3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 4.18.0-372.26.1.el8 (RHEL8.6)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 3.10.0-1062.18.1.el7 (RHEL7.7)
11 3.10.0-1127.19.1.el7 (RHEL7.8)
12 3.10.0-1160.76.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.18.0-348.23.1.el8 (RHEL8.5)
19 4.4.120-92.70 (SLES12 SP2)
20 4.4.180-94.100 (SLES12 SP3)
21 4.4.0-131 (Ubuntu 16.04)
22 4.15.0-32 (Ubuntu 18.04)
23 5.4.0-48 (Ubuntu 20.04)
24 vanilla linux 5.4.0 (ZFS + ldiskfs)
25 vanilla linux 5.4.21 (ZFS + ldiskfs)
26 vanilla linux 5.4.136 (ZFS + ldiskfs)
27 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
28 * Client primary kernels built and tested during release cycle:
29 5.14.0-70.22.1.el9 (RHEL9.0)
30 4.18.0-372.26.1.el8 (RHEL8.6)
31 5.4.0-37 (Ubuntu 20.04)
32 5.14.21-150400.24.21 (SLES15 SP4)
33 * Other clients known to build on these kernels at some point (others may also work):
34 3.10.0-862.14.4.el7 (RHEL7.5)
35 3.10.0-957.27.2.el7 (RHEL7.6)
36 4.14.0-49.13.1.el7a (RHEL7.5)
37 4.14.0-115.2.2.el7a (RHEL7.6)
38 3.10.0-1062.18.1.el7 (RHEL7.7)
39 3.10.0-1127.19.1.el7 (RHEL7.8)
40 3.10.0-1160.76.1.el7 (RHEL7.9)
41 4.18.0-80.11.2.el8 (RHEL8.0)
42 4.18.0-147.8.1.el8 (RHEL8.1)
43 4.18.0-193.28.1.el8 (RHEL8.2)
44 4.18.0-240.22.1.el8 (RHEL8.3)
45 4.18.0-305.25.1.el8 (RHEL8.4)
46 4.18.0-348.23.1.el8 (RHEL8.5)
47 4.4.120-92.70 (SLES12 SP2)
48 4.4.180-94.100 (SLES12 SP3)
49 4.12.14-95.48 (SLES12 SP4)
50 4.12.14-122.130 (SLES12 SP5)
51 4.12.14-197.75 (SLES15 SP1)
52 5.3.18-24.96 (SLES15 SP2)
53 5.3.18-150300.59.93 (SLES15 SP3)
54 4.4.0-131 (Ubuntu 16.04)
55 4.15.0-48 (Ubuntu 18.04)
56 5.8.0-53 (Ubuntu 20.04.2 HWE)
57 5.11.0-31 (Ubuntu 20.04.3 HWE)
58 5.11.0 (vanilla kernel.org)
59 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
60 * Recommended ZFS version: 2.1.5
61 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
62 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
63 stack overflow. For more information, please refer to bugzilla 17630.
64 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
65 the re-exporting nodes to be mounted with "32bitapi" mount option
67 --------------------------------------------------------------------------------
72 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
73 for currently supported client and server kernel versions.
74 * Server primary kernels built and tested during release cycle:
75 4.18.0-348.2.1.el8 (RHEL8.5)
76 * Other server kernels known to build and work at some point (others may also work):
77 3.10.0-862.14.4.el7 (RHEL7.5)
78 3.10.0-957.27.2.el7 (RHEL7.6)
79 3.10.0-1062.18.1.el7 (RHEL7.7)
80 3.10.0-1127.19.1.el7 (RHEL7.8)
81 3.10.0-1160.49.1.el7 (RHEL7.9)
82 4.18.0-80.11.2.el8 (RHEL8.0)
83 4.18.0-147.8.1.el8 (RHEL8.1)
84 4.18.0-193.28.1.el8 (RHEL8.2)
85 4.18.0-240.22.1.el8 (RHEL8.3)
86 4.18.0-305.25.1.el8 (RHEL8.4)
87 4.4.120-92.70 (SLES12 SP2)
88 4.4.180-94.100 (SLES12 SP3)
89 4.4.0-131 (Ubuntu 16.04)
90 4.15.0-32 (Ubuntu 18.04)
91 5.4.0-48 (Ubuntu 20.04)
92 vanilla linux 5.4.0 (ZFS + ldiskfs)
93 vanilla linux 5.4.21 (ZFS + ldiskfs)
94 vanilla linux 5.4.136 (ZFS + ldiskfs)
95 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
96 * Client primary kernels built and tested during release cycle:
97 4.18.0-348.2.1.el8 (RHEL8.5)
98 5.4.0-37 (Ubuntu 20.04)
99 5.3.18-59.27 (SLES15 SP3)
100 * Other clients known to build on these kernels at some point (others may also work):
101 3.10.0-862.14.4.el7 (RHEL7.5)
102 3.10.0-957.27.2.el7 (RHEL7.6)
103 4.14.0-49.13.1.el7a (RHEL7.5)
104 4.14.0-115.2.2.el7a (RHEL7.6)
105 3.10.0-1062.18.1.el7 (RHEL7.7)
106 3.10.0-1127.19.1.el7 (RHEL7.8)
107 3.10.0-1160.49.1.el7 (RHEL7.9)
108 4.18.0-80.11.2.el8 (RHEL8.0)
109 4.18.0-147.8.1.el8 (RHEL8.1)
110 4.18.0-193.28.1.el8 (RHEL8.2)
111 4.18.0-240.22.1.el8 (RHEL8.3)
112 4.18.0-305.25.1.el8 (RHEL8.4)
113 4.4.120-92.70 (SLES12 SP2)
114 4.4.180-94.100 (SLES12 SP3)
115 4.12.14-95.48 (SLES12 SP4)
116 4.12.14-122.91 (SLES12 SP5)
117 4.12.14-197.75 (SLES15 SP1)
118 5.3.18-24.96 (SLES15 SP2)
119 4.4.0-131 (Ubuntu 16.04)
120 4.15.0-48 (Ubuntu 18.04)
121 5.8.0-53 (Ubuntu 20.04.2 HWE)
122 5.11.0-31 (Ubuntu 20.04.3 HWE)
123 5.11.0 (vanilla kernel.org)
124 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
125 * Recommended ZFS version: 2.0.7
126 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
127 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
128 stack overflow. For more information, please refer to bugzilla 17630.
129 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
130 the re-exporting nodes to be mounted with "32bitapi" mount option
132 --------------------------------------------------------------------------------
136 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
137 for currently supported client and server kernel versions.
138 * Server primary kernels built and tested during release cycle:
139 3.10.0-1062.18.1.el7 (RHEL7.7)
140 3.10.0-1127.19.1.el7 (RHEL7.8)
141 3.10.0-1160.6.1.el7 (RHEL7.9)
142 * Other server kernels known to build and work at some point (others may also work):
143 3.10.0-862.14.4.el7 (RHEL7.5)
144 3.10.0-957.27.2.el7 (RHEL7.6)
145 4.18.0-80.11.2.el8 (RHEL8.0)
146 4.18.0-147.8.1.el8 (RHEL8.1)
147 4.18.0-193.28.1.el8 (RHEL8.2)
148 4.18.0-240.1.1.el8 (RHEL8.3)
149 4.4.120-92.70 (SLES12 SP2)
150 4.4.180-94.100 (SLES12 SP3)
151 4.4.0-131 (Ubuntu 16.04)
152 4.15.0-32 (Ubuntu 18.04)
153 5.4.0-48 (Ubuntu 20.04)
154 vanilla linux 5.4.0 (ZFS + ldiskfs)
155 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
156 * Client primary kernels built and tested during release cycle:
157 3.10.0-1062.18.1.el7 (RHEL7.7)
158 3.10.0-1127.19.1.el7 (RHEL7.8)
159 3.10.0-1160.6.1.el7 (RHEL7.9)
160 4.12.14-95.48 (SLES12 SP4)
161 4.12.14-122.20 (SLES12 SP5)
162 4.12.14-197.75 (SLES15 SP1)
163 5.3.18-24.24 (SLES15 SP2)
164 4.15.0-48 (Ubuntu 18.04)
165 * Other clients known to build on these kernels at some point (others may also work):
166 3.10.0-862.14.4.el7 (RHEL7.5)
167 3.10.0-957.27.2.el7 (RHEL7.6)
168 4.14.0-49.13.1.el7a (RHEL7.5)
169 4.14.0-115.2.2.el7a (RHEL7.6)
170 4.18.0-80.11.2.el8 (RHEL8.0)
171 4.18.0-147.8.1.el8 (RHEL8.1)
172 4.18.0-193.28.1.el8 (RHEL8.2)
173 4.18.0-240.1.1.el8 (RHEL8.3)
174 4.4.120-92.70 (SLES12 SP2)
175 4.4.180-94.100 (SLES12 SP3)
176 4.4.0-131 (Ubuntu 16.04)
177 5.4.0-37 (Ubuntu 20.04)
178 5.4.0 (vanilla kernel.org)
179 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
180 * Recommended ZFS version: 2.0.0
181 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
182 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
183 stack overflow. For more information, please refer to bugzilla 17630.
184 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
185 the re-exporting nodes to be mounted with "32bitapi" mount option
187 --------------------------------------------------------------------------------
191 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
192 for currently supported client and server kernel versions.
193 * Server primary kernels built and tested during release cycle:
194 3.10.0-1062.1.1.el7 (RHEL7.7)
195 * Other server kernels known to build and work at some point (others may also work):
196 3.10.0-862.14.4.el7 (RHEL7.5)
197 3.10.0-957.27.2.el7 (RHEL7.6)
198 4.18.0-80.7.1.el8 (RHEL8.0)
199 4.4.120-92.70 (SLES12 SP2)
200 4.4.180-94.100 (SLES12 SP3)
201 4.4.0-131 (Ubuntu 16.04)
202 4.15.0-32 (Ubuntu 18.04)
203 vanilla linux 4.6.7 (ZFS only)
204 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
205 * Client primary unpatched kernels built and tested during release cycle:
206 3.10.0-1062.1.1.el7 (RHEL7.7)
207 4.12.14-95.29 (SLES12 SP4)
208 4.15.0-48 (Ubuntu 18.04)
209 * Other clients known to build on unpatched kernels at some point (others may also work):
210 3.10.0-862.14.4.el7 (RHEL7.5)
211 3.10.0-957.27.2.el7 (RHEL7.6)
212 4.14.0-49.13.1.el7a (RHEL7.5)
213 4.14.0-115.2.2.el7a (RHEL7.6)
214 4.18.0-80.7.1.el8 (RHEL8.0)
215 4.4.120-92.70 (SLES12 SP2)
216 4.4.180-94.100 (SLES12 SP3)
217 4.4.0-131 (Ubuntu 16.04)
218 vanilla linux 4.15.0 (ZFS only)
219 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
220 * Recommended ZFS / SPL version: 0.7.13
221 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
222 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
223 stack overflow. For more information, please refer to bugzilla 17630.
224 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
225 the re-exporting nodes to be mounted with "32bitapi" mount option
226 * Note for Ubuntu 14.04.5:
227 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
228 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
230 --------------------------------------------------------------------------------
234 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
235 for currently supported client and server kernel versions.
