota.sh quotacheck failed: rc = -22
3217 Details : Quotas cannot be enabled on SLES10.
3220 Frequency : liblustre clients only
3222 Description: getdirentries does not give error when run on compute nodes
3223 Details : getdirentries does not fail when the size specified as an argument
3224 is too small to contain at least one entry
3226 Severity : enhancement
3228 Description: Add LNET router traceability for debug purposes
3229 Details : If a checksum failure occurs with a router as part of the
3230 IO path, the NID of the last router that forwarded the bulk data
3231 is printed so it can be identified.
3236 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3237 Details : Due to a race condition, liblustre clients were occasionally
3238 evicted incorrectly.
3240 Severity : enhancement
3242 Description: lfs setstripe use optional parameters instead of postional
3245 Severity : enhancement
3247 Description: Nanosecond timestamp support for ldiskfs
3248 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3249 resolution timestamps. There is not yet support for this at
3250 the Lustre filesystem level.
3253 Frequency : during server recovery
3255 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3256 Details : request with rq_no_resend flag not awake l_wait_event if they get
3260 Frequency : nfs export on patchless client
3262 Description: connectathon hang when test nfs export over patchless client
3263 Details : Disconnected dentry cannot be found with lookup, so we do not need
3264 to unhash it or make it invalid
3267 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3268 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3269 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3274 Description: open req refcounting wrong on reconnect
3275 Details : If reconnect happened between getting open reply from server and
3276 call to mdc_set_replay_data in ll_file_open, we will schedule
3277 replay for unreferenced request that we are about to free.
3278 Subsequent close will crash in variety of ways.
3279 Check that request is still eligible for replay in
3280 mdc_set_replay_data().
3285 Description: disable writes to filesystem when reading health_check file
3286 Details : the default for reading the health_check proc file has changed
3287 to NOT do a journal transaction and write to disk, because this
3288 can cause reads of the /proc file to hang and block HA state
3289 checking on a healthy but otherwise heavily loaded system. It
3290 is possible to return to the previous behaviour during configure
3291 with --enable-health-write.
3293 Severity : enhancement
3295 Description: 64-bit inode version
3296 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3297 to the inode. This will be required for version-based recovery.
3302 Description: MDS fails to start if a duplicate client export is detected
3303 Details : in some rare cases it was possible for a client to connect to
3304 an MDS multiple times. Upon recovery the MDS would detect this
3305 and fail during startup. Handle this more gracefully.
3307 Severity : enhancement
3309 Description: Add -o localflock option to simulate old noflock
3311 Details : This will achieve local-only flock/fcntl locks
3317 Description: log_commit_thread vs filter_destroy race leads to crash
3318 Details : Take import reference before releasing llog record semaphore
3323 Description: Wrong request locking in request set processing
3324 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3325 imp_delayed_list, in this place should be used imp_lock.
3328 Frequency : when reconnection
3330 Description: Grant Leak when osc reconnect to OST
3331 Details : When osc reconnect ost, OST(filter) should check whether it
3332 should grant more space to client by comparing fed_grant and
3333 cl_avail_grant, and return the granted space to client instead
3334 of "new granted" space, because client will call osc_init_grant
3335 to update the client grant space info.
3338 Frequency : when client reconnect to OST
3340 Description: Grant Leak when osc do resend and replay bulk write
3341 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3342 bulk write request, because the grant info while be sync between
3343 OSC and OST when reconnect, and we should ignore the grant info
3344 these of resend/replay write req.
3349 Description: Grant space more than available left space sometimes.
3350 Details : When then OST is about to be full, if two bulk writing from
3351 different clients came to OST. Accord the avaliable space of the
3352 OST, the first req should be permitted, and the second one
3353 should be denied by ENOSPC. But if the seconde arrived before
3354 the first one is committed. The OST might wrongly permit second
3355 writing, which will cause grant space > available space.
3358 Frequency : when client is evicted
3360 Description: Grant might be wrongly erased when osc is evicted by OST
3361 Details : when the import is evicted by server, it will fork another
3362 thread ptlrpc_invalidate_import_thread to invalidate the
3363 import, where the grant will be set to 0. While the original
3364 thread will update the grant it got when connecting. So if
3365 the former happened latter, the grant will be wrongly errased
3366 because of this race.
3371 Description: Checking Stale with correct fid
3372 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3373 to check whether it is stale, because sometimes, we want the
3374 enqueue happened anyway, and op_data.fid2 will not be initialized.
3376 Severity : enhancement
3378 Description: update patchless client
3379 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3382 Frequency : only with 2.4 kernel
3384 Description: random memory corruption
3385 Details : size of struct ll_inode_info is to big for union inode.u and this
3386 can be cause of random memory corruption.
3391 Description: Memory leak in recovery
3392 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3393 It should also check obd_fail before fsfilt_start, otherwise if
3394 fsfilt_start return -EROFS,(failover mds during mds recovery).
3395 then the req will return with repmsg->transno = 0 and rc = EROFS.
3396 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3397 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3398 be freed no matter whether fsfilt_commit success or not.
3401 Frequency : only with huge count clients
3403 Description: Prevents from taking the superblock lock in llap_from_page for
3405 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3406 for llap_from_page call in ll_removepage prevents from taking the
3407 superblock lock for a soon died page.
3412 Description: Not check open intent error before release open handle
3413 Details : in some rare cases, the open intent error is not checked before
3414 release open handle, which may cause
3415 ASSERTION(open_req->rq_transno != 0), because it tries to release
3416 the failed open handle.
3421 Description: Set cat log bitmap only after create log success.
3422 Details : in some rare cases, the cat log bitmap is set too early. and it
3423 should be set only after create log success.
3427 Description: Accessing a block bevice can re-enable I/O when Lustre is
3428 tearing down a device.
3429 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3434 Description: service threads may hog cpus when there are a lot of requests
3435 Details : Insert cond_resched to give other threads a chance to use some CPU
3440 Description: the cat log was not initialized in recovery
3441 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3442 unlink log on mds will not be initialized until mds post
3443 recovery. And also in mds post recovery, the unlink log will
3444 initialization will be done asynchronausly, so there will be race
3445 between add unlink log and unlink log initialization.
3449 Description: brw_stats were being printed incorrectly
3450 Details : brw_stats were being printed as log2 but all of them were not
3451 recorded as log2. Also remove some code duplication arising from
3452 filter_tally_{read,write}.
3456 Frequency : rare, only in recovery.
3457 Description: ASSERTION(req->rq_type != LI_POISON) failed
3458 Details : imp_lock should be held while iterating over imp_sending_list for
3459 prevent destroy request after get timeout in ptlrpc_queue_wait.
3463 Description: replay-single.sh test 52 fails
3464 Details : A lock's skiplist need to be cleanup when it being unlinked
3465 from its resource list.
3469 Description: Short directio read returns full requested size rather than
3471 Details : Direct I/O operations should return actual amount of bytes
3472 transferred rather than requested size.
3474 Severity : enhancement
3476 Description: metadata RPC reduction (e.g. for rm performance)
3477 Details : decrease the amount of synchronous RPC between clients and servers
3478 by canceling conflicing lock before the operation on the client
3479 and packing thier handles into the main operation RPC to server.
3481 Severity : enhancement
3483 Description: Async OSC create to avoid the blocking unnecessarily.
3484 Details : If an OST has no remaining object, system will block on the
3485 creation when it needs to create a new object on this OST. Now,
3486 ways use pre-created objects when available, instead of blocking on
3487 an empty osc while others are not empty. If we must block, we
3488 block for the shortest possible period of time.
3492 Description: improve handling recoverable errors
3493 Details : if request processig with error which can be recoverable on server
3494 request should be resend, otherwise page released from cache and
3497 Severity : enhancement
3499 Description: refine locking for avoid write wrong info into lov_objid file
3500 Details : fix possible races with add new target and write/update data in
3503 --------------------------------------------------------------------------------
3505 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3510 Frequency : on some architectures
3512 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3513 Details : Uninitialized flags sometimes cause configuration commands to
3517 Frequency : patchless clients only
3519 Description: missing __iget() symbol export
3520 Details : The __iget() symbol export is missing. To avoid the need for
3521 this on patchless clients the deathrow inode reaper is turned
3522 off, and we depend on the VM to clean up old inodes. This
3523 dependency was during via the fix for bug 12181.
3525 --------------------------------------------------------------------------------
3527 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3529 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3530 INTEROPERATE with older versions automatically. In many cases a
3531 special upgrade step is needed. Please read the
3532 user documentation before upgrading any part of a live system.
3533 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3534 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3535 * WARNING: Lustre configuration and startup changes are required with
3536 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3540 Severity : enhancement
3542 Description: MountConf
3543 Details : Lustre configuration is now managed via mkfs and mount
3544 commands instead of lmc and lconf. New obd types (MGS, MGC)
3545 are added for dynamic configuration management. See
3546 https://mail.clusterfs.com/wikis/lustre/MountConf for
3549 Severity : enhancement
3551 Description: dynamic OST addition
3552 Details : OSTs can now be added to a live filesystem
3554 Severity : enhancement
3556 Description: startup order invariance
3557 Details : MDTs and OSTs can be started in any order. Clients only
3558 require the MDT to complete startup.
3560 Severity : enhancement
3562 Description: parallel, asynchronous orphan cleanup
3563 Details : orphan cleanup is now performed in separate threads for each
3564 OST, allowing parallel non-blocking operation.
3566 Severity : enhancement
3568 Description: optimized stripe assignment
3569 Details : stripe assignments are now made based on ost space available,
3570 ost previous usage, and OSS previous usage, in order to try
3571 to optimize storage space and networking resources.
3573 Severity : enhancement
3575 Description: Permanently set tunables
3576 Details : All writable /proc/fs/lustre tunables can now be permanently
3577 set on a per-server basis, at mkfs time or on a live system.
3579 Severity : enhancement
3581 Description: Lustre message v2
3582 Details : Add lustre message format v2.
3584 Severity : enhancement
3586 Description: client OST exclusion list
3587 Details : Clients can be started with a list of OSTs that should be
3588 declared "inactive" for known non-responsive OSTs.
3592 Description: SPEC SFS validation failure on NFS v2 over lustre.
3593 Details : Changes the blocksize for regular files to be 2x RPC size,
3594 and not depend on stripe size.
3596 Severity : enhancement
3598 Description: Multiple MD RPCs in flight.
3599 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3600 To avoid overly-overloading MDS, introduce a limit on number of
3601 MDS RPCs in flight for a single client and add /proc controls
3602 to adjust this limit.
3604 Severity : enhancement
3606 Description: client read/write statistics
3607 Details : Add client read/write call usage stats for performance
3608 analysis of user processes.
3609 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3610 file access. extents_stats shows chunk size distribution.
3611 extents_stats_per_process show chunk size distribution per
3614 Severity : enhancement
3616 Description: mds statistics
3617 Details : Add detailed mds operations statistics in
3618 /proc/fs/lustre/mds/*/stats.
3622 Description: Failure of copying files with lustre special EAs.
3623 Details : Client side always return success for setxattr call for lustre
3624 special xattr (currently only "trusted.lov").
3629 Description: Refcount LNET uuids
3630 Details : The global LNET uuid list grew linearly with every startup;
3631 refcount repeated list entries instead of always adding to
3634 Severity : enhancement
3636 Description: Dynamic service threads
3637 Details : Within a small range, start extra service threads
3638 automatically when the request queue builds up.
3641 Frequency : mixed-endian client/server environments
3643 Description: mixed-endian crashes
3644 Details : The new msg_v2 system had some failures in mixed-endian
3647 Severity : enhancement
3649 Description: Easy OST removal
3650 Details : OSTs can be permanently deactivated with e.g. 'lctl
3651 conf_param lustre-OST0001.osc.active=0'
3653 Severity : enhancement
3655 Description: MGS proc entries
3656 Details : Added basic proc entries for the MGS showing what filesystems
3659 Severity : enhancement
3661 Description: provide MGS failover
3662 Details : Added config lock reacquisition after MGS server failover.
3664 Severity : enhancement
3666 Description: add Linux 2.4 support
3667 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3671 Description: a large application tries to do I/O to the same resource and dies
3672 in the middle of it.
3673 Details : Check the req->rq_arrival time after the call to
3674 ost_brw_lock_get(), but before we do anything about
3675 processing it & sending the BULK transfer request. This
3676 should help move old stale pending locks off the queue as
3677 quickly as obd_timeout.
3680 Frequency : when an incorrect nid is specified during startup
3682 Description: ptlrpc connect to non-existant node causes kernel crash
3683 Details : LNET can't be re-entered from an event callback, which
3684 happened when we expire a message after the export has been
3685 cleaned up. Instead, hand the zombie cleanup off to another
3688 Severity : enhancement
3690 Description: plain/inodebits lock performance improvement
3691 Details : Grouping plain/inodebits in granted list by their request modes
3692 and bits policy, thus improving the performance of search through
3696 Frequency : only if OST filesystem is corrupted
3698 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3699 Details : for a short time RPCs with bulk IO are in the replay list,
3700 but replay of bulk IOs is unimplemented. If the OST filesystem
3701 is corrupted due to disk cache incoherency and then replay is
3702 started it is possible to trip an assertion. Avoid putting
3703 committed RPCs into the replay list at all to avoid this issue.
3706 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3708 Description: System hang on startup
3709 Details : This bug allowed the liblustre (e.g. catamount) client to
3710 return to the app before handling all startup RPCs. This
3711 could leave the node unresponsive to lustre network traffic
3712 and manifested as a server ptllnd timeout.
3714 Severity : enhancement
3716 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3717 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3718 boolean to control whether to print peer debug info when a
3719 client's RPC times out.
3722 Frequency : only for kernels with patches from Lustre below 1.4.3
3724 Description: Remove old rdonly API
3725 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3728 Frequency : only for devices with external journals
3730 Description: Set external device read-only also
3731 Details : During a commanded failover stop, we set the disk device
3732 read-only while the server shuts down. We now also set any
3733 external journal device read-only at the same time.
3736 Frequency : when upgrading from 1.4 while trying to change parameters
3738 Description: The wrong (new) MDC name was used when setting parameters for
3739 upgraded MDT's. Also allows changing of OSC (and MDC)
3740 parameters if --writeconf is specified at tunefs upgrade time.
3743 Frequency : when setting specific ost indicies
3745 Description: QOS code breaks on skipped indicies
3746 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3747 created with --index need not be sequential.
3749 Severity : enhancement
3751 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3752 Details : The uninit_groups feature works in conjunction with the kernel
3753 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3754 up the pass1 processing of e2fsck. This is a read-only feature
3755 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3756 mount filesystems that have had this feature enabled.
3758 Severity : enhancement
3760 Description: Improve multi-block allocation algorithm to avoid fragmentation
3761 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3762 allocation locality and avoid filesystem fragmentation.
3764 ------------------------------------------------------------------------------
3766 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3768 * Support for kernels:
3769 2.4.21-47.0.1.EL (RHEL 3)
3770 2.6.5-7.283 (SLES 9)
3771 2.6.9-42.0.10.EL (RHEL 4)
3772 2.6.12.6 vanilla (kernel.org)
3773 2.6.16.27-0.9 (SLES 10)
3774 * Recommended e2fsprogs version: 1.39.cfs5
3776 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3780 Frequency : occasional, depends on client load and configuration
3781 Bugzilla : 12181, 12203
3782 Description: data loss for recently-modified files
3784 Details : In some cases it is possible that recently written or created
3785 files may not be written to disk in a timely manner (this should
3786 normally be within 30s unless client IO load is very high).
3787 The problem appears as zero-length files or files that are a
3788 multiple of 1MB in size after a client crash or client eviction
3789 that are missing data at the end of the file.
3791 This problem is more likely to be hit on clients where files are
3792 repeatedly created and unlinked in the same directory, clients
3793 have a large amount of RAM, have many CPUs, the filesystem has
3794 many OSTs, the clients are rebooted frequently, and/or the files
3795 are not accessed by other nodes after being written.
3797 The presence of the problem can be detected by looking at
3798 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3799 smaller than the second (nr_unused) then dirty files will not
3800 be flushed automatically to disk. "sync; sleep 10" should be
3801 run several times on the node before unmounting it to update
3802 Lustre (this is also safe to run on nodes without this problem).
3804 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3805 that can cause this same problem, so customers using that kernel
3806 also need to update the kernel in addition to Lustre. In order
3807 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3809 It is normal that files written just before a client crash (less
3810 than 30s) may not yet have been flushed to disk, even for local
3814 Frequency : frequent on thin XT3 nodes
3816 Description: UUID collision on thin XT3 Linux nodes
3817 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3818 randomly, since we relied on an insufficiently-seeded PRNG.
3823 Description: OSS hangs after "All ost request buffers busy"
3824 Details : A deadlock between quota and journal operations caused OSS
3825 hangs after printing "All ost request buffers busy."
3828 Frequency : always on liblustre builds
3830 Description: Cleanup compiler warnings on liblustre
3833 Frequency : always on liblustre builds on XT3
3835 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3839 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3840 > 32000 subdirectories
3841 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3842 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3843 subdirectory count crosses 32000. This will aid e2fsck to
3844 correctly handle more than 32000 subdirectories.
3847 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3849 Description: System hang on startup
3850 Details : This bug allowed the liblustre (e.g. catamount) client to
3851 return to the app before handling all startup RPCs. This
3852 could leave the node unresponsive to lustre network traffic
3853 and manifested as a server ptllnd timeout.
3855 Severity : enhancement
3857 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3858 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3859 boolean to control whether to print peer debug info when a
3860 client's RPC times out.
3865 Description: make O_SYNC working on 2.6 kernels
3866 Details : 2.6 kernels use different method for mark pages for write,
3867 so need add a code to lustre for O_SYNC work.
3872 Description: Failure to close file and release space on NFS
3873 Details : Put inode details into lock acquired in ll_intent_file_open.
3874 Use mdc_intent_lock in ll_intent_open to properly
3875 detect all kind of errors unhandled by mdc_enqueue.
3880 Description: proc file read during shutdown sometimes raced obd removal,
3882 Details : Add lock to prevent obd access after proc file removal.
3885 Frequency : Only for files larger than 4GB on 32-bit clients.
3887 Description: improperly doing page alignment of locks
3888 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3889 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3894 Description: RPCs being resent when they shouldn't be
3895 Details : Some RPCs that should not be resent are being resent. This
3896 can cause inconsistencies in the RPC state machine. Do not
3897 resend such requests.
3900 Frequency : rare, only with NFS export
3902 Description: Crash on NFS re-export node
3903 Details : under very unusual load conditions an assertion is hit in
3904 ll_intent_file_open()
3907 Frequency : only if OST filesystem is corrupted
3909 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3910 Details : for a short time RPCs with bulk IO are in the replay list,
3911 but replay of bulk IOs is unimplemented. If the OST filesystem
3912 is corrupted due to disk cache incoherency and then replay is
3913 started it is possible to trip an assertion. Avoid putting
3914 committed RPCs into the replay list at all to avoid this issue.
3919 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3920 Details : Large single O_DIRECT read and write calls can fail to allocate
3921 a sufficiently large buffer to process the request. In case of
3922 allocation failure the allocation is retried with a smaller
3923 buffer and broken into smaller requests.
3925 Severity : enhancement
3927 Description: Add -o localflock option to simulate old noflock behaviour.
3928 Details : This will achieve local-only flock/fcntl locks coherentness.
3933 Description: versioning check is incomplete
3934 Details : Checking the version difference of client vs. server, report
3935 error if the gap is too big.
3940 Description: add support PG_writeback bit
3941 Details : add support for PG_writeback bit for Lustre, for more carefull
3942 work with page cache in 2.6 kernel. This also fix some deadlocks
3943 and remove hack for work O_SYNC with 2.6 kernel.
3945 Severity : enhancement
3947 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3948 Details : The uninit_groups feature works in conjunction with the kernel
3949 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3950 up the pass1 processing of e2fsck. This is a read-only feature
3951 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3952 mount filesystems that have had this feature enabled.
3954 Severity : enhancement
3956 Description: Improve multi-block allocation algorithm to avoid fragmentation
3957 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3958 allocation locality and avoid filesystem fragmentation.
3960 ------------------------------------------------------------------------------
3962 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3964 * Support for kernels:
3965 2.6.9-42.0.3.EL (RHEL 4)
3966 2.6.5-7.276 (SLES 9)
3967 2.4.21-47.0.1.EL (RHEL 3)
3968 2.6.12.6 vanilla (kernel.org)
3969 2.6.16.21-0.8 (SLES10)
3970 * Recommended e2fsprogs version: 1.39.cfs2-0
3972 * The backwards-compatible /proc/sys/portals symlink has been removed
3973 in this release. Before upgrading, please ensure that you change
3974 any configuration scripts or /etc/sysctl.conf files that access
3975 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3976 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3977 in advance of the upgrade on any system running Lustre 1.4.6 or
3978 newer, since /proc/sys/lnet was added in that version.
3979 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3983 Frequency : only when quota is used
3985 Description: avoid scanning export list for quota master
3986 Details : Change the algorithms to avoid scanning export list in order
3987 to improve the efficiency.
3990 Frequency : MDS failover only, very rarely
3992 Description: "went back in time" messages on mds failover
3993 Details : The greatest transno may be lost when the current operation
3994 finishes with an error (transno==0) and the client's last_rcvd
3995 record is over-written. Save the greatest transno in the
3996 mds_last_transno for this case.
3999 Frequency : always for specific kernels and striping counts
4001 Description: client may get "Matching packet too big" without ACL support
4002 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4003 when trying to access files in certain configurations. The
4004 clients should in fact be denied when mounting because they do
4005 not understand ACLs.
4008 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4010 Description: many clients connecting with IO in progress causes connect timeouts
4011 Details : Avoid synchronous journal commits to avoid delays caused by many
4012 clients connecting/disconnecting when bulk IO is in progress.
4013 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4014 OST_IO_PORTAL to avoid delays behind potentially many pending
4018 Frequency : occasionally with multiple writers to a single file
4020 Description: shared writes to file may result in wrong size reported by stat()
4021 Details : Allow growing of kms when extent lock is cancelled
4024 Frequency : always with random mmap IO to multi-striped file
4026 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4027 Details : Only if the hole is at the end of OST object so that kms is too
4028 small. Fix is to increase kms accordingly in ll_nopage.
4031 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4033 Description: writes to a missing object would leak memory on the OST
4034 Details : If there is an inconsistency between the MDS and OST filesystems,
4035 such that the MDS references an object that doesn't exist, writes
4036 to that object will leak memory due to incorrect cleanup in the
4037 error handling path, eventually running out of memory on the OST.
4042 Description: Creating too long symlink causes lustre errors
4043 Details : Check symlink and name lengths before sending requests to MDS.
4046 Frequency : only if flock is enabled (not on by default)
4048 Description: posix locks not released on fd closure on 2.6.9+
4049 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4050 kernels, this caused such locks not to be released on fd close and
4051 then assertions on fs unmount about still used locks.
4054 Frequency : MDS failover only, very rarely
4056 Description: clients may get ASSERTION(granted_lock != NULL)
4057 Details : When request was taking a long time, and a client was resending
4058 a getattr by name lock request. The were multiple lock requests
4059 with the same client lock handle and
4060 mds_getattr_name->fixup_handle_for_resent_request found one of the
4061 lock handles but later failed with ASSERTION(granted_lock != NULL).
4066 Description: handle->h_buffer_credits > 0, assertion failure
4067 Details : h_buffer_credits is zero after truncate, causing assertion
4068 failure. This patch extends the transaction or creates a new
4072 Frequency : NFS re-export or patchless client
4073 Bugzilla : 11179, 10796
4074 Description: Crash on NFS re-export node (__d_move)
4075 Details : We do not want to hash the dentry if we don't have a lock.
4076 But if this dentry is later used in d_move, we'd hit uninitialised
4077 list head d_hash, so we just do this to init d_hash field but
4078 leave dentry unhashed.
4081 Frequency : NFS re-export or patchless client
4083 Description: NFS exports has problem with symbolic link
4084 Details : lustre client didn't properly install dentry when re-exported
4085 to NFS or running patchless client.
4088 Frequency : NFS re-export or patchless client
4090 Description: Various nfs/patchless fixes.
4091 Details : fixes reuse disconected alias for lookup process - this fixes
4092 warning "find_exported_dentry: npd != pd",
4093 fix permission error with open files at nfs.
4094 fix apply umask when do revalidate.
4097 Frequency : occasional
4099 Description: Crash on NFS re-export node
4100 Details : calling clear_page() on the wrong pointer triggered oops in
4101 generic_mapping_read().
4104 Frequency : rarely, using O_DIRECT IO
4106 Description: unaligned directio crashes client with LASSERT
4107 Details : check for unaligned buffers before trying any requests.
4110 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4112 Description: stale data returned from RAID cache
4113 Details : If only a small amount of IO is done to the RAID device before
4114 reading it again it is possible to get stale data from the RAID
4115 cache instead of reading it from disk.
4118 Frequency : always for sles10 kernel
4120 Description: sles10 support
4121 Details : ll_follow_link: compile fixes and using of nd_set_link
4122 under newer kernels.
4125 Frequency : depends on arch, kernel and compiler version, always on sles10
4128 Description: recursive or deep enough symlinks cause stack overflow
4129 Details : getting rid of large stack-allocated variable in
4133 Frequency : depends on hardware
4135 Description: lustre write performance loss in the SLES10 kernel
4136 Details : the performance loss is caused by using of write barriers in the
4137 ext3 code. The SLES10 kernel turns barrier support on by
4138 default. The fix is to undo that change for ldiskfs.
4140 ------------------------------------------------------------------------------
4142 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4144 * Support for kernels:
4145 2.6.9-42.0.3EL (RHEL 4)
4146 2.6.5-7.276 (SLES 9)
4147 2.4.21-47.0.1.EL (RHEL 3)
4148 2.6.12.6 vanilla (kernel.org)
4152 Frequency : quota enabled and large files being deleted
4154 Description: releasing more than 4GB of quota at once hangs OST
4155 Details : If a user deletes more than 4GB of files on a single OST it
4156 will cause the OST to spin in an infinite loop. Release
4157 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4162 Description: statfs data retrieved from /proc may be stale or zero
4163 Details : When reading per-device statfs data from /proc, in the
4164 {kbytes,files}_{total,free,avail} files, it may appear
4165 as zero or be out of date.
4168 Frequency : systems with MD RAID1 external journal devices
4170 Description: lconf's call to blkid is confused by RAID1 journal devices
4171 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4172 of returning all block devices that match the journal UUID.
4175 Frequency : always, for aggregate stripe size over 4GB
4177 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4178 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4179 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4182 Frequency : always if "lfs find" used on a local file/directory
4184 Description: "lfs find" segfaults if used on a local file/directory
4185 Details : The case where a directory component was not specified wasn't
4186 handled correctly. Handle this properly.
4189 Frequency : always on ppc64
4191 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4192 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4193 perform endianness conversion.
4196 Frequency : rarely (truncate to non-zero file size after write under load)
4197 Bugzilla : 10730, 10687
4198 Description: Files padded with zeros to next 4K multiple
4199 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4200 it is possible that files that are truncated to a non-zero size
4201 immediately after being written are filled with zero bytes beyond
4202 the truncated size. No file data is lost.
4204 Severity : enhancement
4206 Description: Allow recovery/failover for liblustre clients.
4207 Details : liblustre clients were unaware of failover configurations until
4210 Severity : enhancement
4212 Description: user file locks should fail when not mounting with flock option
4213 Details : Set up an error-returning stub in ll_file_operations.lock field
4214 to prevent incorrect behaviour when client is mounted without
4215 flock option. Also, set up properly f_op->flock field for
4219 Frequency : always on ia64
4221 Description: "lfs df" loops on printing out MDS statfs information
4222 Details : The obd_ioctl_data was not initialized and in some systems
4223 this caused a failure during the ioctl that did not return
4224 an error. Initialize the struct and return an error on failure.
4227 Frequency : SLES 9 only
4229 Description: Error of copying files with lustre special EAs as root
4230 Details : Client side always return success for setxattr call for lustre
4231 special xattr (currently only "trusted.lov").
4234 Frequency : rarely on clusters with both ia64+i386 clients
4236 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4237 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4238 clients are doing concurrent writes to the same file it is
4239 possible that the ia64 clients may LASSERT because the OST
4240 extent locks are not PAGE_SIZE aligned. Ensure that grown
4241 locks are always aligned on the request boundary.
4244 Frequency : specific use, occasional
4246 Description: Overwriting in use executable truncates on-disk binary image
4247 Details : If one node attempts to overwrite an executable in use by
4248 another node, we now correctly return ETXTBSY instead of
4249 truncating the file.
4251 Severity : enhancement
4253 Description: Async OSC create to avoid the blocking unnecessarily.
4254 Details : If an OST has no remaining object, system will block on the creating
4255 when need to create a new object on this OST. Now, ways use
4256 pre-created objects when available, instead of blocking on an
4257 empty osc while others are not empty. If we must block, we block
4258 for the shortest possible period of time.
4263 Description: chmod on Lustre root is propagated to other clients
4264 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4265 invalid by the follow_mount time.
4268 Frequency : liblustre clients only
4270 Description: Race in 'instant cancel' lock handling could lead to such locks
4271 never to be granted in case of SMP MDS
4272 Details : Do not destroy not yet granted but cbpending locks in
4276 Frequency : replay/resend of open
4278 Description: non null lock assetion failure in mds_intent_policy
4279 Details : Trying to replay/resend lockless open requests resulted in
4280 mds_open() returning 0 with no lock. Now it sets a flag if
4281 a lock is going to be returned.