236 * Server known to build on the following kernels (others may also work):
237 2.6.32-696.18.7.el6 (RHEL6.9)
238 3.10.0-862.14.4.el7 (RHEL7.5)
239 3.10.0-957.el7 (RHEL7.6)
240 4.4.120-92.70 (SLES12 SP2)
241 4.4.162-94.69 (SLES12 SP3)
242 4.4.0-131 (Ubuntu 16.04)
243 4.15.0-32 (Ubuntu 18.04)
244 vanilla linux 4.6.7 (ZFS only)
245 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
246 * Client known to build on unpatched kernels (others may also work):
247 2.6.32-696.18.7.el6 (RHEL6.9)
248 2.6.32-754.6.3.el6 (RHEL6.10)
249 3.10.0-862.14.4.el7 (RHEL7.5)
250 3.10.0-957.el7 (RHEL7.6)
251 4.14.0-49.13.1.el7a (RHEL7.5)
252 4.14.0-115.2.2.el7a (RHEL7.6)
253 4.4.120-92.70 (SLES12 SP2)
254 4.4.162-94.69 (SLES12 SP3)
255 4.4.0-131 (Ubuntu 16.04)
256 4.15.0-32 (Ubuntu 18.04)
257 vanilla linux 4.15.0 (ZFS only)
258 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
259 * Recommended ZFS / SPL version: 0.7.9
260 * Tested with ZFS / SPL version: 0.7.9
261 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
262 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
263 stack overflow. For more information, please refer to bugzilla 17630.
264 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
265 the re-exporting nodes to be mounted with "32bitapi" mount option
266 * Note for Ubuntu 14.04.5:
267 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
268 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
270 --------------------------------------------------------------------------------
272 2018-04-03 Intel Corporation
274 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
275 for currently supported client and server kernel versions.
276 * Server known to build on patched kernels:
277 2.6.32-431.29.2.el6 (RHEL6.5)
278 2.6.32-504.30.3.el6 (RHEL6.6)
279 2.6.32-573.26.1.el6 (RHEL6.7)
280 2.6.32-642.15.1.el6 (RHEL6.8)
281 2.6.32-696.18.7.el6 (RHEL6.9)
282 3.10.0-693.21.1.el7 (RHEL7.4)
283 3.0.101-0.47.71 (SLES11 SP3)
284 3.0.101-107 (SLES11 SP4)
285 3.12.74-60.64.40 (SLES12 SP1)
286 4.4.114-92.64 (SLES12 SP2)
287 4.4.114-94.11 (SLES12 SP3)
288 3.13.0-101 (Ubuntu 14.04, ZFS only)
289 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
290 4.4.0-116 (Ubuntu 16.04)
291 vanilla linux 4.6.7 (ZFS only)
292 * Client known to build on unpatched kernels:
293 2.6.32-431.29.2.el6 (RHEL6.5)
294 2.6.32-504.30.3.el6 (RHEL6.6)
295 2.6.32-573.26.1.el6 (RHEL6.7)
296 2.6.32-642.15.1.el6 (RHEL6.8)
297 2.6.32-696.18.7.el6 (RHEL6.9)
298 3.10.0-693.21.1.el7 (RHEL7.4)
299 3.0.101-0.47.71 (SLES11 SP3)
300 3.0.101-107 (SLES11 SP4)
301 3.12.74-60.64.40 (SLES12 SP1)
302 4.4.114-92.64 (SLES12 SP2)
303 4.4.114-94.11 (SLES12 SP3)
304 3.13.0-101 (Ubuntu 14.04)
305 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
306 4.4.0-116 (Ubuntu 16.04)
308 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
309 * Recommended ZFS / SPL version: 0.7.6
310 * Tested with ZFS / SPL version: 0.7.6
311 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
312 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
313 stack overflow. For more information, please refer to bugzilla 17630.
314 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
315 the re-exporting nodes to be mounted with "32bitapi" mount option
316 * Note for Ubuntu 14.04.5:
317 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
318 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
320 --------------------------------------------------------------------------------
322 2017-07-05 Intel Corporation
324 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
325 for currently supported client and server kernel versions.
326 * Server known to build on patched kernels:
327 2.6.32-431.29.2.el6 (RHEL6.5)
328 2.6.32-504.30.3.el6 (RHEL6.6)
329 2.6.32-573.26.1.el6 (RHEL6.7)
330 2.6.32-642.15.1.el6 (RHEL6.8)
331 2.6.32-696.3.1.el6 (RHEL6.9)
332 3.10.0-514.21.1.el7 (RHEL7.3)
333 3.0.101-0.47.71 (SLES11 SP3)
334 3.0.101-100 (SLES11 SP4)
335 3.12.74-60.64.40 (SLES12 SP1)
336 4.4.59-92.17 (SLES12 SP2)
337 vanilla linux 4.6.7 (ZFS only)
338 * Client known to build on unpatched kernels:
339 2.6.32-431.29.2.el6 (RHEL6.5)
340 2.6.32-504.30.3.el6 (RHEL6.6)
341 2.6.32-573.26.1.el6 (RHEL6.7)
342 2.6.32-642.15.1.el6 (RHEL6.8)
343 2.6.32-696.3.1.el6 (RHEL6.9)
344 3.10.0-514.21.1.el7 (RHEL7.3)
345 3.0.101-0.47.71 (SLES11 SP3)
346 3.0.101-100 (SLES11 SP4)
347 3.12.74-60.64.40 (SLES12 SP1)
348 4.4.59-92.17 (SLES12 SP2)
350 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
351 * Recommended ZFS / SPL version: 0.6.5.9
352 * Tested with ZFS / SPL version: 0.6.5.9
353 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
354 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
355 stack overflow. For more information, please refer to bugzilla 17630.
356 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
357 the re-exporting nodes to be mounted with "32bitapi" mount option
359 --------------------------------------------------------------------------------
361 12-07-2016 Intel Corporation
363 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
364 for currently supported client and server kernel versions.
365 * Server known to build on patched kernels:
366 2.6.32-431.29.2.el6 (RHEL6.5)
367 2.6.32-504.30.3.el6 (RHEL6.6)
368 2.6.32-573.26.1.el6 (RHEL6.7)
369 2.6.32-642.11.1.el6 (RHEL6.8)
370 3.10.0-514.el7 (RHEL7.3)
371 3.0.101-0.47.71 (SLES11 SP3)
372 3.0.101-80 (SLES11 SP4)
373 3.12.62-60.64.8 (SLES12 SP1)
374 vanilla linux 4.5.7 (ZFS only)
375 * Client known to build on unpatched kernels:
376 2.6.32-431.29.2.el6 (RHEL6.5)
377 2.6.32-504.30.3.el6 (RHEL6.6)
378 2.6.32-573.26.1.el6 (RHEL6.7)
379 2.6.32-642.11.1.el6 (RHEL6.8)
380 3.10.0-514.el7 (RHEL7.3)
381 3.0.101-0.47.71 (SLES11 SP3)
382 3.0.101-80 (SLES11 SP4)
383 3.12.62-60.64.8 (SLES12 SP1)
385 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
386 * Recommended ZFS / SPL version: 0.6.5.8
387 * Tested with ZFS / SPL version: 0.6.5.8
388 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
389 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
390 stack overflow. For more information, please refer to bugzilla 17630.
391 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
392 the re-exporting nodes to be mounted with "32bitapi" mount option
394 --------------------------------------------------------------------------------
396 02-29-2016 Intel Corporation
398 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
399 for currently supported client and server kernel versions.
400 * Server known to build on patched kernels:
401 2.6.32-431.29.2.el6 (RHEL6.5)
402 2.6.32-504.30.3.el6 (RHEL6.6)
403 2.6.32-573.12.1.el6 (RHEL6.7)
404 3.10.0-327.3.1.el7 (RHEL7.2)
405 3.0.101-0.47.71 (SLES11 SP3)
406 3.0.101-68 (SLES11 SP4)
407 vanilla linux 4.2.1 (ZFS only)
408 * Client known to build on unpatched kernels:
409 2.6.32-431.29.2.el6 (RHEL6.5)
410 2.6.32-504.30.3.el6 (RHEL6.6)
411 2.6.32-573.12.1.el6 (RHEL6.7)
412 3.10.0-327.3.1.el7 (RHEL7.2)
413 3.0.101-0.47.71 (SLES11 SP3)
414 3.0.101-68 (SLES11 SP4)
417 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
418 * Recommended ZFS / SPL version: 0.6.4.2
419 * Tested with ZFS / SPL version: 0.6.4.2
420 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
421 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
422 stack overflow. For more information, please refer to bugzilla 17630.
423 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
424 the re-exporting nodes to be mounted with "32bitapi" mount option
426 --------------------------------------------------------------------------------
428 03-10-2015 Intel Corporation
430 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
431 for currently supported client and server kernel versions.
432 * Server known to build on patched kernels:
433 2.6.32-431.29.2.el6 (RHEL6.5)
434 2.6.32-504.8.1.el6 (RHEL6.6)
435 3.0.101-0.46 (SLES11 SP3)
436 * Client known to build on unpatched kernels:
437 2.6.32-431.29.2.el6 (RHEL6.5)
438 2.6.32-504.8.1.el6 (RHEL6.6)
439 3.10.0-123.20.1.el7 (RHEL7)
440 3.0.101-0.46 (SLES11 SP3)
441 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
442 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
443 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
444 stack overflow. For more information, please refer to bugzilla 17630.
445 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
446 the re-exporting nodes to be mounted with "32bitapi" mount option
448 Severity : enhancement
450 Description: control OST-index in IDIF via ROCOMPAT flag.
451 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
452 last_rcvd file. For new formatted OST device, it will be auto set;
453 for the case of upgrading from old OST device, you can enable it
454 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
455 enabled, for new created OST-object, its IDIF-in-LMA will contain
456 the OST-index; for the existing OST-object, the OSD will convert
457 old format IDIF as new format IDIF with OST-index stored in the
458 LMA EA when accessing such OST-object or via OI scrub. Once such
459 flag is enabled, it cannot be reverted back, so the system cannot
460 be downgraded to the orignal incompatible version.
462 --------------------------------------------------------------------------------
464 07-30-2014 Intel Corporation
466 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
467 for currently supported client and server kernel versions.
468 * Server known to build on patched kernels:
469 2.6.32-431.20.3.el6 (RHEL6.5)
470 3.0.101-0.31 (SLES11 SP3)
471 * Client known to build on unpatched kernels:
472 2.6.32-431.20.3.el6 (RHEL6.5)
473 3.0.101-0.31 (SLES11 SP3)
476 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
477 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
478 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
479 stack overflow. For more information, please refer to bugzilla 17630.
480 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
481 the re-exporting nodes to be mounted with "32bitapi" mount option
483 --------------------------------------------------------------------------------
485 10-31-2013 Intel Corporation
487 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
488 * Server builds for kernels:
489 2.6.32-358.18.1.el6 (RHEL6.4)
490 * Client builds for unpatched kernels:
491 2.6.32-358.18.1.el6 (RHEL6.4)
492 3.0.92-0.8 (SLES11 SP3)
494 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
495 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
496 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
497 stack overflow. For more information, please refer to bugzilla 17630.
498 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
499 the re-exporting nodes to be mounted with "32bitapi" mount option
501 --------------------------------------------------------------------------------
503 05-30-2013 Intel Corporation
505 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
506 * Server builds for kernels:
507 2.6.32-358.6.2.el6 (RHEL6.4)
508 * Client builds for unpatched kernels:
509 2.6.32-358.6.2.el6 (RHEL6.4)
510 3.0.74-0.6.6 (SLES11 SP2)
512 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
513 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
514 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
515 stack overflow. For more information, please refer to bugzilla 17630.