4283 Severity : enhancement
4285 Description: Checksum enhancements
4286 Details : New checksum enhancements allow for resending RPCs that failed
4289 Severity : enhancement
4291 Description: Tunables on number of dirty pages in cacche
4292 Details : Allow to set limit on number of dirty pages cached.
4297 Description: client crash on unmount - lock still has references
4298 Details : In some error handling cases it was possible to leak a lock
4299 reference on a client while accessing a file. This was not
4300 harmful to the client during operation, but would cause the
4301 client to crash when the filesystem is unmounted.
4304 Frequency : specific case, rare
4306 Description: ETXTBSY on mds though file not in use by client
4307 Details : ETXTBSY is no longer incorrectly returned when attempting to
4308 chmod or chown a directory that the user previously tried to
4309 execute or a currently-executing binary.
4312 Frequency : extremely rare except on liblustre-based clients
4314 Description: Lustre space not freed when files are deleted
4315 Details : Clean up open-unlinked files after client eviction. Previously
4316 the unlink was skipped and the files remained as orphans.
4321 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4322 Details : In some cases it was possible to send a blocking callback to
4323 a client doing a glimpse, even though that client didn't get
4324 a lock granted. When the glimpse lock is cancelled on the OST
4325 the freed lock is left on the waiting list and corrupted the list.
4328 Frequency : all core dumps
4330 Description: Broke core dumps to lustre
4331 Details : Negative dentry may be unhashed if parent does not have UPDATE
4332 lock, but some callers, e.g. do_coredump, expect dentry to be
4333 hashed after successful create, hash it in ll_create_it.
4335 ------------------------------------------------------------------------------
4337 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4339 * Support for kernels:
4340 2.6.9-42.0.2.EL (RHEL 4)
4341 2.6.5-7.276 (SLES 9)
4342 2.4.21-40.EL (RHEL 3)
4343 2.6.12.6 vanilla (kernel.org)
4347 Frequency : always on RHEL 3
4349 Description: Number of open files grows over time
4350 Details : The number of open files grows over time, whether or not
4351 Lustre is started. This was due to a filp leak introduced
4352 by one of our kernel patches.
4354 ------------------------------------------------------------------------------
4356 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4358 * Support for kernels:
4359 2.6.9-42.EL (RHEL 4)
4360 2.6.5-7.276 (SLES 9)
4361 2.4.21-40.EL (RHEL 3)
4362 2.6.12.6 vanilla (kernel.org)
4367 Bugzilla : 5719, 9635, 9792, 9684
4368 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4369 Details : If a server is under heavy load and cannot reply to new
4370 connection requests before the client resends the (re)connect,
4371 the connection handling code can behave badly if two service
4372 threads are concurrently handing separate (re)connections from
4373 the same client. Add better locking to the connection handling
4374 code, and ensure that only a single connection will be processed
4375 for a given client UUID, even if the lock is dropped.
4377 Severity : enhancement
4379 Description: add TCP zero-copy support to kernel
4380 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4381 sends if the hardware supports scatter-gather and checksumming.
4382 This allows socklnd to do client-write and server-read more
4383 efficiently and reduce CPU utilization from skbuf copying.
4386 Frequency : only if NFS exporting from client
4388 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4389 Details : If setting attributes on a file created under NFS that had
4390 never been opened it would be possible to oops the client
4391 if the file had no objects.
4394 Frequency : always for liblustre
4396 Description: liblustre client does MDS+OSTs setattr RPC for each write
4397 Details : When doing a write from a liblustre client, the client
4398 incorrectly issued an RPC to the MDS and each OST the file was
4399 striped over in order to update the timestamps. When writing
4400 with small chunks and many clients this could overwhelm the MDS
4401 with RPCs. In all cases it would slow down the write because
4402 these RPCs are unnecessary.
4404 Severity : enhancement
4406 Description: allow number of MDS service threads to be changed at module load
4407 Details : It is now possible to change the number of MDS service threads
4408 running. Adding "options mds mds_num_threads={N}" to the MDS's
4409 /etc/modprobe.conf will set the number of threads for the next
4410 time Lustre is restarted (assuming the "mds" module is also
4411 reloaded at that time). The default number of threads will
4412 stay the same, 32 for most systems.
4417 Description: OST crash if filesystem is unformatted or corrupt
4418 Details : If an OST is started on a device that has never been formatted
4419 or if the filesystem is corrupt and cannot even mount then the
4420 error handling cleanup routines would dereference a NULL pointer.
4425 Description: NULL pointer deref in llap_from_page.
4426 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4427 incorrect) mapping if the page was truncated/reclaimed while it was
4428 searched for. Check for this condition and skip such pages when
4429 doing readahead. Introduce extra check to llap_from_page() to
4430 verify page->mapping->host is non-NULL (so page is not anonymous).
4433 Frequency : Sometimes when using sys_sendfile
4435 Description: "page not covered by a lock" warnings from ll_readpage
4436 Details : sendfile called ll_readpage without right page locks present.
4437 Now we introduced ll_file_sendfile that does necessary locking
4438 around call to generic_file_sendfile() much like we do in
4442 Frequency : with certain MDS communication failures at client mount time
4444 Description: NULL pointer deref after failed client mount
4445 Details : a client connection request may delayed by the network layer
4446 and not be sent until after the PTLRPC layer has timed out the
4447 request. If the client fails the mount immediately it will try
4448 to clean up before the network times out the request. Add a
4449 reference from the request import to the obd device and delay
4450 the cleanup until the network drops the request.
4453 Frequency : occasionally during client (re)connect
4455 Description: assertion failure during client (re)connect
4456 Details : processing a client connection request may be delayed by the
4457 client or server longer than the client connect timeout. This
4458 causes the client to resend the connection request. If the
4459 original connection request is replied in this interval, the
4460 client may trip an assertion failure in ptlrpc_connect_interpret()
4461 which thought it would be the only running connect process.
4464 Frequency : only with obd_echo servers and clients that are rebooted
4466 Description: kernel BUG accessing uninitialized data structure
4467 Details : When running an obd_echo server it did not start the ping_evictor
4468 thread, and when a client was evicted an uninitialized data
4469 structure was accessed. Start the ping_evictor in the RPC
4470 service startup instead of the OBD startup.
4472 Severity : enhancement
4473 Bugzilla : 10193 (patchless)
4474 Description: Remove dependency on various unexported kernel interfaces.
4475 Details : No longer need reparent_to_init, exit_mm, exit_files,
4476 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4479 Frequency : rare (only users of deprecated and unsupported LDAP config)
4481 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4482 Details : LDAP apparently contains 'lustreName' attributes instead of
4483 'name'. A simple remapping of the name is sufficient.
4486 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4488 Description: waiting_locks_callback trips kernel BUG if client is evicted
4489 Details : Running with the dump_on_timeout debug flag turned on makes
4490 it possible that the waiting_locks_callback() can try to dump
4491 the Lustre kernel debug logs from an interrupt handler. Defer
4492 this log dumping to the expired_lock_main() thread.
4494 Severity : enhancement
4496 Description: Support NFS exporting on 2.6 kernels.
4497 Details : Implement non-rawops metadata methods for NFS server to use without
4498 changing NFS server code.
4501 Frequency : very rare (synthetic metadata workload only)
4503 Description: two racing renames might cause an MDS thread to deadlock
4504 Details : Running the "racer" program may cause one MDS thread to rename
4505 a file from being the source of a rename to being the target of
4506 a rename at exactly the same time that another thread is doing
4507 so, and the second thread has already enqueued these locks after
4508 doing a lookup of the target and is trying to relock them in
4509 order. Ensure that we don't try to re-lock the same resource.
4512 Frequency : only very large systems with liblustre clients
4514 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4515 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4516 all OSTs in parallel. This allows the eviction of stale liblustre
4517 clients to proceed much faster than if they were done in series,
4518 and also offers similar improvements for other set_info RPCs.
4523 Description: excessive CPU usage during initial read phase on client
4524 Details : During the initial read phase on a client, it would agressively
4525 retry readahead on the file, consuming too much CPU and impacting
4526 performance (since 1.4.5.8). Improve the readahead algorithm
4527 to avoid this, and also improve some other common cases (read
4528 of small files in particular, where "small" is files smaller than
4529 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4534 Description: MDS crash when receiving packet with unknown intent.
4535 Details : Do not LBUG in unknown intent case, just return -EFAULT
4537 Severity : enhancement
4538 Bugzilla : 9293, 9385
4539 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4540 Details : Do not serialize getattr (non-intent version) and statfs.
4543 Frequency : occasional, when OST network is overloaded/intermittent
4545 Description: client evicted by OST after bulk IO timeout
4546 Details : If a client sends a bulk IO request (read or write) the OST
4547 may evict the client if it is unresposive to its data GET/PUT
4548 request. This is incorrect if the network is overloaded (takes
4549 too long to transfer the RPC data) or dropped the OST GET/PUT
4550 request. There is no need to evict the client at all, since
4551 the pinger and/or lock callbacks will handle this, and the
4552 client can restart the bulk request.
4555 Frequency : Always when mmapping file with no objects
4557 Description: client crashes when mmapping file with no objects
4558 Details : Check that we actually have objects in a file before doing any
4559 operations on objects in ll_vm_open, ll_vm_close and
4565 Description: Request leak when working with deleted CWD
4566 Details : Introduce advanced request refcount tracking for requests
4567 referenced from lustre intent.
4569 Severity : Enhancement
4571 Description: Cache open file handles on client.
4572 Details : MDS now will return special lock along with openhandle, if
4573 requested and client is allowed to hold openhandle, even if unused,
4574 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4575 closing files for every read/write openration.
4577 Severity : Enhancement
4579 Description: Cache open negative dentries on client when possible.
4580 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4581 negative dentries on lock revocation.
4586 Description: Remounting a client read-only wasn't possible with a zconf mount
4587 Details : It wasn't possible to remount a client read-only with llmount.
4589 Severity : enhancement
4590 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4591 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4592 MPICH is included to add Lustre-specific ADIO interfaces.
4593 This is based closely on the UFS ADIO layer and only differs
4594 in file creation, in order to allow the OST striping to be set.
4595 This is user-contributed code and not supported by CFS.
4600 Description: extended inode attributes (immutable, append-only) work improperly
4601 when 2.4 and 2.6 kernels are used on client/server or vice versa
4602 Details : Introduce kernel-independent values for these flags.
4604 Severity : enhancement
4607 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4608 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4609 granularity. Now, Fractional MB granularity is be supported,
4610 this is very useful for low memory system.
4612 Severity : enhancement
4614 Description: Getattr by fid
4615 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4616 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4617 have locks and avoids getattr rpc requests for it on every stat.
4620 Frequency : Always, for filesystems larger than 2TB
4622 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4623 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4624 filesystems larger than 2TB to be mounted with mballoc enabled.
4627 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4629 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4630 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4631 filesystems larger than 2TB is mounted then the kernel will
4632 silently corrupt the start of the filesystem. CONFIG_LBD is
4633 enabled for all CFS-supported kernels, but the possibility of
4634 this happening with a modified kernel config exists.
4636 Severity : enhancement
4638 Description: add client O_DIRECT support for 2.6 kernels
4639 Details : It is now possible to do O_DIRECT reads and writes to files
4640 in the Lustre client mountpoint on 2.6 kernel clients.
4642 Severity : enhancement
4644 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4645 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4646 parallel, not waiting for response from every OST before sending
4647 a rpc to the next OST.
4652 Description: setattr vs write race when updating file timestamps
4653 Details : Client processes that update a file timestamp into the past
4654 right after writing to the file (e.g. tar) it is possible that
4655 the updated file modification time can be reset to the current
4656 time due to a race between processing the setattr and write RPC.
4658 Severity : enhancement
4660 Description: Bring 'lfs find' closer in line with regular Linux find.
4661 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4662 -print0 options and obtains all the needed info through the lustre
4665 Severity : enhancement
4667 Description: support up to 1024 configured devices on one node
4668 Details : change obd_dev array from statically allocated to dynamically
4669 allocated structs as they are first used to reduce memory usage
4674 Description: Flush dirty partially truncated pages during truncate
4675 Details : Immediatelly flush partially truncated pages in filter_setattr,
4676 this way we completely avoid having any pages in page cache on OST
4677 and can retire ugly workarounds during writes to flush such pages.
4682 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4683 Details : protect inode from truncation within vfs_unlink() context
4684 just take a reference before calling vfs_unlink() and release it
4685 when parent's i_sem is free.
4690 Description: last_id value checked outside lock on OST caused LASSERT failure
4691 Details : If there were multiple MDS->OST object precreate requests in
4692 flight, it was possible that the OST's last object id was checked
4693 outside a lock and incorrectly tripped an assertion. Move checks
4694 inside locks, and discard old precreate requests.
4697 Frequency : always, if extents are used on OSTs
4699 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4700 Details : OSTs using the extents format would not zero the high 16 bits of
4701 the index physical block number. This is not a problem for any
4702 OST filesystems smaller than 16TB, and no kernels support ext3
4703 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4704 new/modified files) and can be fixed for existing filesystems
4705 with e2fsprogs-1.39-cfs1.
4710 Description: import connection selection may be incorrect if timer wraps
4711 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4712 import connections to be ignored if the 32-bit jiffies counter
4713 wraps. Use a 64-bit jiffies counter.
4716 Frequency : very large clusters immediately after boot
4718 Description: LNET request buffers exhausted under heavy short-term load
4719 Details : If a large number of client requests are generated on a service
4720 that has previously never seen so many requests it is possible
4721 that the request buffer growth cannot keep up with the spike in
4722 demand. Instead of dropping incoming requests, they are held in
4723 the LND until the RPC service can accept more requests.
4726 Frequency : Sometimes during replay
4728 Description: Assertion failure in ll_local_open after replay.
4729 Details : If replay happened on an open request reply before we were able
4730 to set replay handler, reply will become not swabbed tripping the
4731 assertion in ll_local_open. Now we set the handler right after
4732 recognising of open request
4735 Frequency : very rare
4737 Description: kernel reports "badness in vsnprintf"
4738 Details : Reading from the "recovery_status" /proc file in small chunks
4739 may cause a negative length in lprocfs_obd_rd_recovery_status()
4740 call to vsnprintf() (which is otherwise harmless). Exit early
4741 if there is no more space in the output buffer.
4743 Severity : enhancement
4745 Description: clear OBD RPC statistics by writing to them
4746 Details : It is now possible to clear the OBD RPC statistics by writing
4747 to the "stats" file.
4752 Description: Client mtime is not the same on different clients after utimes
4753 Details : In some cases, the client was using the utimes() syscall on
4754 a file cached on another node. The clients now validate the
4755 ctime from the MDS + OSTs to determine which one is right.
4760 Description: Inability to activate failout mode
4761 Details : lconf script incorrectly assumed that in pythong string's numeric
4762 value is used in comparisons.
4765 Frequency : always with multiple stripes per file
4767 Description: Inefficient object allocation for mutli-stripe files
4768 Details : When selecting which OSTs to stripe files over, for files with
4769 a stripe count that divides evenly into the number of OSTs,
4770 the MDS is always picking the same starting OST for each file.
4771 Return the OST selection heuristic to the original design.
4776 Description: mount failures may take full timeout to return an error
4777 Details : Under some heavy load conditions it is possible that a
4778 failed mount can wait for the full obd_timeout interval,
4779 possibly several minutes, before reporting an error.