516 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
517 the re-exporting nodes to be mounted with "32bitapi" mount option
519 --------------------------------------------------------------------------------
521 9-30-2012 Whamcloud, Inc.
523 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
524 * Server support for kernels:
525 2.6.32-279.5.1.el6 (RHEL6)
526 * Client support for unpatched kernels:
527 2.6.32-279.5.1.el6 (RHEL6)
529 2.6.18-238.19.1.el5 (RHEL5)
530 2.6.16.60-0.69.1 (SLES10)
531 * Recommended e2fsprogs version: 1.41.90.wc4
532 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
533 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
534 stack overflow. For more information, please refer to bugzilla 17630.
535 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
536 the re-exporting nodes to be mounted with "32bitapi" mount option
538 Severity : enhancement
540 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
542 Severity : enhancement
544 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
546 Severity : enhancement
548 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
550 Severity : enhancement
552 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
554 Severity : enhancement
556 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
558 --------------------------------------------------------------------------------
560 04-30-2012 Whamcloud, Inc.
562 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
563 * Server support for kernels:
564 2.6.32-220.4.2.el6 (RHEL6)
565 * Client support for unpatched kernels:
566 2.6.32-220.4.2.el6 (RHEL6)
568 2.6.18-238.19.1.el5 (RHEL5)
569 2.6.16.60-0.69.1 (SLES10)
570 * Recommended e2fsprogs version: 1.41.90.wc4
571 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
572 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
573 stack overflow. For more information, please refer to bugzilla 17630.
574 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
575 the re-exporting nodes to be mounted with "32bitapi" mount option
577 Severity : enhancement
579 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
581 Severity : enhancement
583 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
585 Severity : enhancement
587 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
589 Severity : enhancement
591 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
593 --------------------------------------------------------------------------------
595 09-30-2011 Whamcloud, Inc.
597 * Server support for kernels:
598 2.6.18-238.19.1.el5 (RHEL5)
599 2.6.32-131.6.1.el6 (RHEL6)
600 * Client support for unpatched kernels:
601 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
602 2.6.18-238.19.1.el5 (RHEL5)
603 2.6.32-131.6.1.el6 (RHEL6)
604 2.6.29.4-167.fc11 (FC11)
605 2.6.16.60-0.69.1 (SLES10)
607 * Recommended e2fsprogs version: 1.41.90.wc3
608 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
609 Lustre filesystem with 4K stack may cause a stack overflow. For more
610 information, please refer to bugzilla 17630.
611 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
612 mounted with "32bitapi" mount option
614 Severity : enhancement
615 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
617 Severity : enhancement
618 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
620 Severity : enhancement
621 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
624 Frequency : only with SLES10
626 Description: Use OFED "KMP" provided by Novell
627 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
628 Lustre is now built against this package. That means you need to
629 install the ofed-kmp package from Novell for the patchless client.
630 Note that the ofed-kmp that Novell ships may not exactly match the
631 kernel version but should still be compatible.
633 Severity : enhancement
635 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
636 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
638 --------------------------------------------------------------------------------
640 2010-07-15 Oracle, Inc.
642 * Server support for kernels:
643 2.6.18-164.11.1.el5 (RHEL 5)
644 2.6.18-164.11.1.0.1.el5 (OEL 5)
645 * Client support for unpatched kernels:
646 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
647 2.6.18-164.11.1.el5 (RHEL 5),
648 2.6.18-164.11.1.0.1.el5 (OEL 5)
649 2.6.16.60-0.42.8 (SLES 10),
651 2.6.29.4-167.fc11 (FC11)
652 * Recommended e2fsprogs version: 1.41.10-sun2
654 Severity : enhancement
656 Description: Distribute lfs_migrate script with lustre
657 Details : The lfs_migrate mentioned in the manual is now included in the
658 lustre tree in lustre/scripts and will distributed in lustre RPMs.
660 Severity : enhancement
662 Description: Failover nid list added to import proc file.
664 Severity : enhancement
666 Description: "som_preview" mount option required to use SOM
667 Details : The clients will need to use som_preview mount option
671 Description: Lustre does not do 1MB IOs to HW RAID
672 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
674 Severity : enhancement
676 Description: lfs quota output is not very convenient for awk/sed-parsing
677 Details : Some positions in lfs quota output table could be empty or
678 non-empty which made it hard to parse it with scripts, now
679 a dash is put instead of space where there is not supposed
684 Description: Update to OFED 1.5.1
685 Details : OFED 1.5.1 will be used in environments that don't use
686 vendor supplied OFED.
690 Description: Remove lnet. prefix from lctl params display
691 Details : Also renamed lnet.memused to lnet_memused to
696 Description: Use conf_param -d to remove permanant settings.
700 Description: Add list_param -R
701 Details : lctl list_param now supports recursive mode.
704 Bugzilla : 21829, 22312
705 Description: Fix llobdstat.
706 Details : 1. Now updating the write_bytes counter
707 2. Port script fixes from 1.8.
711 Description: decrease the usage of memory on clients.
712 Details : 1. On clients, recycle dentries and inodes unused.
713 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
718 Description: llobdstat fix and enhancement
719 Details : add a counter to set a limit to how many samples will be returned
720 fix a wildcard in the path to limit to obdfilter stats only
724 Description: lfs check only allowed for root
725 Details : Code cleanups and added a regresion test of non-root use of
730 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
731 Details : Fix hr->hr_index overload.
735 Description: Fix FIEMAP functionality.
739 Description: FC11 patchless client support added
743 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
744 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
748 Description: New acc-sm script and various test-framework enhancments.
752 Description: Previously landed patch was reverted.
756 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
760 Description: Add ldiskfs maxdirsize mount option.
764 Description: Fix mkfs.lustre support for 16TB LUNs.
768 Description: Print mmp_check_interval
769 Details : print mmp_check_interval and make it possible to abort
770 mount operation in case it takes too long.
774 Description: Add cascading_rw to lustre mpi tests.
778 Description: Store SOM data in EA.
779 Details : Also several SOM related cleanups.
783 Description: Add refcount to md_open_data
787 Description: Use refcount on lov_request_set to prevent memory corruption.
791 Description: Statahead: use parent dir's i_mutex for synchronization.
795 Description: loadgen improvements.
799 Description: Enable peer health detection in ksocklnd and ko2iblnd.
803 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
807 Description: Fix race on obd_devcie:md_stats
811 Description: Fix SLES 11 client builds.
812 Details : Resolved uninitalized pointer gcc warnings.
816 Description: Allow non-root access for "lfs check".
817 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
821 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
823 Severity : enhancement
825 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
826 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
830 Description: Sanity test_27p fails.
831 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
832 Sanity tests 27 were cleaned up and improved.
836 Description: Unify req format on client/servers
837 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
838 buffers and swabbers approach..
843 Description: drop join file stuff.
845 Severity : enhancement
847 Description: Add 2.6.27 and SLES11 (patchless client) support
851 Description: Fix in ptlrpc_expire_one_request() to print the signed time
854 Severity : enhancement
856 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
859 Severity : enhancement
861 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
863 Severity : enhancement
865 Description: correctly handle big reply message.
866 Details : send LNet event if reply is bigger than buffer and adjust this buffer
871 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
873 Severity : enhancement
875 Description: Add support for OEL5.
879 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
881 Severity : enhancement
883 Description: "weak-modules" support
884 Details : Implement "weak-modules" support which enables kernel modules
885 to be used with any kernel that implements the same kABI. In
886 order to achieve this modules are now installed in
887 /lib/modules/$(uname -r)/updates/kernel on all distributions.
891 Description: client could not reconnect to OST because of an active request.
892 Details : The client now retries to reconnect to the same server, if a
893 connect request failed with EBUSY or -EAGAIN. Test case 35b
894 conf-sanity.sh has been created.
898 Description: Don't allow make backward step on assignin osc next id.
899 Details : race between allocation next id and ll_sync thread can be cause
900 of set wrong osc next id and can be kill valid ost objects.
904 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
908 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
912 Description: Error handling in osc_statfs_interpret() has been improved.
913 Details : Check in osc_statfs_interpret() for EBADR.
917 Description: Do not send statfs() requests to OSTs disabled by administrator.
918 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
920 Severity : enhancement
922 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
924 Severity : enhancement
926 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
928 Severity : enhancement
930 Description: Enhancement for lfs(1) command to use numeric uid/gid.
934 Description: Improved cleanup procedure for conf-sanity test_46a.
938 Description: Do not start bulk transfer, if client tried to reconnect.
939 Details : Do not start bulk transfer, if client tried to reconnect,
940 but failed due to active rpcs. Check for exp_abort_active_req
941 has been implemented.
945 Description: Prevent inconsistences between linux and lustre mount structures.
946 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
947 Make the sleep interruptible.
949 Severity : enhancement
951 Description: provide server to client comms path
952 Details : server to client communications path via new LDLM_SET_INFO rpc
956 Description: truncate starts GFP_FS allocation under transaction causing deadlock
957 Details : ldiskfs_truncate calls grab_cache_page which may start page
958 allocation under an open transaction. This may lead to
959 calling prune_icache with consequent lustre reentrance.
963 Description: send correctly lsm on open replay
964 Details : MDS is trust to LSM size on replay open, but client can set wrong size
969 Description: enable client interface failover
970 Details : When a child reconnects from another NID, properly update export
971 nid hash position and ldlm reverse import.
973 Severity : enhancement
975 Description: Communicate OST degraded/readonly state via statfs to MDS
976 Details : Flags in the statfs returned from OSTs indicate whether the
977 OST is in a degraded RAID state, or if the filesystem has
978 turned read-only after a filesystem error is detected.
983 Description: don't panic if EPROTO was hit when reading symlink
984 Details : correctly handling request reference in error cases.
986 Severity : enhancement
988 Description: Add LustreNetLink, a kernel-userspace communcation path.
990 Severity : enhancement
992 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
997 Description: access to llog context before init.
998 Details : move handling CATALOGS file at osc layer and forbid access to llog
1002 Frequency : rare, only when client is writing to OST on local node
1004 Description: Avoid deadlock for local client writes
1005 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1006 memory freeing context. This allows OST threads to set the
1007 PF_MEMALLOC flag on task structures in order to allocate memory
1008 from reserved pools and complete IO.
1009 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1010 so that the OST threads generate memory pressure and allow
1011 inactive pages to be reclaimed.
1013 Severity : enhancement
1015 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1020 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1021 Details : move ldlm namespace creation in setup phase to avoid grab
1022 _lprocfs_lock with cli_sem held.
1026 Description: Temporarily disable grant shrink.
1027 Details : Disable the feature for debugging.
1030 Frequency : only when formatting test filesystems
1032 Description: Unable to run several mkfs.lustre on loop devices at the same time
1033 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1034 formatting. The solution is to proper handle the error.
1036 Severity : enhancement
1038 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1040 Severity : enhancement
1042 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1047 Description: resolve race between obd_disconnect and class_disconnect_exports
1048 Details : if obd_disconnect will be called to already disconnected export he
1049 forget release one reference and osc module can't unloaded.
1051 Severity : enhancement
1053 Description: Allow tuning service thread via /proc
1054 Details : For each service a new
1055 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1056 created that can be used to set min/max thread counts, and get the
1057 current number of running threads.
1062 Descriptoin: don't return error if have partially created objects for file.
1063 Details : lov_update_create_set uses set->set_success as index for created
1064 objects, so if some requests will be failed, they will have hole
1065 at end of array and qos_shrink_lsm for allocate correct lsm.