4780 Instead return an error as soon as the status is known.
4782 Frequency : quota enabled and large files being deleted
4784 Description: releasing more than 4GB of quota at once hangs OST
4785 Details : If a user deletes more than 4GB of files on a single OST it
4786 will cause the OST to spin in an infinite loop. Release
4787 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4792 Description: statfs data retrieved from /proc may be stale or zero
4793 Details : When reading per-device statfs data from /proc, in the
4794 {kbytes,files}_{total,free,avail} files, it may appear
4795 as zero or be out of date.
4798 Frequency : systems with MD RAID1 external journal devices
4800 Description: lconf's call to blkid is confused by RAID1 journal devices
4801 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4802 of returning all block devices that match the journal UUID.
4805 Frequency : always, for aggregate stripe size over 4GB
4807 Description: assertion fails when trying to use 4GB stripe size
4808 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4809 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4812 Frequency : always on ppc64
4814 Description: the first write on an ext3 filesystem with mballoc got stuck
4815 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4816 perform endianness conversion.
4818 ------------------------------------------------------------------------------
4820 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4822 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4823 INTEROPERATE with older versions automatically. Please read the
4824 user documentation before upgrading any part of a live system.
4825 * WARNING: Lustre networking configuration changes are required with
4826 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4829 * Support for kernels:
4830 2.6.9-22.0.2.EL (RHEL 4)
4831 2.6.5-7.244 (SLES 9)
4832 2.6.12.6 vanilla (kernel.org)
4835 Severity : enhancement
4836 Bugzilla : 7981/8208
4837 Description: Introduced Lustre Networking (LNET)
4838 Details : LNET is new networking infrastructure for Lustre, it includes
4839 a reorganized network configuration mode (see the user
4840 documentation for full details) as well as support for routing
4841 between different network fabrics. Lustre Networking Devices
4842 (LNDs) for the supported network fabrics have also been
4843 created for this new infrastructure.
4845 Severity : enhancement
4846 Description: Introduced Access control lists
4847 Details : clients can set ACLs on files and directories in order to have
4848 more fine-grained permissions than the standard Unix UGO+RWX.
4849 The MDS must be started with the "-o acl" mount option.
4851 Severity : enhancement
4852 Description: Introduced filesystem quotas
4853 Details : Administrators may now establish per-user quotas on the
4856 Severity : enhancement
4858 Description: Configuration change for the XT3
4859 The PTLLND is now used to run Lustre over Portals on the XT3
4860 The configure option(s) --with-cray-portals are no longer used.
4861 Rather --with-portals=<path-to-portals-includes> is used to
4862 enable building on the XT3. In addition to enable XT3 specific
4863 features the option --enable-cray-xt3 must be used.
4868 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4869 Details : A race between allocating a new llcd and re-getting the llcd_lock
4870 allowed another thread to grab newly-allocated llcd.
4872 Severity : enhancement
4874 Description: 2.6 OST async journal commit and locking fix to improve performance
4875 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4876 2.6 kernels are now async as they already were in 2.4 kernels so
4877 that they can commit concurrently with the network bulk transfer.
4878 For block-allocated files the filter allocation semaphore is held
4879 to avoid filesystem fragmentation during allocation. BKL lock
4880 removed for 2.6 xattr operations where it is no longer needed.
4885 Description: lconf incorrectly determined whether two IP networks could talk
4886 Details : In some more complicated routing and multiple-network
4887 configurations, lconf will avoid trying to make a network
4888 connection to a disjoint part of the IP space. It was doing the
4889 math incorrectly for one set of cases.
4894 Description: Fix for potential infinite loop processing records in an llog.
4895 Details : If an llog record is corrupted/zeroed, it is possible to loop
4896 forever in llog_process(). Validate the llog record length
4897 and skip the remainder of the block on error.
4900 Frequency : occasional (liblustre only)
4902 Description: liblustre could not open files whose last component is a symlink
4903 Details : sysio_path_walk() would incorrectly pass the open intent to
4904 intermediate path components.
4907 Frequency : rare (liblustre only with non-standard tuning)
4908 Bugzilla : 7201 (7350)
4909 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4910 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4911 after it has been released, only do this for referenced locks
4913 Severity : enhancement
4915 Description: specifying an (invalid) directory default stripe_size of -1
4916 would reset the directory default striping
4917 Details : stripe_size -1 was used internally to signal directory stripe
4918 removal, now use "all default" to signal dir stripe removal
4919 as a directory striping of "all default" is not useful
4922 Frequency : common for large clusters running liblustre clients
4924 Description: doing an ls when liblustre clients are running is slow
4925 Details : sending a glimpse AST to a liblustre client waits for every AST
4926 to time out, as liblustre clients will not respond. Since they
4927 cannot cache data we refresh the OST lock LVB from disk instead.
4929 Severity : enhancement
4931 Description: doing an ls at the same time as file IO can be slow
4932 Details : enqueue and other "small" requests can be blocked behind many
4933 large IO requests. Create a new OST IO portal for non-IO
4934 requests so they can be processed faster.
4937 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4939 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4940 created file with mode 000
4941 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4942 allow an ftruncate() as a non-root user to a file with mode 000.
4943 Lustre now respects this flag to disable mode checks when
4944 truncating a file owned by the user
4947 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4950 Description: Revoking locks from clients that went dead or catatonic might take
4952 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4953 cancellation of such locks instant on servers without waiting for
4954 any reply from clients. Clients drops these locks when cancel
4955 notification from server is received without replying.
4958 Frequency : liblustre-only, when liblustre client dies or becomes busy
4960 Description: Doing ls on Linux clients can take a long time with active
4962 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4963 granting such locks to it in the first place if possible. Locks
4964 are taken by proxy on the OST during the read or write and
4965 dropped immediately afterward. Add connect flags handling, do
4966 not grant locks to liblustre clients for glimpse ASTs.
4968 Severity : enhancement
4970 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4971 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4972 sequential reads and behaves badly with applications doing random
4973 reads. Improve it by reading ahead at least read region, and
4974 avoiding excessive large RPC for small reads.
4976 Severity : enhancement
4978 Description: Creating more than 1000 files for a single job may cause a load
4979 imbalance on the OSTs if there are also a large number of OSTs.
4980 Details : qos_prep_create() uses an OST index reseed value that is an
4981 even multiple of the number of available OSTs so that if the
4982 reseed happens in the middle of the object allocation it will
4983 still utilize the OSTs as uniformly as possible.
4988 Description: OST or MDS may oops in ping_evictor_main()
4989 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4990 but doesn't restart at beginning of obd_exports_timed list
4993 Severity : enhancement
4995 Description: improve by-nid export eviction on the MDS and OST
4996 Details : allow multiple exports with the same NID to be evicted at one
4997 time without re-searching the exports list.
5000 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5002 Description: MDS may oops in groups_free()
5003 Details : in rare race conditions a newly allocated group_info struct is
5004 freed again, and this can be NULL. The 2.4 compatibility code
5005 for groups_free() checked for a NULL pointer, but 2.6 did not.
5008 Frequency : common for liblustre clients doing little filesystem IO
5009 Bugzilla : 9352, 7313
5010 Description: server may evict liblustre clients accessing contended locks
5011 Details : if a client is granted a lock or receives a completion AST
5012 with a blocking AST already set it would not reply to the AST
5013 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5016 Frequency : lfs setstripe, only systems with more than 160 OSTs
5018 Description: unable to set striping with a starting offset beyond OST 160
5019 Details : llapi_create_file() incorrectly limited the starting stripe
5020 index to the maximum single-file stripe count.
5023 Frequency : LDAP users only
5025 Description: lconf did not handle in-kernel recovery with LDAP properly
5026 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5028 Severity : enhancement
5030 Description: bind OST threads to NUMA nodes to improve performance
5031 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5032 node and do their allocations there to localize memory access
5034 Severity : enhancement
5036 Description: llmount can determine client NID directly from Myrinet (GM)
5037 Details : the client NID code from gmnalnid was moved directly into
5038 llmount, removing the need to use this or specifying the
5039 client NID explicitly when mounting GM clients with zeroconf
5042 Frequency : if client is started with down MDS
5044 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5045 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5046 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5047 ptlrpc_import_delayed_req() because we didn't check if the
5048 request was interrupted, and we also didn't break out of the
5049 event loop if there was no timeout
5054 Description: data loss during non-page-aligned writes to a single file from
5055 both multiple nodes and multiple threads on one node at same time
5056 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5057 inconsistency led to false short-reads, that were cached and later
5058 used by ->prepare_write() to fill in partially written page,
5059 leading to data loss.
5062 Frequency : always, if lconf --abort_recovery used
5064 Description: lconf --abort_recovery fails with 'Operation not supported'
5065 Details : lconf was attempting to abort recovery on the MDT device and not
5068 Severity : enhancement
5070 Description: remove cleanup logs
5071 Details : replace lconf-generated cleanup logs with lustre internal
5072 cleanup routines. Eliminates the need for client-cleanup and
5075 Severity : enhancement
5077 Description: add support for EAs (user and system) on lustre filesystems
5078 Details : it is now possible to store extended attributes in the Lustre
5079 client filesystem, and with the user_xattr mount option it
5080 is possible to allow users to store EAs on their files also
5082 Severity : enhancement
5084 Description: Add possibility (config option) to show minimal available OST free
5086 Details : When compiled with --enable-mindf configure option, statfs(2)
5087 (and so, df) will return least minimal free space available from
5088 all OSTs as amount of free space on FS, instead of summary of
5089 free spaces of all OSTs.
5091 Severity : enhancement
5093 Description: do not expand extent locks acquired on OST-side
5094 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5095 by server: they are not cached anyway.
5098 Frequency : when mmap is used/binaries executed from Lustre
5100 Description: Unmmap pages before throwing them away from read cache.
5101 Details : llap_shrink cache now attempts to unmap pages before discarding
5102 them (if unmapping failed - do not discard). SLES9 kernel has
5103 extra checks that trigger if this unmapping is not done first.
5108 Description: lconf didn't resolve symlinks before checking to see whether a
5109 given mountpoint was already in use
5112 Frequency : when migrating failover services
5113 Bugzilla : 6395, 9514
5114 Description: When migrating a subset of services from a node (e.g. failback
5115 from a failover service node) the remaining services would
5116 time out and evict clients.
5117 Details : lconf --force (implied by --failover) sets the global obd_timeout
5118 to 5 seconds in order to quickly disconnect, but this caused
5119 other RPCs to time out too quickly. Do not change the global
5120 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5122 Severity : enhancement
5123 Frequency : if MDS is started with down OST
5124 Bugzilla : 9439,5706
5125 Description: Allow startup/shutdown of an MDS without depending on the
5126 availability of the OSTs.
5127 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5128 Add appropriate locking and lov-osc refcounts for safe
5129 cleaning. Add osc abort_inflight calls in case the
5130 synchronize never started.
5133 Frequency : occasional (Cray XT3 only)
5135 Description: root not authorized to access files in CRAY_PORTALS environment
5136 Details : The client process capabilities were not honoured on the MDS in
5137 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5138 been accessed by an authorized user then root was able to access
5139 the file on the local client also. The root user capabilities
5140 are now allowed on the MDS, as this environment has secure UID.
5143 Frequency : occasional
5145 Description: ldiskfs "too long searching" message happens too often
5146 Details : A debugging message (otherwise harmless) prints too often on
5147 the OST console. This has been reduced to only happen when
5148 there are fragmentation problems on the filesystem.
5153 Description: Division by zero in statfs when all OSCs are inactive
5154 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5155 lov_statfs divides by value returned by lov_get_stripecnt().
5159 Bugzilla : 9489, 3273
5160 Description: First write from each client to each OST was only 4kB in size,
5161 to initialize client writeback cache, which caused sub-optimal
5162 RPCs and poor layout on disk for the first written file.
5163 Details : Clients now request an initial cache grant at (re)connect time
5164 and so that they can start streaming writes to the cache right
5165 away and always do full-sized RPCs if there is enough data.
5166 If the OST is rebooted the client also re-establishes its grant
5167 so that client cached writes will be honoured under the grant.
5172 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5173 Details : Now I/O RPCs go to different portal number and (presumably) fast
5174 lock requests (and glimses) and other RPCs get their own service
5175 threads pool that should be able to service those RPCs
5178 Severity : enhancement
5180 Description: Ability to exchange lustre version between client and servers and
5181 issue warnings at client side if client is too old. Also for
5182 liblustre clients there is ability to refuse connection of too old
5184 Details : New 'version' field is added to connect data structure that is
5185 filled with version info. That info is later checked by server and
5189 Frequency : rare, liblustre only.
5190 Bugzilla : 9296, 9581
5191 Description: Two simultaneous writes from liblustre at offset within same page
5192 might proceed at the same time overwriting eachother with stale
5194 Details : I/O lock withing llu_file_prwv was released too early, before data
5195 actually was hitting the wire. Extended lock-holding time until
5196 server acknowledges receiving data.
5199 Frequency : extremely rare. Never observed in practice.
5201 Description: avoid generating lustre_handle cookie of 0.
5202 Details : class_handle_hash() generates handle cookies by incrementing
5203 global counter, and can hit 0 occasionaly (this is unlikely, but
5204 not impossible, because initial value of cookie counter is
5205 selected randonly). Value of 0 is used as a sentinel meaning
5206 "unassigned handle" --- avoid it. Also coalesce two critical
5207 sections in this function into one.
5209 Severity : enhancement
5211 Description: allow liblustre clients to delegate truncate locking to OST
5212 Details : To avoid overhead of locking, liblustre client instructs OST to
5213 take extent lock in ost_punch() on client's behalf. New connection
5214 flag is added to handle backward compatibility.
5216 Severity : enhancement
5217 Bugzilla : 4928, 7341, 9758
5218 Description: allow number of OST service threads to be specified
5219 Details : a module parameter allows the number of OST service threads
5220 to be specified via "options ost ost_num_threads={N}" in the
5221 OSS's /etc/modules.conf or /etc/modprobe.conf.
5225 Bugzilla : 6146, 9635, 9895
5226 Description: servers crash with bad pointer in target_handle_connect()
5227 Details : In rare cases when a client is reconnecting it was possible that
5228 the connection request was the last reference for that export.
5229 We would temporarily drop the export reference and get a new
5230 one, but this may have been the last reference and the export
5231 was just destroyed. Get new reference before dropping old one.
5233 Severity : enhancement
5234 Frequency : if client is started with failover MDS
5236 Description: Allow multiple MDS hostnames in the mount command
5237 Details : Try to read the configuration from all specified MDS
5238 hostnames during a client mount in case the "primary"
5241 Severity : enhancement
5243 Description: Stop sending data to evicted clients as soon as possible.
5244 Details : Check if the client we are about to send or are sending data to
5245 was evicted already. (Check is done every second of waiting,
5246 for which l_wait_event interface was extended to allow checking
5247 of exit condition at specified intervals).