1067 Severity : enhancement
1069 Description: Update OFED support to 1.4.1
1074 Description: Reduce small size read RPC
1075 Details : Set read-ahead limit for every file and only do read-ahead when
1076 available read-ahead pages are bigger than 1M to avoid small size
1079 Severity : enhancement
1081 Description: move AT tunable parameters for more consistent usage
1082 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1084 Severity : enhancement
1086 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1087 Details : allow skip disconnected ost for send statfs request and hide error
1093 Description: fix racy locking of mballoc block bitmaps causing BUG
1094 Details : The locking of the mballoc buddy bitmap and the in-memory
1095 block bitmap was using two different spin locks in some
1096 cases. This made it possible to incorrectly access the
1097 mballoc bitmap while another process was modifying it,
1098 causing a sanity assertion to fail. While no on-disk corruption
1099 was reported, there was some risk of this happening.
1102 Frequency : rare, on llog test 6
1104 Descriptoin: don't allow connect to already connected import
1105 Details : allowing connect to already connected import is hide connecting problem.
1110 Descriptoin: allow kill process which wait statahead result
1111 Details : for some reasons 'ls' can stick in waiting result from statahead,
1112 in this case need way for kill this process.
1114 Severity : enhancement
1116 Description: Add state history info file, enhance import info file
1117 Details : Track import connection state changes in a new osc/mdc proc file;
1118 add overview-type data to the osc/mdc import proc file.
1120 Severity : enhancement
1122 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1124 Severity : enhancement
1126 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1131 Descriptoin: OST grant too much space to client even there are not enough space.
1132 Details : Client will shrink its grant cache to OST if there are no write
1133 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1134 this grant cache if there are already not enough avaible space
1135 (left_space < total_clients * 32M).
1138 Frequency : start MDS on uncleanly shutdowned MDS device
1140 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1141 Details : stay in waiting mds<>ost recovery finished produce random bugs
1142 due race between two ll_sync thread for one lov target. send
1143 ACTIVATE event only if connect realy finished and import have
1147 Frequency : rare, connect and disconnect target at same time
1149 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1150 Details : don't call obd_disconnect under lov_lock. this long time
1151 operation and can block ptlrpcd which answer to connect request.
1156 Descriptoin: don't lose wakeup for imp_recovery_waitq
1157 Details : recover_import_no_retry or invalidate_import and import_close can
1158 both sleep on imp_recovery_waitq, but we was send only one wakeup
1162 Frequency : always with long access acl
1164 Descriptoin: mds can't pack reply with long acl.
1165 Details : mds don't control size of acl but they limited by reint/getattr
1168 Severity : enhancement
1170 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1172 Severity : enhancement
1174 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1177 Frequency : start MDS on uncleanly shutdowned MDS device
1179 Descriptoin: aborting recovery hang on MDS
1180 Details : don't throttle destroy RPCs for the MDT.
1183 Frequency : on remount
1185 Description: external journal device not working after the remount
1186 Details : clear dev_rdonly flag for external journal devices in
1192 Description: shutdown vs evict race
1193 Details : client_disconnect_export vs connect request race.
1194 if client will evicted at this time - we start invalidate
1195 thread without referece to import and import can be freed
1199 Frequency : rare, need acl's on inode.
1201 Description: client can't handle ost additional correctly
1202 Details : if ost was added after client connected to mds client can have
1203 hit lnet_try_match_md ... to big messages to wide striped files.
1204 in this case need teach client to handle config events about add
1205 lov target and update client max ea size at that event.
1207 Severity : enhancement
1209 Description: Changelogs
1210 Details : Changelogs are a lightweight mechanism to track filesystem
1211 metadata and namespace changes. The changelog is recorded
1212 permanently on the MDTs, and is periodically "consumed" / purged
1213 when records are no longer needed.
1215 Severity : enhancement
1217 Description: compact fld format with extents
1218 Details : Store range of seq rather than every seq in FLD. Seq
1219 controller update FLD rather than clients. In Case of CMD, mdt0
1220 has FLD, all other metadata server act as non persistent proxy
1221 for FLD queries and cache fld entries in fld cache.
1226 Description: don't skip ost target if they assigned to file
1227 Details : Drop slow OSCs if we can, but not for requested start idx.
1228 This means "if OSC is slow and it is not the requested
1229 start OST, then it can be skipped, otherwise skip it only
1230 if it is inactive/recovering/out-of-space.
1234 Description: more cleanup in mds_lov
1235 Details : not send LOV EA under replay, we can't know about they size at this
1236 time. Don't allow client connect to mds before any ost connected,
1237 for avoid problems with LOV EA size and returning EIO to client.
1239 Severity : enhancement
1241 Description: Interoperability at server side (Disk interoperability)
1243 Severity : enhancement
1245 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1247 Severity : enhancement
1249 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1251 Severity : enhancement
1253 Description: New client IO stack (CLIO).
1255 Severity : enhancement
1257 Description: Commit on sharing. Eliminate inter-client dependencies between
1258 uncommitted transactions by doing transaction commits.
1259 Thereby clients may recovery independently.
1262 Frequency : Create a symlink file with a very long name
1264 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1265 Details : If there is no extra space in the request for early cancels,
1266 ldlm_req_handles_avail() returns 0 instead of a negative value.
1268 Severity : enhancement
1270 Description: Add /proc entry for import status
1271 Details : The mdc, osc, and mgc import directories now have
1272 an import directory that contains useful import data for debugging
1273 connection problems.
1275 Severity : enhancement
1277 Description: Re-disable certain /proc logging
1278 Details : Enable and disable client's offset_stats, extents_stats and
1279 extents_stats_per_process stats logging on the fly.
1282 Frequency : Only on FC kernels 2.6.22+
1284 Description: oops in statahead
1285 Details : Do not drop reference count for the dentry from VFS when lookup,
1286 VFS will do that by itself.
1288 Severity : enhancement
1290 Description: Generic /proc file permissions
1291 Details : Set /Proc file permissions in a more generic way to enable non-
1292 root users operate on some /proc files.
1296 Description: Hitting mdc_commit_close() ASSERTION
1297 Details : Properly handle request reference release in
1298 ll_release_openhandle().
1302 Description: quota recovery deadlock during mds failover
1303 Details : This patch includes att18982, att18236, att18237 in bz14840.
1305 1. fix osts hang when mds does failover with quotaon
1306 2. prevent watchdog storm when osts threads wait for the
1311 Frequency : only patchless client
1312 Description: add workaround for race between add/remove dentry from hash
1314 Severity : enhancement
1316 Description: Allow OST glimpses to return PW locks
1320 Description: LBUG when llog conf file is full
1321 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1325 Description: Prevent import from entering FULL state when server in recovery
1329 Description: service mount cannot take device name with ":"
1330 Details : Only when device name contains ":/" will mount treat it as
1336 Description: replace ptlrpcd with the statahead thread to interpret the async
1337 statahead RPC callback
1341 Frequency : on recovery
1342 Description: I/O failures after umount during fail back
1343 Details : if client reconnected to restarted server we need join to recovery
1344 instead of find server handler is changed and process self
1345 eviction with cancel all locks.
1347 Severity : enhancement
1349 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1351 Severity : enhancement
1353 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1355 Severity : enhancement
1357 Description: Upcall on Lustre log has been dumped
1358 Details : Allow for a user mode script to be called once a Lustre log has
1359 been dumped. It passes the filename of the dumped log to the
1360 script, the location of the script can be specified via
1361 /proc/sys/lnet/debug_log_upcall.
1366 Description: avoid idr_remove called for id which is not allocated.
1367 Details : Move assigment s_dev for clustered nfs to end of initialization,
1368 to avoid problem with error handling.
1373 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1374 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1375 into CONN_USED_HASH and this prodice warning when put connection
1376 again in unused hash.
1379 Severity : enhancement
1381 Description: File striping can now be set to use an arbitrary pool of OSTs.
1383 Severity : enhancement
1385 Description: Export bytes_read/bytes_write count on OSC/OST.
1389 Description: Early reply size mismatch, MGC loses connection
1390 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1391 the connect flags are properly negotiated.
1396 Description: quotas are not honored with O_DIRECT
1397 Details : all writes with the flag O_DIRECT will use grants which leads to
1398 this problem. Now using OBD_BRW_SYNC to guard this.
1402 Description: add quota statistics
1403 Details : 1. sort out quota proc entries and proc code.
1404 2. add quota statistics
1406 Severity : enhancement
1408 Description: enable quota support for HEAD.
1412 Description: Properly propagate oinfo flags from lov to osc for statfs
1413 Details : restore missing copy oi_flags to lov requests.
1415 Severity : enhancement
1417 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1418 : l_getgroups(8), lst(8), routerstat(8)
1419 Details : included man pages for llobdstat(8), llstat(8),
1420 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1422 Severity : enhancement
1424 Description: configure's --enable-quota should check the
1425 : kernel .config for CONFIG_QUOTA
1426 Details : configure is terminated if --enable-quota is passed but
1427 : no quota support is in kernel
1431 Description: Remove portals compatibility
1432 Details : Remove portals compatibility, not interoperable with releases
1437 Description: Resolve device initialization race
1438 Details : Prevent proc handler from accessing devices added to the
1439 obd_devs array but yet be intialized.
1441 Severity : enhancement
1443 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1445 Severity : enhancement
1447 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1452 Description: Using wrong pointer in osc_brw_prep_request
1453 Details : Access to array[-1] can produce panic if kernel compiled with
1454 CONFIG_PAGE_ALLOC enabled
1458 Description: Client runs out of low memory
1459 Details : Consider only lowmem when counting initial number of llap pages
1463 Description: *optional* service tags registration
1464 Details : if the "service tags" package is installed on a Lustre node
1465 When the filesystem is mounted, a local-node service tag will
1466 be created. See http://inventory.sun.com/ for more information
1467 about the Service Tags asset management system.
1471 Description: Kernel BUG tries to release flock
1472 Details : Lustre does not destroy flock lock before last reference goes
1473 away. So always drop flock locks when client is evicted and
1474 perform unlock regardless of successfulness of speaking to MDS.
1478 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1482 Description: sanity test 65a fails if stripecount of -1 is set
1483 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1488 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1489 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1492 Severity : enhancement
1494 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1498 Description: do not process already freed flock
1499 Details : flock can possibly be freed by another thread before it reaches
1500 to ldlm_flock_completion_ast.
1504 Description: LBUG during stress test
1505 Details : Need properly lock accesses the flock deadlock detection list.
1509 Description: oops in page fault handler
1510 Details : kernel page fault handler can return two special 'pages' in error
1511 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1515 Description: timeout with invalidate import.
1516 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1517 should be handled by ptlrpcd. This produce long age waiting and
1518 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1520 Severity : enhancement
1522 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1526 Description: enable MGS and MDT services start separately
1527 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1528 a co-located MGS without starting the MGS, which is a complement
1529 to 'nosvc' mount option.
1533 Frequency : after recovery
1534 Description: precreate to many object's after del orphan.
1535 Details : del orphan st in oscc last_id == next_id and this triger growing
1536 count of precreated objects. Set flag LOW to skip increase count
1537 of precreated objects.
1541 Frequency : rare, on clear nid stats
1542 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1543 Details : when clean nid stats sometimes try destroy live entry,
1544 and this produce panic in free.
1548 Description: Stack overflow during MDS log replay
1549 ease stack pressure by using a thread dealing llog_process.
1553 Description: wait until IO finished before start new when do lock cancel.
1554 Details : VM protocol want old IO finished before start new, in this case
1555 need wait until PG_writeback is cleared until check dirty flag and
1556 call writepages in lock cancel callback.
1558 Severity : enhancement
1560 Description: using special macro for print time and cleanup in includes.
1564 Description: mds_mfd_close() ASSERTION(rc == 0)
1565 Details : In mds_mfd_close(), we need protect inode's writecount change
1566 within its orphan write semaphore to prevent possible races.
1570 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1574 Frequency : rare, on shutdown ost
1575 Description: don't hit live lock with umount ost.
1576 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1577 use shrink_dcache_sb instead.
1581 Description: don't panic with use echo client
1582 Details : echo client pass NULL as client nid pointer and this produce null
1583 pointer dereference.
1587 Description: fix build on ppc32
1588 Details : compile code with -m64 flag produce wrong object file for ppc32.
1592 Description: add message levels for liblustreapi
1596 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1597 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1598 will result in an -EAGAIN and in the exit call path we call
1599 alloc_rr() which will with increasing aggressiveness attempt to
1600 aquire precreated objects on the minimum number of required OSCs.
1604 Description: Use old size assignment to avoid deadlock
1605 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1606 the scheduling while holding a spinlock. We do not need locking
1607 for size in ll_update_inode() because size is only updated from
1608 the MDS for directories or files without objects, so there is no
1609 other place to do the update, and concurrent access to such inodes
1610 are protected by the inode lock.
1614 Description: resolve "_IOWR redefined" build error on SLES10
1618 Description: dump the memory debugging after all modules are unloaded to
1619 suppress false negative in conf_sanity test 39
1621 Severity : enhancement
1623 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1628 Description: fix SLES kernel versioning
1629 Details : the kernel version for our SLES 10 kernel did not include a "-"
1630 before the "smp" at the end. while this was not a problem in
1631 general, it did mean that software trying to use the kernel
1632 version to try to detect a vendor specific kernel would fail.
1633 this was most evident by the OFED build scripts.
1637 Description: Don't update lov_desc members until making sure they are valid
1638 Details : When updating lov_desc members via proc fs, need fix their
1639 validities before doing the real update.
1643 Description: don't put request into delay list while invalidate in flight.
1644 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1645 import in flight. this produce timeout for invalidate and sometimes
1646 can cause stale data.
1650 Frequency : on ppc only
1651 Description: not convert ost objects for directory because it's not exist.
1652 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1656 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1658 Severity : enhancement
1660 Description: Update to support 2.6.22.14 vanilla kernel.
1664 Frequency : rare, on recovery
1665 Description: read procfs can produce deadlock in some situation
1666 Details : Holding lprocfs lock which send rpc can produce block for destroy
1667 obd objects and this also block reconnect with -EALREADY. This
1668 isn't fix all lprocfs bugs - but make it rare.
1670 Severity : enhancement
1672 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1675 Frequency : frequent on X2 node
1677 Description: mdc_set_open_replay_data LBUG
1678 Details : Set replay data for requests that are eligible for replay.
1682 Description: lustre_mgs: operation 101 on unconnected MGS
1683 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1684 MGC, and late on MGC cannot successfully connect to MGS and a lot
1685 of the error messages complaining that MGS is not connected.
1689 Frequency : on network error
1690 Description: panic with double free request if network error
1691 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1692 but it's true only for synchronus enqueue, for async enqueue
1693 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1696 Severity : enhancement
1698 Description: client-side metadata stat-ahead during readdir(directory readahead)
1699 Details : perform client-side metadata stat-ahead when the client detects
1700 readdir and sequential stat of dir entries therein
1703 Frequency : on start mds
1705 Description: Implement get_info(last_id) in obdfilter.
1708 Frequency : occasional
1710 Description: Correctly check stale fid, not start epoch if ost not support SOM
1711 Details : open with flag O_CREATE need set old fid in op_fid3 because
1712 op_fid2 was overwritten with new generated fid, but mds can answer
1713 with one of these two fids and both is not stale. Setattr
1714 incorrectly started an epoch and assume will be called
1715 done_writing, but without SOM done_writing ever being called.
1718 Frequency : rare, depends on device drivers and load
1720 Description: MDS or OSS nodes crash due to stack overflow
1721 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1722 In some cases, in conjunction with device drivers that use a lot
1723 of stack the MDS (or possibly OSS) service threads could overflow
1724 the stack. One change which was identified to consume additional
1725 stack has been reworked to avoid the extra stack usage.
1728 Frequency : occasional
1730 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1731 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1732 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1733 oscc_has_objects and see OSC has no objests and call
1734 oscc_internal_create to resend create request.
1736 Severity : enhancement
1738 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1740 Severity : enhancement
1742 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1745 Frequency : very rare
1747 Description: Fix replay if there is an un-replied request and open
1748 Details : In some cases, older replay request will revert the
1749 mcd->mcd_last_xid on MDS which is used to record the client's
1750 latest sent request.
1752 Severity : enhancement
1754 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1756 Severity : enhancement
1758 Description: Add rhel5 support to HEAD.
1760 Serverity : enhancement
1762 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1767 Description: Don't allow skipping OSTs if index has been specified.
1768 Details : Don't allow skipping OSTs if index has been specified, make
1769 locking in internal create lots better.
1773 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1774 Details : access to bitfield in structure is always rounded to long
1775 and this produce problem with not atomic change any bit.
1779 Description: Lustre make rpms failed.
1780 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1781 builds Lustre rpms from tarball.
1784 Frequency : rare on shutdown ost
1786 Description: If llog cancel was not send before clean_exports phase, this can
1787 produce deadlock in llog code.
1788 Details : If llog thread has last reference to obd and call class_import_put
1789 this produce deadlock because llog_cleanup_commit_master wait when
1790 last llog_commit_thread exited, but this never success because was
1791 called from llog_commit_thread.
1795 Description: allow userland application know is lost one of stripes.
1796 Details : fill lvb_blocks with error code on ost and return it to
1797 application if error flag found.
1801 Description: NULL lov_tgts causing MDS oops
1802 Details : more safe checks for NULL lov_tgts for avoid oops.
1804 Severity : enhancement
1806 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1810 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1811 Details : If an OST is down the MDS will hang indefinitely in
1812 obd_statfs_rqset() waiting for the statfs data. While for
1813 MDS QOS usage of statfs, it should not stuck in waiting.
1815 Severity : enhancement
1817 Description: remote_acl support
1818 Details : Support ACL-based permission check for remote user.
1819 Support setfacl/getfacl for remote user with the utils
1820 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1821 the system "{s,g}etfacl" utils.
1823 Severity : enhancement
1825 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1827 Severity : enhancement
1829 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1833 Description: make mgs_setparam() handle fsname containing dash
1834 Details : fsname containing a dash does not work with lctl conf_param
1836 Severity : enhancement
1838 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1840 Severity : enhancement
1842 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1846 Frequency : rare, at shutdown
1847 Description: access already free / zero obd_namespace.
1848 Details : if client_disconnect_export was called without force flag set,
1849 and exist connect request in flight, this can produce access to
1850 NULL pointer (or already free pointer) when connect_interpret
1851 store ocd flags in obd_namespace.
1855 Frequency : only at startup
1856 Description: not alloc memory with spinlock held.
1857 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1858 if any spinlock held.
1860 Severity : enhancement
1862 Description: make lustre randomly fail allocating memory
1863 Details : Make lustre randomly failed allocating memory for testing purpose.
1865 Severity : enhancement
1867 Description: lost problems with lov objid file
1868 Details : Fixes some scability and access to not inited memory problems
1869 in work with lov objdid file.
1874 Description: lfs find does not continue on file error
1875 Details : Continue other files processing when a file/dir is absent.
1879 Description: Inconsistent usage of lustre_pack_reply()
1880 Details : Standardize the usage of lustre_pack_reply() such that it
1881 always generate a CERROR on failure.
1884 Frequency : occasional
1886 Description: MDS hang or stay in waiting lock
1887 Details : If client receive lock with CBPENDING flag ldlm need send lock
1888 cancel as separate rpc, to avoid situation when cancel request
1889 can't processed due all i/o threads stay in wait lock.
1893 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1897 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1901 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1905 Description: Update to RHEL4 latest kernel.
1907 Severity : enhancement
1909 Description: Build SLES10 patchless client fails
1910 Details : The configure was broken by run ./configure with
1911 --with-linux-obj=.... argument for patchless client. When the
1912 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1913 can't search header adequately. Use absolute path such as
1914 -I($LINUX)/include instead.
1918 Description: interrupt oig_wait produce painc on resend.
1919 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1920 private set, and this produce situation when rq_ptlrpcd_data not
1921 copyed to new allocated request and triggered LBUG on assert
1922 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1923 ptlrpcd set if request is changed and can be safety changed to use
1928 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1929 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1930 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1931 these macros and replace them with their difinitions instead.
1935 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1936 Details : Update which_patch & target file for SLES10 latest kernel.
1939 Description: Remove client patches from SLES 10 kernel.
1940 Details : This causes SLES 10 clients to behave as patchless clients
1941 even on a Lustre-patched (server) kernel.
1943 Severity : enhancement
1945 Description: self-adjustable client's lru lists
1946 Details : use adaptive algorithm for managing client cached locks lru
1947 lists according to current server load, other client's work
1948 pattern, memory activities, etc. Both, server and client
1949 side namespaces provide number of proc tunables for controlling
1952 Severity : enhancement
1954 Description: light-weight GSS support
1955 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1956 and reduce performance overhead.
1958 Severity : enhancement
1960 Description: Linux keyring support
1961 Details : Support using service of Linux keyring for Lustre GSS internal
1962 context refresh/cache mechanism.
1966 Description: Fix errors in lfs documentation
1967 Details : Fixes man pages
1971 Description: don't use GFP_* in generic Lustre code.
1972 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1976 Description: obdclass is limited by single OBD_ALLOC(idarray)
1977 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1981 Description: warnings with build patchless client with vanila 2.6.19 and up
1982 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1983 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1987 Frequency : Only for SLES
1988 Description: sanity_quota fail test_1
1989 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1990 file, which makes TSTID[2] inunique.
1994 Frequency : Only for SLES9
1995 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1996 Details : Update target/ChangeLog/which_patch .
2000 Frequency : Only for test_10 in sanity-quota.sh
2001 Description: a bug in quota test code
2002 Details : checking a test flag in a wrong place
2006 Frequency : Only for test_13 in sanity-quota.sh
2007 Description: a bug in quota test script
2008 Details : shouldn't check overfull allocated quota
2012 Description: jbd statistics
2013 Details : Port older jbd statistics patch for sles10
2015 Severity : enhancement
2017 Description: Add printing inode info into message about error in writepage.
2022 Description: reply_lock_interpret crash due to race with it and lock cancel.
2023 Details : Do not replay locks that are being cancelled. Do not reference
2024 locks by their address during replay, just by their handle.
2028 Frequency : When flocks are used.
2029 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2030 Details : Flock locks might destroy just granted lock if it could be merged
2031 with another existing flock, this is done in completion handler,
2032 so teach ldlm_cli_enquque_fini that this is a valid case for
2038 Description: Oops in read and write path when failing to allocate lock.
2039 Details : Check if lock allocation failed and return error back.
2043 Description: lstripe command fails for valid OST index
2044 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2045 instead of lov->desc.ld_active_tgt_count.