5250 Frequency : rare, normally only when NFS exporting is done from client
5252 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5254 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5257 Frequency : always, if srand() is called before liblustre initialization
5259 Description: Liblustre uses system PRNG disturbing its usage by user application
5260 Details : Introduce internal to lustre fast and high-quality PRNG for
5261 lustre usage and make liblustre and some other places in generic
5262 lustre code to use it.
5264 Severity : enhancement
5265 Bugzilla : 9477, 9557, 9870
5266 Description: Verify that the MDS configuration logs are updated when xml is
5267 Details : Check if the .xml configuration logs are newer than the config
5268 logs stored on the MDS and report an error if this is the case.
5269 Request --write-conf, or allow starting with --old_conf.
5271 Severity : enhancement
5273 Description: Handle symlinks in the path when checking if Lustre is mounted.
5274 Details : Resolve intermediate symlinks when checking if a client has
5275 mounted a filesystem to avoid duplicate client mounts.
5280 Description: lconf can hit an error exception but still return success.
5281 Details : The lconf command catches the Command error exception at the top
5282 level script context and will exit with the associated exit
5283 status, but doesn't ensure that this exit status is non-zero.
5288 Description: failure of ptlrpc thread startup can cause oops
5289 Details : Starting a ptlrpc service thread can fail if there are a large
5290 number of threads or the server memory is very fragmented.
5291 Handle this without oopsing.
5294 Frequency : always, only if liblustre and non-default acceptor port was used
5296 Description: liblustre cannot connect to servers with non-default acceptor port
5297 Details : tcpnal_set_default_params() was not called and was therefore
5298 ignoring the environment varaible TCPNAL_PORT, as well as other
5299 TCPNAL_ environment variables
5304 Description: two objects could be created on the same OST for a single file
5305 Details : If an OST is down, in some cases it was possible to create two
5306 objects on a single OST for a single file. No problems other
5307 than potential performance impact and spurious error messages.
5311 Bugzilla : 5681, 9562
5312 Description: Client may oops in ll_unhash_aliases
5313 Details : Client dcache may become inconsistent in race condition.
5314 In some cases "getcwd" can fail if the current directory is
5320 Description: Inode refcounting problems in NFS export code
5321 Details : link_raw functions used to call d_instantiate without obtaining
5322 extra inode reference first.
5326 Bugzilla : 9942, 9903
5327 Description: Referencing freed requests leading to crash, memleaks with NFS.
5328 Details : We used to require that call to ll_revalidate_it was always
5329 followed by ll_lookup_it. Also with revalidate_special() it is
5330 possible to call ll_revalidate_it() twice for the same dentry
5331 even if first occurence returned success. This fix changes semantic
5332 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5333 reference on a request referred from the intent.
5334 ll_intent_release() then releases such a request.
5337 Frequency : rare, normally benchmark loads only
5339 Description: unlinked inodes were kept in memory on the client
5340 Details : If a client is repeatedly creating and unlinking files it
5341 can accumulate a lot of stale inodes in the inode slab cache.
5342 If there is no other client load running this can cause the
5343 client node to run out of memory. Instead flush old inodes
5344 from client cache that have the same inode number as a new inode.
5347 Frequency : SLES9 2.6.5 kernel and long filenames only
5348 Bugzilla : 9969, 10379
5349 Description: utime reports stale NFS file handle
5350 Details : SLES9 uses out-of-dentry names in some cases, which confused
5351 the lustre dentry revalidation. Change it to always use the
5355 Frequency : rare, unless heavy write-truncate concurrency is continuous
5356 Bugzilla : 4180, 6984, 7171, 9963, 9331
5357 Description: OST becomes very slow and/or deadlocked during object unlink
5358 Details : filter_destroy() was holding onto the parent directory lock
5359 while truncating+unlinking objects. For very large objects this
5360 may block other threads for a long time and slow overall OST
5361 responsiveness. It may also be possible to get a lock ordering
5362 deadlock in this case, or run out of journal credits because of
5363 the combined truncate+unlink. Solution is to do object truncate
5364 first in one transaction without parent lock, and then do the
5365 final unlink in a new transaction with the parent lock. This
5366 reduces the lock hold time dramatically.
5369 Frequency : rare, 2.4 kernels only
5371 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5372 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5373 before unmounting their filesystem in order to avoid deadlock.
5374 The kernel_locked() function in 2.4 kernels only checks whether
5375 the kernel lock is held, not whether it is this process that is
5376 holding it as 2.6 kernels do.
5381 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5382 Details : The client ptlrpc code may be trying to reconnect to a down
5383 server before a previous connection attempt has timed out.
5384 Increase the reconnect interval to be longer than the connection
5385 timeout interval to avoid sending duplicate connections to
5389 Frequency : echo_client brw_test command
5391 Description: fix echo_client to work with OST preallocated code
5392 Details : OST preallocation code (5137) didn't take echo_client IO path
5393 into account: echo_client calls filter methods outside of any
5394 OST thread and, hence, there is no per-thread preallocated
5395 pages and buffers to use. Solution: hijack pga pages for IO. As
5396 a byproduct, this avoids unnecessary data copying.
5400 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5401 Description: Client can oops in mdc_commit_close() after open replay
5402 Details : It was possible for the MDS to return an open request with no
5403 transaction number in mds_finish_transno() if the client was
5404 evicted, but without actually returning an error. Clients
5405 would later try to replay that open and may trip an assertion
5406 Simplify the client close codepath, and always return an error
5407 from the MDS in case the open is not successful.
5410 Frequency : rare, 2.6 OSTs only
5412 Description: OST may deadlock under high load on fragmented files
5413 Details : If there was a heavy load and highly-fragmented OST filesystems
5414 it was possible to have all the OST threads deadlock waiting on
5415 allocation of biovecs, because the biovecs were not released
5416 until the entire RPC IO was completed. Instead, release biovecs
5417 as soon as they are complete to ensure forward IO progress.
5419 Severity : enhancement
5421 Description: Support for specifying external journal device at mount
5422 Details : If an OST or MDS device is formatted with an external journal
5423 device, this device major/minor is stored in the ext3 superblock
5424 and may not be valid for failover. Allow detecting and
5425 specifying the external journal at mount time.
5430 Description: Mounting an MDS with pending unlinked files may cause oops
5431 Details : target_finish_recovery() calls mds_postrecov() which returned
5432 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5433 considers this an error and immediately begins cleaning up the
5434 lov, just after starting the mds_lov process
5436 Severity : enhancement
5438 Description: Implement 'lfs df' to report actual free space on per-OST basis
5439 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5440 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5441 usage of objects; '-h' to report in human readable format.
5443 ------------------------------------------------------------------------------
5445 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5452 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5453 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5454 at startup during orphan unlinking. Instead, skip buddy update
5455 before setup, as it will be regenerated after recovery is complete.
5460 Description: If an OST is inactive, its locks might reference stale inodes.
5461 Details : lov_change_cbdata() must iterate over all namespaces, even if
5462 they are inactive to clear inode references from the lock.
5464 Severity : enhancement
5465 Frequency : occasional, if non-standard max_dirty_mb used
5467 Description: Client will block write RPCs if not enough grant
5468 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5469 then the client will block writes while waiting for another RPC
5470 to complete instead of consuming its dirty limit. With change
5471 we get improved performance when max_dirty_mb is small.
5473 Severity : enhancement
5474 Bugzilla : 3389, 6253
5475 Description: Add support for supplementary groups on the MDS.
5476 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5477 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5478 lookups for user supplementary groups into a cache.
5482 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5483 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5484 for liblustre O_EXCL is enforced by MDS.
5488 Description: readdir never returns NULL in liblustre.
5489 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5490 dentry in struct dirent.
5494 Frequency : liblustre only, depends on application IO pattern
5495 Description: liblustre clients evicted if not contacting servers
5496 Details : Don't put liblustre clients into the ping_evictor list, so
5497 they will not be evicted by the pinger ever.
5499 Severity : enhancement
5501 Description: Add ability to evict clients by NID from MDS.
5502 Details : By echoing "nid:$NID" string into
5503 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5504 $NID would be instantly evicted from this MDS and from all active
5505 OSTs connected to it.
5509 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5510 Details : lookup_it_finish() used to query file size from OSTs that was not
5515 Description: service threads change working directory to that of init
5516 Details : Starting lustre service threads may pin the working directory
5517 of the parent thread, making that filesystem busy. Threads
5518 now change to the working directory of init to avoid this.
5522 Frequency : during shutdown only
5523 Description: shutdown with a failed MDS or OST can cause unmount to hang
5524 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5529 Frequency : 2.6 only
5530 Description: chmod/chown may include an extra supplementary group
5531 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5532 supplementary group and if none is specified this is used.
5535 Bugzilla : 5479 (6816)
5537 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5538 Details : If lookup is in progress on a file that is unlinked we might try
5539 to revalidate the inode and fail in revalidate after lookup is
5540 complete and ll_file_open() enqueues the open again but
5541 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5544 Frequency : always, if lconf --abort_recovery used
5546 Description: lconf --abort_recovery fails with 'Operation not supported'
5547 Details : lconf was attempting to abort recovery on the MDT device and not
5550 ------------------------------------------------------------------------------
5552 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5557 Frequency : rare (only unsupported configurations with a node running as an
5559 Bugzilla : 6514, 5137
5560 Description: Mounting a Lustre file system on a node running as an OST could
5562 Details : OSTs now preallocates memory needed to write out data at
5563 startup, instead of when needed, to avoid having to
5564 allocate memory in possibly low memory situations.
5565 Specifically, if the file system is mounted on on OST,
5566 memory pressure could force it to try to write out data,
5567 which it needed to allocate memory to do. Due to the low
5568 memory, it would be unable to do so and the node would
5569 become unresponsive.
5571 Severity : enhancement
5573 Description: Addition of lconf --service command line option
5574 Details : lconf now accepts a '--service <arg>' option, which is
5575 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5577 Severity : enhancement
5579 Description: Failover mode is now the default for OSTs.
5580 Details : By default, OSTs will now run in failover mode. To return to
5581 the old behaviour, add '--failout' to the lmc line for OSTs.
5583 Severity : enhancement
5585 Description: Health checks are now provided for MDS and OSTs
5586 Details : Additional detailed health check information on MSD and OSTs
5587 is now provided through the procfs health_check value.
5590 Frequency : occasional, depends on IO load
5592 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5593 after numerous create/delete cycles
5594 Details : The ext3 inode allocation policy would not allocate new inodes
5595 very well on the OSTs because there are no new directories
5596 being created. Instead we look for groups with free space if
5597 the parent directories are nearly full.
5602 Description: Network or server problems during mount may cause partially
5603 mounted clients instead of returning an error.
5604 Details : The config llog parsing code may overwrite the error return
5605 code during mount error handling, returning success instead
5611 Description: MDS can fail to allocate large reply buffers
5612 Details : After long uptimes the MDS can fail to allocate large reply
5613 buffers (e.g. zconf client mount config records) due to memory
5614 fragmentation or consumption by the buffer cache. Preallocate
5615 some large reply buffers so that these replies can be sent even
5616 under memory pressure.
5620 Frequency : rare (liblustre)
5621 Description: fsx running with liblustre complained that using truncate() to
5622 extend the file doesn't work. This patch corrects that issue.
5623 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5624 ATTR_SIZE and lsm use in llu_setattr_raw.
5628 Frequency : rare, only 2.6 kernels
5629 Description: Unusual file access patterns on the MDS may result in inode
5630 data being lost in very rare circumstances.
5631 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5632 code in the 2.6 kernel caused the inode and/or EA data not to
5633 be read from disk, causing single-file corruption.
5637 Frequency : rare, only 2.6 filesystems using extents
5638 Description: Heavy concurrent write and delete load may cause data corruption.
5639 Details : It was possible under high-load situations to have an extent
5640 metadata block in the block device cache from a just-unlinked
5641 file overwrite a newly-allocated data block. We now unmap any
5642 metadata buffers that alias just-allocated data blocks.
5646 Frequency : filesystems with default stripe_count larger than 77
5647 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5648 Details : lconf specifies an inode size of 4096 bytes when the default
5649 stripe_count is larger than 77. This conflicts with the default
5650 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5651 this case to avoid pinning too much memory for large EAs.
5653 ------------------------------------------------------------------------------
5655 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5660 Frequency : rare (extremely heavy IO load with hundreds of clients)
5662 Description: Client is evicted, gets IO error writing to file
5663 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5664 caused clients to be evicted for AST timeouts. The fixes in
5665 bug 5192 mean we no longer need to have such short AST timeouts
5666 so ldlm_timeout has been increased.
5669 Frequency : occasional during --force or --failover shutdown under load
5670 Bugzilla : 5949, 4834
5671 Description: Server oops/LBUG if stopped with --force or --failover under load
5672 Details : a collection of import/export refcount and cleanup ordering
5673 issues fixed for safer force cleanup
5676 Frequency : only filesystems larger than 120 OSTs
5677 Bugzilla : 5990, 6223
5678 Description: lfs getstripe would oops on a very large filesystem
5679 Details : lov_getconfig used kfree on vmalloc'd memory
5682 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5683 Bugzilla : 6242, 6243
5684 Description: reading from files that had been truncated to a non-zero size
5685 but never opened returned no data
5686 Details : ll_file_read() reads zeros from no-object files to EOF
5691 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5692 Details : There's little harm in aborting MDS/OSS recovery and letting it
5693 try again, so I removed the LASSERT and return an error instead.
5695 Severity : enhancement
5697 Description: New debugging infrastructure for tracking down data corruption
5698 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5699 (b) cover more of the client-side code path, and (c) try to narrow
5700 down where problems occurred
5704 Bugzilla : 3819, 4364, 4397, 6313
5705 Description: Racing close and eviction MDS could cause assertion in mds_close
5706 Details : It was possible to get multiple mfd references during close and
5707 client eviction, leading to one thread referencing a freed mfd.
5709 Severity: : enhancement
5710 Bugzilla : 3262, 6359
5711 Description: Attempts to reconnect to servers are now more aggressive.
5712 Details : This builds on the enhanced upcall-less recovery that was added
5713 in 1.4.2. When trying to reconnect to servers, clients will
5714 now try each server in the failover group every 10 seconds. By
5715 default, clients would previously try one server every 25 seconds.
5720 Description: After recovery, certain operations trigger a failed
5721 assertion on a client.
5722 Details : Failing over an mds, using lconf -d --failover, while a
5723 client was doing a readdir() call would cause the client to
5724 LBUG after recovery completed and the readdir() was resent.
5726 Severity : enhancement
5728 Description: Default groups are now added by lconf
5729 Details : You can now run lconf --group <servicename> without having to
5730 manually add groups with lmc.
5733 Frequency : occasional
5735 Description: Nodes with an elan id of 0 trigger a failed assertion
5738 Frequency : always when accessing e.g. tty/console device nodes
5740 Description: tty and some other devices nodes cannot be used on lustre
5741 Details : file's private_data field is used by device data and lustre
5742 values in there got lost. New field was added to struct file to
5743 store fs-specific private data.