2049 Description: sanity.sh failed test 103
2050 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2055 Description: df doesn't work properly if diskfs blocksize != 4K
2056 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2060 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2061 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2063 Severity : enhancement
2065 Description: Adaptive timeouts
2066 Details : RPC timeouts adapt to changing server load and network
2067 conditions to reduce resend attempts and improve recovery time.
2071 Description: llapi_file_create() does not allow some changes
2072 Details : add llapi_file_open() that allows specifying the mode and
2073 open flags, and also returns an open file handle.
2077 Description: merge and cleanup kernel patches.
2078 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2082 Description: Add journal checksum support.(Kernel part)
2083 Details : The journal checksum feature adds two new flags i.e
2084 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2085 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2086 indicates that the commit block contains the checksum for
2087 the blocks described by the descriptor blocks. Now commit
2088 record can be sent to disk without waiting for descriptor
2089 blocks to be written to disk. This behavior is controlled
2090 using JBD2_FEATURE_ASYNC_COMMIT flag.
2094 Description: OSS needs mutliple precreate threads
2095 Details : Add ability to start more than one create thread per OSS.
2099 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2100 Details : execute lfs setstripe on client
2104 Description: mds_obd_create error creating tmp object
2105 Details : When the user sets quota on root, llog will be affected and can't
2106 create files and write files.
2109 Frequency : Always on ia64 patchless client, and possibly others.
2111 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2112 Details : This allows the patchless client to be loaded on architectures
2113 without this export.
2117 Description: RedHat Update kernel for RHEL5
2118 Details : Add the kernel config file for RHEL5.
2122 Description: RedHat Update kernel for RHEL5
2123 Details : Modify the kernel config file more closer RHEL5.
2127 Description: Build failure against Centos5 (RHEL5)
2128 Details : Define PAGE_SIZE when it isn't present.
2132 Description: test_7 on sanity-quota.sh can't be used on separate machines
2133 Details : add facet to handle it
2137 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2138 Details : with NFS, the anon dentry's parent was set to itself in
2139 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2140 corresponding dentry other than use rec->ur_name.
2142 Severity : enhancement
2144 Description: lfs setstripe enhancement
2145 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2149 Description: enable data checksumming by default
2150 Details : enable checksum by default, allow --disable-checksum
2151 configure option and "-o nochecksum" mount option. Checksums
2152 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2156 Description: lustre support for RHEL5
2157 Details : Add support for RHEL5.
2161 Description: Client eviction due to failover config
2162 Details : after a connection loss, the lustre client should attempt to
2163 reconnect to the last active server first before trying the
2164 other potential connections.
2168 Description: when mds and osts use different quota unit(32bit and 64bit),
2169 quota will be released repeatly.
2170 Details : void sending multiple quota reqs to mds, which will keep the
2171 status between the reqs.
2175 Description: osts not allocated evenly to files
2176 Details : change the condition to increase offset_idx
2179 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2181 Description: Data corruption for OSTs that are formatted larger than 2TB
2183 Details : When generating the bio request for lustre file writes the
2184 sector number would overflow a temporary variable before being
2185 used for the IO. The data reads correctly from Lustre (which
2186 will overflow in a similar manner) but other file data or
2187 filesystem metadata may be corrupted in some cases.
2191 Description: Tune the kernel for good SCSI performance.
2192 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2193 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2198 Description: rewrite ext2-derived code in obdclass/uuid.c
2199 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2200 from scratch preserving functionality.
2204 Description: "lfs find -obd UUID" prints directories
2205 Details : "lfs find -obd UUID" will return all directory names instead
2206 of just file names. It is incorrect because the directories
2207 do not reside on the OSTs.
2211 Description: use i_size_read and i_size_write in 2.6 port
2212 Details : replace inode->i_size access with i_size_read/write()
2215 Frequency : when removing large files
2217 Description: scheduling issue during removal of large Lustre files
2218 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2219 It causes scheduling issues when removing large files (17TB in the
2223 Frequency : only with liblustre clients on XT3
2225 Description: evictions taking too long
2226 Details : allow llrd to evict clients directly on OSTs
2229 Frequency : only on ppc
2231 Description: /proc/fs/lustre/devices broken on ppc
2232 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2233 We should make sure the type of 'index' is loff_t and then cast
2234 explicitly as needed below. Do not assign an explicitly cast
2238 Frequency : only for rhel5
2240 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2241 Details : Modify the target file & which_kernel.
2243 Severity : enhancement
2245 Description: omit set fsid for export NFS
2246 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2247 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2251 Frequency : Always, for kernels after 2.6.16
2252 Description: Fix warning idr_remove called for id=.. which is not allocated.
2253 Details : Last kernels save old s_dev before kill super and not allow
2254 to restore from callback - restore it before call kill_anon_super.
2258 Description: buffer overruns could theoretically occur
2259 Details : llapi_semantic_traverse() modifies the "path" argument by
2260 appending values to the end of the origin string, and a
2261 overrun may occur. Adding buffer overrun check in liblustreapi.
2265 Description: Fix error on 'ls .' at the top of the Lustre mount.
2266 Details : Don't revalidate dentry if it is a root dentry.
2270 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2271 Details : Modify vm-tunables-rhel4.patch.
2275 Description: Kernel config for 2.6.18-vanilla.
2276 Details : Modify targets/2.6-vanilla.target.in.
2277 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2278 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2279 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2280 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2284 Description: improve handling recoverable errors
2285 Details : if request processig with error which can be recoverable on server
2286 request should be resend, otherwise page released from cache and
2291 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2292 Details : Modify target file & which_patch.
2293 A flaw was found in the IA32 system call emulation provided
2294 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2295 value could be stored in the %RAX register, which could trigger an
2296 out-of-bounds system call table access. An untrusted local user
2297 could exploit this flaw to run code in the kernel
2298 (ie a root privilege escalation). (CVE-2007-4573).
2302 Description: O_DIRECT bypasses client statistics.
2303 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2304 read_ahead_stats were updated. copied the stats section from
2305 osc_send_oap_rpc() into async_internal().
2309 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2313 Description: change order of libsysio includes
2314 Details : '#include sysio.h' should always come before '#include xtio.h'
2318 Description: handle "serious error: objid * already exists" more gracefully
2319 Details : If LAST_ID value on disk is smaller than the objects existing in
2320 the O/0/d* directories, it indicates disk corruption and causes an
2321 LBUG(). If the object is 0-length, then we should use the existing
2322 object. This will help to avoid a full fsck in most cases.
2324 Severity : enhancement
2326 Description: adapt the lustre_config script to support the upgrade case
2327 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2328 server targets to 1.6 in parallel.
2332 Description: To avoid grant space > avaible space when the disk is almost
2333 full. Without this patch you might see the error "grant XXXX >
2334 available" or some LBUG about grant, when the disk is almost
2336 Details : In filter_check_grant, for non_grant cache write, we should
2337 check the left space by if (*left > ungranted + bytes), instead
2338 of (*left > ungranted), because only we are sure the left space
2339 is enough for another "bytes", then the ungrant space should be
2340 increase. In client, we should update cl_avail_grant only there
2341 is OBD_MD_FLGRANT in the reply.
2345 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2346 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2347 described in CVE-2007-4573. This problem could allow untrusted
2348 local users to gain root access.
2351 Frequency : when using O_DIRECT and quotas
2353 Description: Incorrect file ownership on O_DIRECT output files
2354 Details : block usage reported by 'lfs quota' does not take into account
2355 files that have been written with O_DIRECT.
2360 Description: touch file failed when fs is not full
2361 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2362 otherwise we can get ENOSP while fs is not full.
2366 Description: parallel lock callbacks
2367 Details : Instead of sending blocking and completion callbacks as separated
2368 requests, adding them to a set and sending in parallel.
2371 Frequency : only for Cray XT3
2372 Bugzilla : 12829/13455
2373 Description: Changing primary group doesn't change the group lustre assigns to
2375 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2376 overridden with the primary group provided by the group upcall,
2377 whereas the supplied fsgid can be trusted if it is in the list of
2378 supplementary groups returned by the group upcall.
2380 Severity : enhancement
2382 Description: Allow masking D_WARNING, D_ERROR messages from console
2383 Details : Console messages can now be disabled via lnet.printk.
2387 Description: User code with malformed file open parameter crashes client node
2388 Details : Before packing join_file req, all the related reference should be
2389 checked carefully in case some malformed flags cause fake
2390 join_file req on client.
2394 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2395 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2396 mode is equal to the granted mode. However, it should also check
2397 that LDLM_FL_CP_REQD is not set.
2401 Description: Heavy nfs access might result in deadlocks
2402 Details : After ELC code landed, it is now improper to enqueue any mds
2403 locks under och_sem, because enqueue might want to decide to
2404 cancel open locks for same inode we are holding och_sem for.
2408 Description: Client eviction while running blogbench
2409 Details : A lot of unlink operations with concurrent I/O can lead to a
2410 deadlock causing evictions. To address the problem, the number of
2411 oustanding OST_DESTROY requests is now throttled to
2412 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2413 callbacks are processed in priority.
2417 Description: enable ACLs on MDS by default
2418 Details : ACLs must be enabled on MDS by default.
2421 Frequency : PPC/PPC64 only
2423 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2424 Details : fix duplicated definitions between asm-ppc64/types.h and
2425 lustre_types.h on PPC.
2428 Frequency : PPC/PPC64 only
2430 Description: asm-ppc/segment.h does not exist
2431 Details : fix compile issue on PPC.
2435 Description: better handle error messages in extents code
2438 Frequency : RHEL4 only
2440 Description: mkfs is very slow on IA64/RHEL4
2441 Details : A performance regression has been discovered in the MPT Fusion
2442 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2443 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2444 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2446 Severity : enhancement
2448 Description: SNMP support enhancement
2449 Details : Adding total number of sampled request for an MDS node in snmp
2452 Severity : enhancement
2454 Description: Optimize ldlm waiting list processing for PR extent locks
2455 Details : When processing waiting list for read extent lock and meeting read
2456 lock that is same or wider to it that is not contended, skip
2457 processing rest of the list and immediatelly return current
2458 status of conflictness, since we are guaranteed there are no
2459 conflicting locks in the rest of the list.
2463 Description: Time out and refuse to reconnect
2464 Details : When the failover node is the primary node, it is possible
2465 to have two identical connections in imp_conn_list. We must
2466 compare not conn's pointers but NIDs, otherwise we can defeat
2467 connection throttling.
2471 Description: port llog fixes from b1_6 into HEAD
2472 Details : Port llog reference couting and some llog cleanups from b1_6
2473 (bug 10800) into HEAD, for protect from panic and access to already
2474 free llog structures.
2478 Description: Detect stride IO mode in read-ahead
2479 Details : When a client does stride read, read-ahead should detect that and
2480 read-ahead pages according to the detected stride pattern.
2484 Description: data checksumming impacts single node performance
2485 Details : add support for several checksum algorithm. Currently, only CRC32
2486 and Adler-32 are supported. The checksum type can be changed on
2487 the fly via /proc/fs/lustre/osc/*/checksum_type.
2491 Description: use adler32 for page checksums
2492 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2497 Description: build for x2 fails
2498 Details : fix compile issue on Cray systems.
2502 Description: Properly match for duplicate locks
2503 Details : Due to different lock order from skiplists code, we need to
2504 traverse entire list for now
2507 Frequency : only on PPC/SLES10
2509 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2510 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2511 be defined. Add a hack in mkfs_lustre.c to work around this compile
2516 Description: LASSERT on MDS when client holding flock lock dies
2517 Details : ldlm pool logic depends on number of granted locks equal to
2518 number of released locks which is not true for flock locks, so
2519 just exclude such locks from consideration.
2523 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2524 Details : Use fsfilt_sync() for both the whole filesystem sync and
2525 individual file sync to eliminate dangerous inode locking
2526 with I_LOCK that can lead to a deadlock.
2530 Description: performance in 1.6.3
2531 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2532 sure that 1MB requests are not fragmented by the block layer.
2534 Severity : enhancement
2536 Description: organize the server-side client stats on per-nid basis
2537 Details : Change the structure of stats under obdfilter and mds to
2545 The "uuid"s file would list the uuids of _active_ exports.
2546 And the clear entry is to clear all stats and stale nids.
2548 Severity : enhancement
2550 Description: eliminate client locks in face of contention
2551 Details : file contention detection and lockless i/o implementation
2552 for contended files.
2556 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2559 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2561 Description: MDS may lose file striping (and hence file data) in some cases
2562 Details : If there are additional extended attributes stored on the MDS,
2563 in particular ACLs, SELinux, or user attributes (if user_xattr
2564 is specified for the client mount options) then there is a risk
2565 of attribute loss. Additionally, the Lustre file striping
2566 needs to be larger than default (e.g. striped over all OSTs),
2567 and an additional attribute must be stored initially in the
2568 inode and then increase in size enough to be moved to the
2569 external attribute block (e.g. ACL growing in size) for file
2574 Description: skiplist implementation simplification
2575 Details : skiplists are used to group compatible locks on granted list
2576 that was implemented as tracking first and last lock of each
2577 lock group the patch changes that to using doubly linked lists
2581 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2582 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2583 because LASSERT on both the data supplied by a client, and the
2584 data on disk is dangerous and incorrect.
2586 Severity : enhancement
2588 Description: Slow truncate/writes to huge files at high offsets.
2589 Details : Directly associate cached pages to lock that protect those pages,
2590 this allows us to quickly find what pages to write and remove
2591 once lock callback is received.
2595 Description: more ldlm soft lockups
2596 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2597 starve other threads from the resource lock for a long time in
2598 case of long waiting queue, so change the debug level from
2599 D_OTHER to the less frequently used D_INFO.
2601 Severity : enhancement
2603 Description: add -gid, -group, -uid, -user options to lfs find
2607 Description: Hung threads in invalidate_inode_pages2_range
2608 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2609 once one is completed. As a result, some RPCs are stuck in the
2610 queue and are never sent.
2614 Description: filter threads hungs on waiting journal commit
2615 Details : Cleanup filter group llog code, then only filter group llog will
2616 be only created in the MDS/OST syncing process.
2620 Description: Procfs and llog threads access destoryed import sometimes.
2621 Details : Sync the import destoryed process with procfs and llog threads by
2622 the import refcount and semaphore.
2624 Severity : enhancement
2626 Description: openlock cache of b1_6 port to HEAD
2631 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2632 Details : If insertion of an extent fails, then discard the inode
2633 preallocation and free data blocks else it can lead to duplicate
2638 Description: don't always update ctime in ext3_xattr_set_handle()
2639 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2640 In some cases the ctime should not be updated, for example for
2641 2.0->1.8 compatibility it is necessary to delete an xattr and it
2642 should not update the ctime.
2646 Bugzilla : 15713/16362
2647 Description: Assertion in iopen_connect_dentry in 1.6.3
2648 Details : looking up an inode via iopen with the wrong generation number can
2649 populate the dcache with a disconneced dentry while the inode
2650 number is in the process of being reallocated. This causes an
2651 assertion failure in iopen since the inode's dentry list contains
2652 both a connected and disconnected dentry.
2656 Description: assertion failure in ldlm_handle2lock()
2657 Details : fix a race between class_handle_unhash() and class_handle2object()
2658 introduced in lustre 1.6.5 by bug 13622.
2663 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2664 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2669 Description: oops in sd_iostats_seq_show()
2670 Details : unloading/reloading the scsi low level driver triggers a kernel
2671 bug when trying to access the sd iostat file.
2676 Description: Kernel panics during QLogic driver reload
2677 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2678 patch, causing memory corruption.
2683 Description: journal_dev option does not work in b1_6
2684 Details : pass mount option during pre-mount.
2686 Severity : enhancement
2688 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2689 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2690 extent information of a file. It can be used to map logical blocks
2691 in a file to physical blocks in the block device.
2695 Description: LDLM soft lockups - improvement
2696 Details : It is be possible to send the lock handle along with each read
2697 or write request because the client is already doing a lock match
2698 itself so there isn't any reason the OST should have to re-do that
2702 Frequency : only on Cray X2
2704 Description: X2 build failures
2705 Details : fix build failures on Cray X2.
2709 Description: xid & resent requests
2710 Details : Initialize RPC XID from clock at startup (randomly if clock is
2713 Severity : enhancement
2715 Description: Add lustre_start utility to start or stop multiple Lustre servers
2720 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2721 Details : In case of memory pressure, list_del() can be called twice on
2722 req->rq_history_list, causing a kernel oops.
2726 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2727 Details : fix stack overflow in the distributed lock manager by defering
2728 export eviction after a failed AST to the elt thread instead of
2729 handling it in the dlm interpret routine.
2733 Description: Convert some comments to new format.
2734 Details : Update documenting comments to match doxygen conventions.
2738 Description: Grammar fixes.
2739 Details : A couple of trivial spelling fixes.
2743 Description: OSD_COUNTERS-mandatory
2744 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2745 to simplify the code.
2749 Description: simplify cmm_device freeing logic.
2750 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2754 Description: Add lockdep support to dt_object_operations locking interface.
2755 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2756 parameter indicating lock ordering. Update mdd code to use new
2761 Description: Introduce failloc constants for lockless IO tests.
2762 Details : Add two new failloc constants to test lockless IO. Only one of
2763 them in implemented---another is checked in yet to be landed
2768 Description: Add lockdep support for inode mutex.
2769 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2770 in the situations where Lustre has to lock more than one inode
2775 Description: Add optional invariants checking support.
2776 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2777 configure switch. This macro is to be used for consistency and
2778 sanity checks that are too expensive to be left in `production'
2783 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2784 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2789 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2790 Details : Use spin_lock_nested() in (the only) situation where more than
2791 one ldlm_lock is locked simultaneously. Also, fix possible
2792 dead-lock in ldlm_lock_change_resource() by enforcing particular
2797 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2798 Details : Instead of specifying each ldlm_lock call-back through separate
2799 parameter, wrap them into struct ldlm_callback_suite.
2803 Description: Kill join_lru obd method and its callers.
2804 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2805 in memory---a requirement implemented through ->o_join_lru() obd
2806 method in HEAD. Remove this method and its users.
2810 Description: Add asynchronous ldlm ENQUEUE completion handler.
2811 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2812 case that stock ldlm_completion_ast() cannot handle as it waits
2813 until lock is granted. Introduce new ldlm_completion_ast_async()
2814 for this. Also comment ldlm_completion_ast().
2818 Description: ldlm_error <-> errno conversion.
2819 Details : Add functions to map (rather arbitrary) between LDLM error codes
2820 and standard errno values. CLIO needs this to prevent LDLM specific
2821 constants from escaping ldlm and osc.
2825 Description: Kill unused ldlm_handle2lock_ns() function.
2826 Details : Kill unused ldlm_handle2lock_ns() function.
2830 Description: Add lu_ref support to ldlm_lock
2831 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2832 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2833 and ldlm_resource. LDLM interface has to be changed, because code that
2834 releases a reference on a lock, has to "know" what reference this is.
2835 In the most frequent case
2837 lock = ldlm_handle2lock(handle);
2839 LDLM_LOCK_PUT(lock);
2841 no changes are required. When any other reference (received _not_ from
2842 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2843 instead of LDLM_LOCK_PUT().
2845 Arguably, changes are pervasive, and interface requires some discipline
2846 for proper use. On the other hand, it was very instrumental in finding
2847 a few leaked lock references.
2851 Description: Add ldlm_lock_addref_try().
2852 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2853 attempts to addref a lock that might be being canceled
2858 Description: Add ldlm_weigh_callback().
2859 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2860 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2861 hard-coded `number of pages' logic.
2865 Description: Add lockdep annotations to llog code.
2866 Details : Use appropriately tagged _nested() locking calls in the places
2867 where llog takes more than one ->lgh_lock lock.
2871 Description: Add loi_kms_set().
2872 Details : Wrap kms updates into a helper function.
2876 Description: Constify instances of struct lsm_operations.
2877 Details : Constify instances of struct lsm_operations.
2881 Description: lu_conf support.
2882 Details : On a server, a file system object is uniquely identified
2883 by a fid, which is sufficient to locate and load all object
2884 state (inode). On a client, on the other hand, more data are
2885 necessary instantiate an object. Change lu_object_find() and
2886 friends to take additional `lu_conf' argument describing object.
2887 Typically this includes layout information.
2891 Description: lu_context fixes.
2892 Details : Introduce new lu_context functions that are needed on the client
2893 side, where some system threads (ptlrpcd) are shared by multiple
2894 modules, and so cannot be stopped during module shutdown.
2898 Description: Add start and stop methods to lu_device_type_operations.
2899 Details : Introduce two new methods in lu_device_type_operations, that are
2900 invoked when first instance of a given type is created and last one
2901 is destroyed respectively. This is need by CLIO.
2905 Description: Add lu_ref support to struct lu_device.
2906 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2907 track leaked references.
2911 Description: Introduce lu_kmem_descr.
2912 Details : lu_kmem_descr and its companion interface allow to create
2913 and destroy a number of kmem caches at once.
2917 Description: Fix lu_object finalization race.
2918 Details : Fix a race between lu_object_find() finding an object and its
2919 concurrent finalization. This race is (most likely) not possible
2920 on the server, but might happen on the client.
2924 Description: Introduce lu_ref interface.
2925 Details : lu_ref is a debugging module allowing to track references to
2926 a given object. It is quite cpu expensive, and has to be
2927 explicitly enabled with --enable-lu_ref. See usage description
2932 Description: Factor lu_site procfs stats into a separate function.
2933 Details : Separate lu_site stats printing code into a separate function
2934 that can be reused on a client.
2938 Description: Constify instances of struct {lu,dt,md}_device_operations.
2939 Details : Constify instances of struct {lu,dt,md}_device_operations.
2943 Description: Introduce struct md_site and move meta-data specific parts of
2944 struct lu_site here.
2945 Details : Move md-specific fields out of struct lu_site into special struct
2946 md_site, so that lu_site can be used on a client.
2950 Description: Kill mdd_lov_destroy().
2951 Details : Remove unused mdd code.
2955 Description: Add st_block checking to multistat.c.
2956 Details : Add st_block checking to multistat.c.
2960 Description: Add lu_ref support to struct obd_device.
2961 Details : Add lu_ref tracking to obd_device.
2965 Description: Kill obd_set_fail_loc().
2966 Details : Remove unused code.
2970 Description: Add special type for ptlrpc_request interpret functions.
2971 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2972 there. Actual usage will be in CLIO.
2976 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2977 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2978 doesn't work with lockdep.
2982 Description: Add rwv.c test program.
2983 Details : New testing program exercising readv(2) and writev(2) (Qian).
2987 Description: Add sendfile.c test program.
2988 Details : New testing program exercising sendfile(2) (Jay).
2992 Description: Ratelimit a message that can be very frequent.
2993 Details : Ratelimit a memory allocation failure message that can
2998 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2999 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3003 Description: Liblustre build fixes.