5746 Frequency : when exporting Lustre via NFS
5748 Description: NFSD failed occasionally when looking up a path component
5749 Details : NFSD is looking up ".." which was broken in ext3 directories
5750 that had grown large enough to become hashed.
5753 Frequency : Clusters with multiple interfaces not on the same subnet
5755 Description: Nodes will repeatedly try to reconnect to an interface which it
5756 cannot reach and report an error to the log.
5757 Details : Extra peer list entries will be created by lconf with some peers
5758 unreachable. lconf now validates the peer before adding it.
5761 Frequency : Only if a default stripe is set on the filesystem root.
5763 Description: Setting a default stripe on the filesystem root prevented the
5764 filesystem from being remounted.
5765 Details : The client was sending extra request flags in the root getattr
5766 request and did not allocate a reply buffer for the dir EA.
5769 Frequency : occasional, higher if lots of files are accessed by one client
5770 Bugzilla : 6159, 6097
5771 Description: Client trips assertion regarding lsm mismatch/magic
5772 Details : While revalidating inodes the VFS looks up inodes with ifind()
5773 and in rare cases can find an inode that is being freed.
5774 The ll_test_inode() code will free the lsm during ifind()
5775 when it finds an existing inode and then the VFS later attaches
5776 this free lsm to a new inode.
5780 Bugzilla : 6422, 7030
5781 Description: MDS deadlock between mkdir and client eviction
5782 Details : Creating a new file via mkdir or mknod (starting a transaction
5783 and getting the ns lock) can deadlock with client eviction
5784 (gets ns lock and trying to finish a synchronous transaction).
5787 Frequency : occasional
5788 Description: While starting a server, the fsfilt_ext3 module could not be
5790 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5791 kernels. Previously, lconf would still use the ext3 name
5792 when trying to load modules. Now, it will correctly use
5793 ext3 on 2.4 and ldiskfs on 2.6.
5795 Severity : enhancement
5796 Description: The default stripe count has been changed to 1
5797 Details : The interpretation of the default stripe count (0, to lfs
5798 or lmc) has been changed to mean striping across a single
5799 OST, rather than all available. For general usage we have
5800 found a stripe count of 1 or 2 works best.
5802 Severity : enhancement
5803 Description: Add support for compiling against Cray portals.
5804 Details : Conditional compiling for some areas that are different
5808 Frequency : occasional
5809 Bugzilla : 6409, 6834
5810 Description: Creating files with an explicit stripe count may lead to
5811 a failed assertion on the MDS
5812 Details : If some OSTs are full or unavailable, creating files may
5813 trigger a failed assertion on the MDS. Now, Lustre will
5814 try to use other servers or return an error to the
5818 Frequency : occasional
5820 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5821 caused lock timeout during readahead (since 1.4.2).
5822 Details : Processes doing readahead might match a lock that hasn't been
5823 granted yet if there are overlapping and conflicting lock
5824 requests. The readahead process waits on ungranted lock
5825 (original lock is CBPENDING), while OST waits for that process
5826 to cancel CBPENDING read lock and eventually evicts client.
5828 Severity : enhancement
5830 Description: Initial enabling of flock support for clients
5831 Details : Implements fcntl advisory locking and file status functions.
5832 This feature is provided as an optional mount flag (default
5833 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5834 locking are implemented yet, and those that are are not guaranteed
5835 to be completely correct in production environments.
5836 mount -t lustre -o [flock|noflock] ...
5839 Frequency : occasional
5841 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5842 assertion in the ext3 JBD (journaling) layer.
5843 Details : The b_committed_data struct is protected by the big kernel lock
5844 in 2.4 kernels, serializing journal_commit_transaction() and
5845 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5846 access to this struct. In 2.6 kernels there is finer grained
5847 locking to improve SMP performance of the JBD layer.
5851 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5856 Description: merge and cleanup kernel patches.
5859 -----------------------------------------------------------------------------
5861 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5863 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5864 versions of Lustre. Please update all servers and clients to
5865 version 1.4.2 or later at the same time. You must also run
5866 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5867 to update the configuration logs.
5869 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5870 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5871 - swab llog records of type '0' so we get proper header size/idx (5861)
5872 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5873 - fix rename of one directory over another leaking an inode (5953)
5874 - avoid SetPageDirty on 2.6 (5981)
5875 - don't re-add just-being-destroyed locks to the waiting list (5653)
5876 - when creating new directories, inherit the parent's custom
5877 striping settings if present parent (3048)
5878 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5879 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5880 - don't hold client locks on temporary worklist from l_lru (5666)
5881 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5882 - automatically evict dead clients (5921)
5883 - Update file size properly in create+truncate+fstat case (6196)
5884 - Do not unhash mountpoint dentries, do not allow removal of
5886 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5887 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5888 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5889 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5890 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5891 - Correctly handle reads of files with no objects (6243)
5892 - lctl recover will also mark a device active if deactivate used (5933)
5894 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5895 - allow --write-conf on an MDS with different nettype than client (5619)
5896 - don't write config llogs to MDS for mounts not from that MDS (5617)
5897 - lconf should create multiple TCP connections from a client (5201)
5898 - init scripts are now turned off by default; run chkconfig --on
5899 lustre and chkconfig --on lustrefs to use them
5900 - upcalls are no longer needed for clients to recover to failover
5902 - add --abort-recovery option to lconf to abort recovery on device
5904 - add support for an arbitrary number of OSTs (3026)
5905 - Quota support protocol changes.
5906 - forward compatibility changes to wire structs (6007)
5907 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5908 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5909 - improved "lustre status" script
5910 - initialize blocksize for non-regular files (6062)
5911 - added --disable-server and --disable-client configure options (5782)
5912 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5913 - Vanilla 2.4.29 support
5914 - increase maximum number of obd devices to 520 (6242)
5915 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5916 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5917 SLES 9 2.6.5 kernels
5918 - limit stripes per file to 160 (the maximum EA size) (6093)
5920 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5923 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5924 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5925 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5926 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5927 - small ext3 extents cleanups and fixes (5733)
5928 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5929 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5930 - avoid needless client->OST connect, fix handle mismatch (5317)
5931 - fix DLM error path that led to out-of-sync client, long delays (5779)
5932 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5933 - fix several locking issues related to i_size (5492,5624,5654,5672)
5934 - don't move pending lock onto export if it is already evicted (5683)
5935 - fix kernel oops when creating .foo in unlinked directory (5548)
5936 - fix deadlock in obdfilter statistics vs. object create (5811)
5937 - use time_{before,after} to avoid timer jiffies wrap (5882)
5938 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5939 - Do not leak request if server was not able to process it (5154)
5940 - If mds_open unable to find parent dir, make that negative lookup(5154)
5941 - don't create new directories with extent-mapping (5909, 5936)
5943 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5944 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5945 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5946 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5947 - support for RHEL4 kernel on i686 (5773)
5948 - provide error messages when incompatible logs are encountered (5898)
5950 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5951 * version 1.4.0.10 (1.4.1 release candidate 1)
5953 - don't keep a lock reference when lock is not granted (4238)
5954 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5955 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5956 - drop import inflight refcount on signal_completed_replay error (5255)
5957 - unlock page after async write error during send (3677)
5958 - handle missing objects in filter_preprw_read properly (5265)
5959 - no transno return for symlink open, don't save no-trasno open (3440)
5960 - don't try to complete elan receive that already failed (4012)
5961 - free RPC server reply state on error (5406)
5962 - clean up thread from ptlrpc_start_thread() on error (5160)
5963 - readahead could read extra page into cache that wasn't ejected (5388)
5964 - prevent races in class_attach/setup/cleanup/detach (5260)
5965 - don't dereference de->d_inode after l_dput of de (5458)
5966 - use "int" for stripe value returned from lock_to_stripe (5544)
5967 - mballoc allocation and error-checking fixes in 2.6 (5504)
5968 - block device patches to fix I/O request sizes in 2.6 (5482)
5969 - look up hostnames for IB nals (5602)
5970 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5971 - don't start multiple acceptors for the same port (5277)
5972 - fix incorrect LASSERT in mds_getattr_name (5635)
5973 - export a proc file for general "ping" checking (5628)
5974 - fix "lfs check" to not block when the MDS is down (5628)
5976 - service request history (4965)
5977 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5978 - create an "evict_client" /proc entry on OSTs, like the MDS has
5979 - fix mount usage message, return errors per mount(8) (5168)
5980 - change grep [] to grep "[]" in tests so they work in more UMLs
5981 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5982 - remove ancient LOV_MAGIC_V0 EA support (5047)
5983 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5984 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5985 - allow readahead allocations to fail when low on memory (5383)
5986 - mmap locking landed again, after considerable improvement (2828)
5987 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5989 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5992 - send OST transaction number in read/write reply to free req (4966)
5993 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5994 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5995 - the watchdog thread now runs as interruptible (5246)
5996 - flock/lockf fixes (but it's still disabled, pending 5135)
5997 - don't use EXT3 constants in llite code (5094)
5998 - memory shortage at startup could cause assertion (5176)
6000 - reorganization of lov code
6001 - single portals codebase
6003 - add extents/mballoc support (5025)
6004 - direct I/O reads in the obdfilter (4048)
6005 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6007 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6010 - send OST transaction number in read/write reply to free req (4966)
6011 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6012 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6013 - the watchdog thread now runs as interruptible (5246)
6014 - handle missing objects in filter_preprw_read properly (5265)
6015 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6016 - drop import inflight refcount on signal_completed_replay error (5255)
6017 - unlock page after async write error during send (3677)
6018 - return original error code on reconstructed replies (3761)
6019 - no transno return for symlink open, don't save no-trasno open (3440)
6021 - add pid to ldlm debugging output (4922)
6022 - bump the watchdog timeouts -- we can't handle 30sec yet
6023 - extra debugging for orphan dentry/inode bug (5259)
6025 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6028 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6029 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6030 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6031 - fix NULL dereference in /proc/sys/portals/routes (4827)
6032 - allow failed mdc_close() operations to be interrupted (4561)
6033 - stop precreate on OST before MDS would time out on it (4778)
6034 - don't send partial-page writes before EOF from client (4410)
6035 - discard client grant for sub-page writes on large-page clients (4520)
6036 - don't free dentries not owned by NFS code, check generation (4806)
6037 - fix lsm leak if mds_create_objects() fails (4801)
6038 - limit debug_daemon file size, always print CERROR messages (4789)
6039 - use transno after validating reply (3892)
6040 - process timed out requests if import state changes (3754)
6041 - update mtime on OST during writes, return in glimpse (4829)
6042 - add mkfsoptions to LDAP (4679)
6043 - use ->max_readahead method instead of zapping global ra (5039)
6044 - don't interrupt __l_wait_event() during strace
6046 - add software watchdogs to catch hung threads quickly (4941)
6047 - make lustrefs init script start after nfs is mounted
6048 - fix CWARN/ERROR duplication (4930)
6049 - return async write errors to application if possible (2248)
6050 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6051 - print NAL number in %x format (4645)
6052 - update barely-supported suse-2.4.21-171 series (4842)
6053 - support for sles 9 %post scripts
6054 - support for building 2.6 kernel-source packages
6055 - support for sles km_* packages
6057 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6060 - ignore -ENOENT errors in osc_destroy (3639)
6061 - notify osc create thread that OSC is being cleaned up (4600)
6062 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6063 - reconstruct ost_handle() like mds_handle() (4657)
6064 - create a new thread to do import eviction to avoid deadlock (3969)
6065 - let lconf resolve symlinked-to devices (4629)
6066 - don't unlink "objects" from directory with default EA (4554)
6067 - hold socknal file ref over connect in case target is down (4394)
6068 - allow more than 32000 subdirectories in a single directory (3244)
6069 - fix blocks count for O_DIRECT writes (3751)
6070 - OST returns ENOSPC from object create when no space left (4539)
6071 - don't send truncate RPC if file size isn't changing (4410)
6072 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6073 - bind to privileged port in socknal and tcpnal (3689)
6075 - rate limit CERROR/CWARN console message to avoid overload (4519)
6076 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6077 - basic mmap support (3918)
6078 - kernel patch series update from b1_4 (4711)
6080 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6083 - avoid crash during MDS cleanup with OST shut down (2775)
6084 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6085 - don't use bad inodes on the MDS (3744)
6086 - dynamic object preallocation to improve recovery speed (4236)
6087 - don't hold spinlock over lock dumping or change debug flags (4401)
6088 - don't zero obd_dev when it is force cleaned (3651)
6089 - print grants to console if they go negative (4431)
6090 - "lctl deactivate" will stop automatic recovery attempts (3406)
6091 - look for existing locks in ldlm_handle_enqueue() (3764)
6092 - don't resolve lock handle twice in recovery avoiding race (4401)
6093 - revalidate should check working dir is a directory (4134)
6095 - don't always mark "slow" obdfilter messages as errors (4418)
6097 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6100 - don't close LustreDB during write_conf until it is done (3860)
6101 - fix typo in lconf for_each_profile (3821)
6102 - allow dumping logs from multiple threads at one time (3820)
6103 - don't allow multiple threads in OSC recovery (3812)
6104 - fix debug_size parameters (3864)
6105 - fix mds_postrecov to initialize import for llog ctxt (3121)
6106 - replace config semaphore with spinlock (3306)
6107 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6108 - don't allow enqueue to complete on a destroyed export (3822)
6109 - down write_lock before checking llog header bitmap (3825)
6110 - recover from lock replay timeout (3764)
6111 - up llog sem before sending rpc (3652)
6112 - reduce ns lock hold times when setting kms (3267)
6113 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6114 - fix NULL deref and obd_dev leak on setup error (3312)
6115 - replace some LBUG about llog ops with error handling (3841)
6116 - don't match INVALID dentries from d_lookup and spin (3784)
6117 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6118 - fix invalid assertion in ptlrpc_set_wait (3880)
6120 - add libwrap support for the TCP acceptor (3996)
6121 - add /proc/sys/portals/routes for non-root route listing (3994)
6122 - allow setting MDS UUID in .xml (2580)
6123 - print the stack of a process that LBUGs (4228)
6125 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6128 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6129 - make sure to unset replay flag from failed open requests (3440)
6130 - if default stripe count is 0, use OST count for inode size (3636)
6131 - update parent mtime/ctime on client for create/unlink (2611)
6132 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6133 - free recovery state on server during a forced cleanup (3571)
6134 - unregister_reply for resent reqs (3063)
6135 - loop back devices mounting and status check on 2.6 (3563)
6136 - fix resource-creation race that can provoke i_size == 0 (3513)
6137 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6138 - more debugging for page-accounting assertion (3746)
6139 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6140 - avoid deadlock after precreation failure (3758)
6141 - fix race and lock order deadlock in orphan handling (3450, 3750)
6142 - add validity checks when grabbing inodes from l_ast_data (3599)
6144 - add /proc/.../recovery_status to obdfilter (3428)
6145 - lightweight CDEBUG infrastructure, debug daemon (3668)
6146 - change default OSC RPC parameters to be better on small clusters
6147 - turn off OST read cache for files smaller than 32MB
6148 - install man pages and include them in rpms (3100)
6149 - add new init script for (un)mounting lustre filesystems (2593)
6150 - run chkconfig in %post for init scripts (3701)
6151 - drop scimac NAL (unmaintained)
6153 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6156 - clean kiobufs before and after use (3485)
6157 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6158 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6159 - change init script to fail more clearly if not run as root (1528)
6160 - allow clients to reconnect during replay (1742)
6161 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6162 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6163 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6164 - fix cond_resched (3554)
6165 - only evict unfinished clients after recovery (3515)
6166 - allow bulk resend, prevent data loss (3570)
6167 - dynamic ptlrpc request buffer allocation (2102)
6168 - don't allow unlinking open directory if it isn't empty (2904)
6169 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6170 - remove extraneous obd dereference causing LASSERT failure (3334)
6171 - don't use get_cycles() when creating temp. files on the mds (3156)
6172 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6173 - handle EEXIST for set-stripe, set proper directory name (3336)
6175 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6176 - fix ksocknal_fmb_callback() error messages (2918)
6178 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6181 - don't copy lvb into (possibly NULL) reply on error (2983)
6182 - don't deref dentry after dput, don't free lvb on error (2922)
6183 - use the kms to determine writeback rpc length (2947)
6184 - increment oti_logcookies when osc is inactive (2948)
6185 - update client's i_blocks count via lvb messages (2543)
6186 - handle intent open/close of special files properly (1557)
6187 - mount MDS with errors=remount-ro, like obdfilter (2009)
6188 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6189 - don't use cancelling-locks' kms values (2947)
6190 - use highest lock extent for kms, not last one (2925)
6191 - don't dereference ERR_PTR() dentry in error handling path (3107)
6192 - fix thread race in portals_debug_dumplog() (3122)
6193 - create lprocfs device entries at setup instead of at attach (1519)
6194 - common AST error handler, don't evict client on completion race (3145)
6195 - zero nameidata in detach_mnt in 2.6 (3118)
6196 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6197 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6198 - initialize RPC timeout timer earlier for 2.6 (3219)
6199 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6200 - print nal/nid for unknown nid (3258)
6201 - additional checks for oscc recovery before doing precreate (3284)
6202 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6203 - don't crash in mdc_close for bad permissions on open (3285)
6204 - zero i_rdev for non-device files (3147)
6205 - clear page->private before handing to FS, better assertion (3119)
6206 - tune the read pipeline (3236)
6207 - fix incorrect decref of invalidated dentry (2350)
6208 - provide read-ahead stats and refine rpc in flight stats (3328)
6209 - don't hold journal transaction open across create RPC (3313)
6210 - update atime on MDS at close time (3265)
6211 - close LDAP connection when recovering to avoid server load (3315)
6212 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6213 - don't leak open file on MDS after open resend (3325)
6214 - serialize filter_precreate and filter_destroy_precreated (3329)
6215 - loop device shouldn't call sync_dev() for nul device (3092)
6216 - clear page cache after eviction (2766)
6217 - resynchronize MDS->OST in background (2824)
6218 - refuse to mount the same filesystem twice on same mountpoint (3394)
6219 - allow llmount to create routes for mounting behind routers (3320)
6220 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6221 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6222 - fix rare problem with rename on htree directories (3417)
6224 - allow default OST striping configuration per directory (1414)
6225 - fix compilation for qswnal for 2.6 kernels (3125)
6226 - increase maximum number of MDS request buffers for large systems
6227 - change liblustreapi to be useful for external progs like lfsck (3098)
6228 - increase local configuration timeout for slow disks (3353)
6229 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6231 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6234 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6235 - don't overwrite extent policy data in reply if lock was blocked (2901)
6236 - drop filter export grants atomically with removal from device (2663)
6237 - del obd_self_export from work_list in class_disconnect_exports (2908)
6238 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6239 - swab reply message in mdc_close, other PPC fixes (2464)
6240 - fix destroying of named logs (2325)
6241 - overwrite old logs when running lconf --write_conf (2264)
6242 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6243 - fix race in target_handle_connect (2898)
6244 - mds_reint_create() should take same inode create lock (2926)
6245 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6246 - don't close files for self_export to avoid uninitialized obd (2936)
6247 - allow MDS with the same name as client node (2939)
6248 - hold dentry reference for closed log files for unlink (2325)
6249 - reserve space for all logs during transactions (2059)
6250 - don't evict page beyond end of stripe extent (2925)
6251 - don't oops on a deleted current working directory (2399)
6252 - handle hard links to targets without a parent properly (2517)
6253 - don't dereference NULL lock when racing during eviction (2867)
6254 - don't grow lock extents when lots of conflicting locks (2919)
6256 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6259 - account for cache space usage on clients to avoid data loss (974)
6260 - lfsck support in lustre kernel code (2349)
6261 - reduce journal credits needed for BRW writes (2370)
6262 - orphan handling to avoid losing space on client/server crashes
6263 - ptlrpcd can be blocked, stopping ALL progress (2477)
6264 - use lock value blocks to assist in proper KMS, faster stat (1021)
6265 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6266 - recovery for initial connections (2355)
6267 - fixes for mds_cleanup_orphans (1934)
6268 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6269 - block all file creations until orphan recovery completes (1901)
6270 - client remove rq_connection from request struct (2423)
6271 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6272 - recovery timer race (2670)
6273 - mdc_close recovey bug (2532)
6274 - ptlrpc cleanup bug (2710)
6275 - mds timeout on local locks (2588)
6276 - namespace lock held during RPCs (2431)
6277 - handle interrupted sync write properly (2503)
6278 - don't try to handle a message that hasn't been replied to (2699)
6279 - client assert failure during cleanup after abort recovery (2701)
6280 - leak mdc device after failed mount (2712)
6281 - ptlrpc_check_set allows timedout requests to complete (2714)
6282 - wait for inflight reqs when ptlrpcd finishes (2710)
6283 - make sure unregistered services are removed from the srv_list
6284 - reset bulk XID's when resending them (caught by 1138 test)
6285 - unregister_bulk after timeout
6286 - fix lconf error (2694)
6287 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6288 - readahead locks pages, leaves pending causing memory pressure (2673)
6289 - increase OST request buffers to 4096 on large machines (2729)
6290 - fix up permission of existing directories in simple_mkdir (2661)
6291 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6292 - don't assign transno to errored transactions (2742)
6293 - don't delete objects on OST if given a bogus objid from MDS (2751)
6294 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6295 - if rq_no_resend, then timeout request after recovery (2432)
6296 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6297 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6298 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6299 - MDS recovery shouldn't delete objects that it has given out (2730)
6300 - if enqueue arrives after completion, don't clobber LVB (2819)
6301 - don't unlock pages twice when trigger_group_io returns error (2814)
6302 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6303 - don't write pages to disk if there was an error (1450)
6304 - don't ping imports that have recovery disabled (2676)
6305 - take buffered bytes into account when balancing socknal conn (2817)
6306 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6307 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6308 - remove little-endian swabbing of llog records (1987)
6309 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6310 - retry reposting request buffers if they fail (1191)
6311 - grow extent at grant time to avoid granting a revoked lock (2809)
6312 - lock revoke doesn't evict page if covered by a second lock (2765)
6313 - disable VM readahead to avoid reading outside lock extents (2805)
6315 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6316 - updated kernel patches for hp-2.4.20 kernel (2681)
6318 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6321 - fix truncated write corruption (2366)
6322 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6324 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6325 - ensure len(uuid) < 37 in lmc (1171)
6326 - fix ia64 OOPS in llog_test (2255)
6327 - zero end of page at obdfilter for partial page writes (2648)
6328 - don't leave stale dentries around after renames (bug 2428)
6329 - fix timeouts when evicting a client with a single lock held (2642)
6330 - set deadline for the initial HELLO message to drain (2634)
6331 - print out dotted-quad IP addresses in the socknal (2302)
6333 - additional debugging for MDS client eviction problem (2443)
6334 - fix mkfsoptions support for osts (2603, 2604)
6336 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6339 - add series for the vanilla 2.6.0 kernel
6340 - add series for the vanilla 2.4.24 kernel
6341 - add series for a cray x86/64 UL kernel drop
6342 - fix xattr patches for the vanilla 2.4.19 series
6344 - generate true UUIDs in lmc (1171)
6345 - have portals stack dumping break in UML (2466)
6346 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6347 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6348 - wake callback waiting threads on client eviction (2460)
6349 - Add --ptldebug and --subsystem to lmc (1719)
6350 - update assertion to allow safe interrupt allocation
6351 - set rq_no_resend for cancel requests (2432)
6352 - recalculate ptlrpcd timeout after resend (2494)
6353 - call vfs_rmdir when removing pending directories (2368)
6354 - fix renaming a file to itself (2429)
6355 - lmc creates a default one-stripe lov (2454)
6356 - expand procfs space to handle large clusters (2326)
6357 - increase UML stack to avoid overflow
6358 - update lconf's list of debug and subsystem masks
6359 - fix lfs find --obd (2510)
6360 - /proc tunable for disabling filter read caching (2591)
6361 - stop rpm packages from altering slapd.conf (2301)
6362 - disable nagle in the socknal under 0conf (2578)
6363 - choose mds inode size based on stripe count (2572)
6364 - fix kernel-source rpm problems (2516)
6366 - add --disable-doc to avoid pdf generation (2421)
6367 - update documentation, tests, type-os, comments
6368 - avoid format warnings on ia64
6369 - remove the TOE NAL
6370 - tiny code cleanups by removing unused fields
6372 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6375 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6376 - remove the most blatant lies from BUILDING file (2371)
6377 - change default debug level to reasonable production setting
6378 - reduce client side cache size to reduce cache flush time
6379 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6380 - make TCP zerocopy and pinger support enabled by default (2476)
6381 - sync writes completed after process exits caused crashes (2319)
6382 - maintain correct mount count on the MDS (2356)
6383 - backout 1557, because 2316 wasn't really fixed
6384 - better file I/O statistics gathering in /proc
6385 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6386 - another kernel patch to fix zero-copy TCP function export
6387 - don't take duplicate lock when processing re-sent getattr (2420)
6388 - lctl uses obd_self_export instead of creating new conn (2353)
6389 - MDS/OST recovery case which requires object creation asserted (2425)
6390 - move lfs from /usr/sbin to /usr/bin in packages
6391 - fix race between mds_client_add and mds_client_free (2417)
6392 - use kmalloc instead of slabs in portals (2430)
6393 - don't create duplicate records when a failover MDS is present (2442)
6394 - remove unnecessary mount age check (2332)
6395 - don't remove directory inodes from locks prematurely (2451)
6396 - don't break if MDS service name is the same as hostname (2103)
6397 - fix races in client write RPC generation when cache full (2482)
6399 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6402 - remove now-unused request->rq_obd (278)
6403 - if an allocation fails, print out how much memory we've used (1933)
6404 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6405 - add the "configurable stack size" patch to most series files (1256)
6406 - ability to write large log records, for 100+ OST configs (2306)
6407 - fix NULL deref when filter_prep fails (2314)
6408 - fix operator precedence error in filter_sync
6409 - dynamic allocation of socknal TX descriptors (2315)
6410 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6411 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6412 - ensure that CWARN messages in Portals always get to the syslog
6413 - __init/__exit are not for prototype decls (ldlm_init/exit)
6414 - x86-64 compile warning fixes
6415 - fix gateway LMC keyword conflict (2318)
6416 - fix MDS lock inversions in getattr/reint paths (1844)
6417 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6418 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6419 - DLM race condition prevented some lock evictions (2328)
6420 - ENOMEM detection and retry on socknal sends (2230)
6421 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6422 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6423 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6424 - better error messages when a client is rejected during recovery (1505)
6425 - avoid cancelling locks which were never granted, after failure (2330)
6426 - fix i_sem/journal inversion in mds_client_add (2333)
6427 - fix truncate/getattr lock cycle deadlock (2334)
6428 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6429 - fix two rare exit paths which could leak an l_lock() ref (2321)
6430 - fencepost error in MDS/OST orphan recovery (2226)
6431 - make log record alignment 8 bytes (1988)
6432 - lstripe now fails when requested offset > ost_count (2237)
6433 - ensure that all kernel series have a complete list.h (1607)
6434 - fix crashes in special-file operations (2316)
6435 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6437 - allow configurable automake binary, for testing new versions
6438 - small update to the lfs documentation
6440 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6442 * fix negative export reference count in fsfilt_sync (2312)
6444 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6445 * release candidate 0.9.1
6447 - orphans are moved into the PENDING directory for possible recovery
6448 - replayed opens now open by fid for orphan/rename safety (1042)
6449 - last close of an orphan inode generates a transno (683)
6450 - chdir() and mount() now pin the directory entry (1020)
6451 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6452 - discard very old requests without processing them (1502)
6453 - remove some common, well-understood CERRORs (1505)
6454 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6455 - clear "grant" flags in OST replies until OST grant code lands (1644)
6456 - fix read performance by not clobbering i_blksize on client (1598)
6457 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6458 - make LRU size a /proc tunable, clears locks when reduced (707)
6459 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6460 - server locks take references on exports now (1558)
6461 - build fixes for 2.4.20-rh trees (1663)
6462 - return an error from lov_create if all OSCs are inactive (1751)
6463 - fix import levels when a reconnect happens without a timeout (1597)
6464 - exit early from mds_open if we get a lookup error (1749)
6465 - partial page read at EOF wouldn't wait for disk before sending (1642)
6466 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6467 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6468 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6469 - write/truncate lock inversion (1639)
6470 - Don't auto-load obdclass, portals modules during cleanup (1495)
6471 - fix timestamps from jumping to "now" (1763)
6472 - extra journal assertions (1648)
6473 - add an extra multiunlink test (1771)
6474 - fix read_record/write_record API (1776)
6475 - fix leak of offset_extent, possible incorrect i_size later (1772)
6476 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6477 - Debugging for the kqswnal_get_idle_tx problems (1820)
6478 - Allow recovery to be attempted multiple times (1536)
6479 - Write out MDS last_rcvd file after it is first created (1600)
6480 - Fix tx_descriptor leak in failed transmit situations (1827)
6481 - ext3 journaling fixes for assertion failure after IO error (1871)
6482 - class_export_put() on freed export after completion AST error (1896)
6483 - Fix revalidate looping in VFS (1322)
6484 - Don't access a freed export during MDS_REINT timeout (1521)
6485 - Add open-unlink recovery support on the MDS (1673,1764)
6486 - Return an error if no MDS data was read from last_rcvd (1946)
6487 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6488 - Don't setup a disk device that doesn't match exported UUID (317)
6489 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6490 - avoid committing NULL handle in force close
6491 - local.sh is now a one-stripe LOV configuration
6492 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6493 - fix ext3 htree duplicate directory entry corruption (1516)
6494 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6495 - update to new LOV EA format (2097)
6496 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6498 2003-06-15 Phil Schwan <phil@clusterfs.com>
6501 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6502 - per-import/export recovery handling (958, 931, 959)
6503 - multiple last-rcvd slots, for serving multiple FSes (949)
6504 - connections are again shared between multiple imp/exports (963, 964)
6505 - "umount -f" would hang if any requests needed to be sent (393, 978)
6506 - avoid pinning large req buffer by copying for queued messages (989)
6507 - add "uuid" to "lctl device" command to help upcalls (991)
6508 - "open" RPCs with transnos would confuse recovery counters (1037)
6509 - do proper endian conversion of all wire messages (288, 340, 891)
6510 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6511 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6512 - fix ptlrpc_connection leak in target_handle_connect (1174)
6513 - fix import refcounting bug in OST and MDS cleanup (1134)
6514 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6515 - fix very unlikely obd_types race condition (501)
6516 - remove osc_open hack for echo_client (1187)
6517 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6518 - a failure in read_inode2 leads to deadlock (1139)
6519 - cancel ack-locks as soon as transaction is committed (1072)
6520 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6521 - make sure to commitrw after any preprw to avoid deadlock (1162)
6522 - failing to execute a file in a lustre FS would lock inode (1203)
6523 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6524 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6525 - obd_uuid2tgt would walk off the end of the list (1255)
6526 - on IA64 the osc would give portals incorrect bulk size (1258)
6527 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6528 - fix lock inversion caused by new llite matching code (1282)
6529 - limit the number of dirty pages on a client to 10MB (1286)
6530 - timed out locks were not being corrected cancelled (1289)
6531 - fix O_DIRECT above 4GB on IA-32 (1292)
6532 * major user-visible changes
6533 - fail out/fail over policy now controlled by the upcall (993)
6535 - add OBD_PING to check server availability and failure (954)
6536 - lustre messages are now sent in sending host order (288, 340, 891)
6537 - add eadatalen to MDS getattr reply (340)
6538 - OST read replies may contain second buffer, with per-page status (593)
6540 2003-03-11 Phil Schwan <phil@clusterfs.com>
6543 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6544 - failed open()s could cause deadlock; fixed (867, 869)
6545 - stop cancelling OST locks when files are closed (481)
6546 - overlapping XID spaces caused network corruption (851, 853)
6547 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6548 - setattr_raw added, to do single-RPC, server-side setattrs
6549 - lmc/lconf syntax change for OST UUIDs
6550 - fix crashy race condition between ptlrpc_free_req and osc_close
6551 - don't use request in mdc_enqueue if we hit a timeout (889)
6552 - don't set the inode i_size for regular files from the MDS (896)
6553 - handle out of order completion AST (842)
6554 - don't LBUG if a lock request times out after receiving AST (913)
6555 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6556 - if a bad lock AST arrives, send an error instead of dropping entirely
6557 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6558 - fix leak in bulk IO when only partially completed (899, 900, 926)
6559 - fix O_DIRECT for ia64 (55)
6560 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6561 - C-z after timeout could hang a process forever; fixed (977)
6563 - client-side I/O cache (678, 924, 929, 941, 970)
6565 - READPAGE and SETATTRs which don't take server-side locks get
6568 2003-02-11 Phil Schwan <phil@clusterfs.com>
6571 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6572 - Fix fsfilt_extN_readpage() to read a full page of directory
6573 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6574 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6575 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6576 - Most intent operations were removed, in favour of a new RPC mode
6577 that does a single RPC to the server and bypasses most of the VFS
6578 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6579 - Aggressively cancel local locks on DLM servers
6580 - mds_reint_unlink sends EA to the client if it's the last nlink.
6581 client uses that EA to unlink OST objects.
6582 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6583 - recursive symlinks were fixed (439)
6584 - fixed NULL deref in DEBUG_REQ
6585 - filter_update_lastobjid no longer calls sync, which annoyed extN
6586 - fixed multi-client small-writes to a single file problem (445)
6587 - fixed mtime updates during file writes (607)
6588 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6589 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6590 - fixed Linux OST size reporting problem (444, 656)
6591 - OST now updates object mtime with writes or setattr (607, 619)
6592 - client verifies file size before zeroing page past EOF (445)
6593 - OST now writes last allocated objid to disk with allocation (108)
6594 - LOV on echo now works (409)
6596 - mds_reint_unlink sends a new buffer, with the EA included. this
6597 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6598 set if a regular file was being unlinked, and it was the last link
6599 - use PtlGet from the target for bulk writes (315)
6600 - OST now updates object mtime with writes or setattr (607, 619)
6601 - LDLM now has a grant-time callback to revalidate locked items, if
6603 - Many MDS operations were reorganized to combat race conditions
6605 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6606 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6608 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6611 - Fully reactivate OST imports after reconnection (512, others)
6612 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6613 - More graceful error handling for truncating on dead OST (515)
6614 - Don't error out unless we're actually accessing dead stripes (474)
6615 - Fix garbage sizes when stripes are missing (410)
6616 - LRU counters were broken, causing constant lock purge (433, 432)
6617 - garbage on read from stripes with failed OSTs (441)
6618 - mark OSCs as active before reconnecting during recovery (438)
6619 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6620 - lfind did not preserve OST order in output (443)
6621 - symlinks cause hung clients, incorrect data (439)
6622 - stop dereferencing request after dropping refcount (457)
6623 - don't LASSERT(spin_is_locked) on non-SMP (455)
6624 - fixes for many rename() bugs
6625 - fstat didn't correctly synchronize attributes (399)
6626 - server must handle lock cancellation during blocking AST prep (487)
6627 - bulk descriptors were free()d too soon (511)
6628 - fix paths in lconf, which would load incorrect modules (451, 507)
6629 - fix confusing lconf 'host not found' error message (386)
6630 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6631 - fix race condition in mdc_blocking_ast() for inode access (526)
6632 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6633 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6634 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6635 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6636 - fix lstripe in several ways (kernel side) (527)
6637 - fix request leak in ldlm_cli_enqueue (262)
6638 - incorrect OSC was marked inactive after OST failure
6639 - call mds_fs_cleanup before unmounting filesystem (524)
6640 - fix races between taking ns_lock and ldlm_lock_change_resource
6641 - fix races updating LOV export open file list
6642 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6643 - fix recovery NULL deref in ldlm_cli_cancel_unused
6644 - fix some DLM races by using new hash table for lock handles (419)
6645 - permit the client to specify desired inodes, at replay
6646 - duplicate requests when we queue them for replay reintegration
6647 - fix last_rcvd offset calculation
6648 - sync after each recovered transaction, so we always make progress
6649 - never, not always, ERESTART requests without transnos
6650 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6652 - skip replay if the MDS says that the client is already connected
6653 - don't check for a recovery-enabled export to match lctl's UUID
6654 - don't INC_USE_COUNT for phantom exports
6655 - don't crash when cleaning up phantom exports (567)
6656 - don't double-finish or set replay data for errored mdc_open requests
6657 - abort requests when they time out, so we don't get old replies
6658 - send/receive replies for AST messages again
6659 - if the client says that it doesn't have the lock, cancel it on the
6661 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6662 mark it as destroyed, it will all work out when decref is called
6663 - fix module use counts (22, 581)
6665 - ASTs now expect a reply (server cancels lock on error reply)
6667 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6670 - fix many simultaneous client startup (392)
6671 - fix dentry->d_it clobbering
6672 - credentials weren't being shipped for readdir/getattr operations
6673 - remove invalid assertions triggered during some concurrent MD
6675 - proper Lustre versions added (336, 389)
6676 - fix memory leak for create error case (398)
6677 - fix LOV locking bug that would get cli/srv out of sync
6678 - fix echo client over LOV (409)
6679 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6680 - fix double-O_EXCL intent crash (424)
6681 - avoid sending multiple lock CANCELs (352)
6683 - MDS can do multi-client recovery (modulo bugs in new code)
6685 - many updates, edits, cleanups
6687 2002-11-18 Phil Schwan <phil@clusterfs.com>
6690 - fix null d_it dereference (346)
6691 - fix full OST/dbench hang (333)
6692 - fix permission problem with file removal (286)
6693 - fix removal of OSCs from LOV when they fail
6694 - fix NULL deref during bulk timeout (214)
6695 - fix problems related to multiple filesystems on one MDS (241)
6696 - fixed serious subtle metadata locking bugs
6697 - free locks on clients when inodes are removed due to memory
6699 - fix inode pointer in lock data (285)
6700 - partial support for multiple MDS on a single host (241)
6701 - data locks weren't cancelled at clear_inode time (290, 311)
6702 - intent locks could lead to unbounded lock growth (205)
6703 - added a maximum lock count, an LRU list, and a flusher
6704 - fix multiple rename (365)
6705 - properly abstracted the echo client
6706 - OSC locked 1 byte too many; fixed
6707 - rewrote brw callback code:
6708 - fixed recovery bugs related to LOVs (306)
6709 - fixed too-many-pages-in-one-write crash (191)
6710 - fixed (again) crash in sync_io_timeout (214)
6711 - probably fixed callback-related race (385)
6713 - Add capability to MDS protocol
6714 - LDLM cancellations and callbacks on different portals
6716 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6719 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6720 - defer open object destruction to close time (601981, 138)
6721 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6722 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6723 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6724 - don't call obd_setattr in ll_file_release for destroyed objects
6725 * protocol change to lustre_msg: move |version| and add |flags|
6726 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6727 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6728 * added replay of create, unlink, link and rename operations during
6729 MDS failover; recovery should be much more robust now
6730 * remove failed OSCs from LOVs (only lov_create uses this so far)
6731 * the lustre-HOWTO was brought (more) up to date (582544)
6733 2002-10-23 Phil Schwan <phil@clusterfs.com>
6736 - in-use dentries weren't being reused properly (617851)
6737 - prevent multiple LDLM setup (599178)
6738 - fix LOV size calculations for truncate (617853)
6739 - fix client handling of MDS intent errors (POSIX)
6740 - fix permission bug in lovstripe.c test (624321)
6741 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6742 - truncate past end of file could corrupt data
6743 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6744 - a race in recovery could return ETIMEDOUT to apps (623947)
6745 - building outside the source directory was fixed
6746 * the lustre-HOWTO was brought (more) up to date (582544)
6747 * major progress was made on recovery functionality
6749 2002-10-10 Phil Schwan <phil@clusterfs.com>
6752 - recovery deadlock fix
6753 - rm -rf causes LBUG fix (617817)
6754 - file open by multiple tasks fix (618962)
6755 - directory permissions bugs (602707 and 620007)
6756 - journal_stop fixed with locking (611313)
6757 - O_APPEND failures resolved (618273, perhaps 614459)
6758 - lconf PATH fix (619770)
6759 - IA64 build fix (621450)
6760 - RPC buffer sizes scale with amount of memory
6762 2002-10-01 Phil Schwan <phil@clusterfs.com>
6765 - locks would be cancelled without throwing away data pages,
6766 resulting in inconsistent data (605627)
6767 - inode attributes were not always being refreshed (605627, 612449)
6768 - lconf now continues to cleanup after lctl reports an error
6769 - MDS now enforces user permissions (602707)
6770 - lprocfs cleanup fixed, but not yet enabled (614157)
6771 - fixed infinite server hang, should a client not respond to an AST
6772 - avoid going into recovery if user calls readlink() with a buffer
6773 that's too small (613941)
6774 - AST RPCs no longer require replies (614867) -- this may be changed
6775 - don't crash server if client sends an IOV that's too big (611336)
6776 - fixed lock conversion deadlock (611892)
6777 - fixed the following of symlinks (614622)
6778 * recovery: the server can remove locks from a client that dies, other
6779 clients can make progress
6780 * more extN patch fixes
6781 * compile-time configurable ptlrpc buffer allocations
6783 - collaborative read cache document
6784 - Lustre Lite Performance CDR document-in-progress
6786 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6789 - fix typo in patch-2.4.18
6791 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6794 - clear ptlrpc request each time in handle_incoming_request()
6795 - unlink of files now destroys the object on the OST
6797 2002-09-19 Peter Braam <braam@clusterfs.com>
6799 * add hard link support
6800 * change obdfile creation method
6801 * kernel patch changed
6803 2002-09-19 Peter Braam <braam@clusterfs.com>
6806 - stack overflow bug in extN fixed
6808 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6810 * documentation updates
6811 - add man pages for config tools
6812 - update tests/README to describe testing with new config tools
6813 - finish metadata API descriptions
6814 * bug fixes and cleanups
6815 - statfs workaround for 16TB limit
6816 - LOV stripe allocation improved, can stripe on subset of OSTs
6817 - LOV file size/IO offset was wrong for files > 4GB in size
6818 - object EA data was being dropped, caused files to be unreadable
6819 - memory overflow with non-LOV OST caused memory corruption
6820 - fixed regression tests to work with new config tools, obdfilter
6821 - fixed bug when directory size became larger than 1 block
6822 - fixed bug (for single client case) when PWD was deleted
6823 - invalidate local directory pages when doing intent-based ops
6824 - avoid LDLM oops when lock callback contained bad data
6826 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6828 * documentation updates
6829 * bug fixes and cleanups
6830 - configuration tools
6833 - 64-bit compile warnings
6834 - 64-bit internal statfs data
6836 * test_brw on persistent OST devices
6838 * lprocfs (disabled)
6840 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6842 * documentation updates
6843 * bug fixes and cleanups
6844 * configuration tools
6846 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6849 * many small fixes to 0_5_4
6850 * io/network handling
6851 * thinkos in MDS operations
6853 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6856 * crucial basic fixes to 0.5.3
6857 * IOR, Iozone work over Elan
6860 2002-08-07 Phil Schwan <phil@clusterfs.com>
6861 * version 0_5_3, our first alpha
6862 * we use the new Portals iovs
6863 * documentation updates
6864 * bug fixes and cleanups
6865 * small changes in the DLM wire protocol
6867 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6868 * version 0_5_1 with some initial stability,
6869 * locking on MD and file I/O.
6870 * documentation updates
6871 * several bug fixes since 0.5.0
6872 * small changes in wire protocol
6874 2002-07-18 Phil Schwan <phil@clusterfs.com>
6876 * delivered as Lustre Light Alpha
6877 * fixed a crash after handling invalid MDS requests
6878 * fixed directory pages for architectures with non-4k pages sizes
6880 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6881 * release version v0_4_4
6882 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6884 * Fixes a number of interruption problems with OST operations.
6885 * Update documentation for portals header changes
6886 * Move all wire protocol structs/defines to lustre_idl.h
6887 * Fixes symlink length bug.
6888 * Add tcpdump to repository.
6890 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6891 * release version v0_4_3
6892 * Fixes statfs for inodes on extN.
6893 * Fixes bug in runtests which would delete /etc/hosts.
6894 * Use 64-bit object IDs wherever possible (not into VFS though)
6895 Remove ost_get_info, which is unused by lustre, and out of date.
6897 2002-07-03 Peter Braam <braam@clusterfs.com>
6898 * release version v0_4_2 Fixes a lookup error (type not passed)
6899 * move forward to head of Portals
6900 * move forward to latest Lustre kernel
6902 2002-06-25 Peter Braam <braam@clusterfs.com>
6903 * release version v0_4_1. Hopefully stable on single node use.