3004 Details : Liblustre build fixes.
3008 Description: libcfs: add cfs_{need,cond}_resched() interface.
3009 Details : libcfs: add cfs_{need,cond}_resched() definition and
3010 implementations for Linux, NT, and liblustre.
3012 Severity : enhancement
3014 Description: More exported tunables for mballoc
3015 Details : Add support for tunable preallocation window and new tunables for
3016 large/small requests
3020 Description: Detect corruption of block bitmap and checking for preallocations
3021 Details : Checks validity of on-disk block bitmap. Also it does better
3022 checking of number of applied preallocations. When corruption is
3023 found, it turns filesystem readonly to prevent further corruptions.
3027 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3028 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3029 window to 0. If it does hit cache miss, and read-pattern is still
3030 stride-io mode, does not reset the stride window, but also does not
3031 increase the stride window length in this case.
3035 Frequency : only for big-endian servers
3036 Description: Check if system is big-endian while mounting fs with extents feature
3037 Details : Mounting a filesystem with extents feature will fail on big-endian
3038 systems since ext3-based ldiskfs is not supported on big-endian
3039 systems. This can be over-riden with "bigendian_extents" mount option.
3041 Severity : enhancement
3043 Description: The root squash functionality
3044 Details : A security feature, which is to prevent users from being able
3045 to mount lustre on their desktop, run as root, and delete
3046 all of the files in the filesystem. The goal is accomplished by
3047 remapping user id (UID) and group id (GID) of the root user to
3048 a UID and GID specified by the system administartor via Lustre
3049 configuration management server (MGS). The functionality also
3050 allows to specify sets of clients for which the remapping does
3055 Description: Excessive recovery window
3056 Details : With AT enabled, the recovery window can be excessively long (6000+
3057 seconds). To address this problem, we no longer use
3058 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3059 timeout no longer depends on the service time, it is set to
3060 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3061 time via pb_service_time.
3065 Description: Watchdog triggered on MDS failover
3066 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3067 the OSTs know that the MDS "UUID" can be reused for the same export
3068 from a different NID, so we do not need to wait for the export to be
3072 Frequency : rare, only if using MMP with Linux RAID
3074 Description: MMP doesn't work with Linux RAID
3075 Details : While using HA for Lustre servers with Linux RAID, it is possible
3076 that MMP will not detect multiple mounts. To make this work we
3077 need to unplug the device queue in RAID when the MMP block is being
3078 written. Also while reading the MMP block, we should read it from
3079 disk and not the cached one.
3081 Severity : enhancement
3083 Description: open file using fid
3084 Details : A file can be opened using just its fid, like
3085 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3088 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3091 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3092 Details : add proper locking for creation of jbd_4k slab cache
3096 Description: MMP check in ext3_remount() fails without displaying any error
3097 Details : When multiple mount protection fails during remount, proper error
3100 Severity : enhancement
3102 Description: Allow stripe size to be up to 4G-64k
3103 Details : Fix math logic to allow large stripe sizes.
3107 Description: add check for >8TB ldiskfs filesystems
3108 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3109 Don't allow >8TB ldiskfs filesystems to be mounted without
3110 force_over_8tb mount option
3115 Descriptoin: OSS complains "too many blocks" and crashes
3116 Details : In filter_commitrw_write():
3117 1. lquota_chkquota();
3118 2. obdo_from_inode();
3119 3. lquota_pending_commit();
3120 "2." has changed uid and gid in this case, then caused
3121 lquota_chkquota() and lquota_pending_commit() to use
3122 different uid and gid. That is the root of the bug.
3126 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3127 Details : Detect and print out the 'stripe count = -1' case properly.
3131 Description: Out or order replies might be lost on replay
3132 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3133 request with tid smaller than one currently being added, add it
3134 to the start, not end of the list.
3136 --------------------------------------------------------------------------------
3138 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3140 * Support for kernels:
3141 2.6.5-7.283 (SLES 9),
3142 2.6.9-55.EL (RHEL 4),
3143 2.6.16.46-0.14 (SLES 10),
3144 2.6.18.8 vanilla (kernel.org)
3145 * Client support for unpatched kernels:
3146 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3147 2.6.16 - 2.6.22 vanilla (kernel.org)
3148 * Due to recently discovered recovery problems, we do not recommend
3149 using patchless RHEL 4 clients with this or any earlier release.
3150 * Recommended e2fsprogs version: 1.39.cfs8
3151 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3152 * Starting with this release, the ldiskfs backing filesystem required
3153 by Lustre is now in its own package, lustre-ldiskfs. This package
3154 should be installed. It is versioned separately from Lustre and
3155 may be released separately in future.
3159 Description: block reactivating mgc import until all deactivates complete
3160 Details : Fix race when failing back MDT/MGS to itself (testing)
3163 Frequency : at statup only
3165 Description: mds_lov_synchronize race leads to various problems
3166 Details : simultaneous MDT->OST connections at startup can cause the
3167 sync to abort, leaving the OSC in a bad state.
3169 Severity : enhancement
3171 Description: add optional extra BUILD_VERSION info
3172 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3173 to override the lustre version.
3176 Frequency : 2.6.18 servers only
3178 Description: ll_kern_mount() doesn't release the module reference
3179 Details : The ldiskfs module reference count never drops down to 0
3180 because ll_kern_mount() doesn't release the module reference.
3185 Description: server LBUG when using old ost_num_threads parameter
3186 Details : Accept the old ost_num_threads parameter but warn that it
3187 is deprecated, and fix an off-by-one error that caused an LBUG.
3192 Description: Transient SCSI error results in persistent IO issue
3193 Details : iobuf->dr_error is not reinitialized to 0 between two
3197 Frequency : sometimes when underlying device returns I/O errors
3199 Description: OSTs not going read-only during write failures
3200 Details : OSTs are not remounted read-only when the journal commit threads
3201 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3202 instead of the ext3 wrappers.
3206 Description: poor connect scaling with increasing client count
3207 Details : Don't run filter_grant_sanity_check for more than 100 exports
3208 to improve scaling for large numbers of clients.
3211 Frequency : SLES10 only
3213 Description: sanity-quota.sh quotacheck failed: rc = -22
3214 Details : Quotas cannot be enabled on SLES10.
3217 Frequency : liblustre clients only
3219 Description: getdirentries does not give error when run on compute nodes
3220 Details : getdirentries does not fail when the size specified as an argument
3221 is too small to contain at least one entry
3223 Severity : enhancement
3225 Description: Add LNET router traceability for debug purposes
3226 Details : If a checksum failure occurs with a router as part of the
3227 IO path, the NID of the last router that forwarded the bulk data
3228 is printed so it can be identified.
3233 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3234 Details : Due to a race condition, liblustre clients were occasionally
3235 evicted incorrectly.
3237 Severity : enhancement
3239 Description: lfs setstripe use optional parameters instead of postional
3242 Severity : enhancement
3244 Description: Nanosecond timestamp support for ldiskfs
3245 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3246 resolution timestamps. There is not yet support for this at
3247 the Lustre filesystem level.
3250 Frequency : during server recovery
3252 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3253 Details : request with rq_no_resend flag not awake l_wait_event if they get
3257 Frequency : nfs export on patchless client
3259 Description: connectathon hang when test nfs export over patchless client
3260 Details : Disconnected dentry cannot be found with lookup, so we do not need
3261 to unhash it or make it invalid
3264 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3265 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3266 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3271 Description: open req refcounting wrong on reconnect
3272 Details : If reconnect happened between getting open reply from server and
3273 call to mdc_set_replay_data in ll_file_open, we will schedule
3274 replay for unreferenced request that we are about to free.
3275 Subsequent close will crash in variety of ways.
3276 Check that request is still eligible for replay in
3277 mdc_set_replay_data().
3282 Description: disable writes to filesystem when reading health_check file
3283 Details : the default for reading the health_check proc file has changed
3284 to NOT do a journal transaction and write to disk, because this
3285 can cause reads of the /proc file to hang and block HA state
3286 checking on a healthy but otherwise heavily loaded system. It
3287 is possible to return to the previous behaviour during configure
3288 with --enable-health-write.
3290 Severity : enhancement
3292 Description: 64-bit inode version
3293 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3294 to the inode. This will be required for version-based recovery.
3299 Description: MDS fails to start if a duplicate client export is detected
3300 Details : in some rare cases it was possible for a client to connect to
3301 an MDS multiple times. Upon recovery the MDS would detect this
3302 and fail during startup. Handle this more gracefully.
3304 Severity : enhancement
3306 Description: Add -o localflock option to simulate old noflock
3308 Details : This will achieve local-only flock/fcntl locks
3314 Description: log_commit_thread vs filter_destroy race leads to crash
3315 Details : Take import reference before releasing llog record semaphore
3320 Description: Wrong request locking in request set processing
3321 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3322 imp_delayed_list, in this place should be used imp_lock.
3325 Frequency : when reconnection
3327 Description: Grant Leak when osc reconnect to OST
3328 Details : When osc reconnect ost, OST(filter) should check whether it
3329 should grant more space to client by comparing fed_grant and
3330 cl_avail_grant, and return the granted space to client instead
3331 of "new granted" space, because client will call osc_init_grant
3332 to update the client grant space info.
3335 Frequency : when client reconnect to OST
3337 Description: Grant Leak when osc do resend and replay bulk write
3338 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3339 bulk write request, because the grant info while be sync between
3340 OSC and OST when reconnect, and we should ignore the grant info
3341 these of resend/replay write req.
3346 Description: Grant space more than available left space sometimes.
3347 Details : When then OST is about to be full, if two bulk writing from
3348 different clients came to OST. Accord the avaliable space of the
3349 OST, the first req should be permitted, and the second one
3350 should be denied by ENOSPC. But if the seconde arrived before
3351 the first one is committed. The OST might wrongly permit second
3352 writing, which will cause grant space > available space.
3355 Frequency : when client is evicted
3357 Description: Grant might be wrongly erased when osc is evicted by OST
3358 Details : when the import is evicted by server, it will fork another
3359 thread ptlrpc_invalidate_import_thread to invalidate the
3360 import, where the grant will be set to 0. While the original
3361 thread will update the grant it got when connecting. So if
3362 the former happened latter, the grant will be wrongly errased
3363 because of this race.
3368 Description: Checking Stale with correct fid
3369 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3370 to check whether it is stale, because sometimes, we want the
3371 enqueue happened anyway, and op_data.fid2 will not be initialized.
3373 Severity : enhancement
3375 Description: update patchless client
3376 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3379 Frequency : only with 2.4 kernel
3381 Description: random memory corruption
3382 Details : size of struct ll_inode_info is to big for union inode.u and this
3383 can be cause of random memory corruption.
3388 Description: Memory leak in recovery
3389 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3390 It should also check obd_fail before fsfilt_start, otherwise if
3391 fsfilt_start return -EROFS,(failover mds during mds recovery).
3392 then the req will return with repmsg->transno = 0 and rc = EROFS.
3393 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3394 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3395 be freed no matter whether fsfilt_commit success or not.
3398 Frequency : only with huge count clients
3400 Description: Prevents from taking the superblock lock in llap_from_page for
3402 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3403 for llap_from_page call in ll_removepage prevents from taking the
3404 superblock lock for a soon died page.
3409 Description: Not check open intent error before release open handle
3410 Details : in some rare cases, the open intent error is not checked before
3411 release open handle, which may cause
3412 ASSERTION(open_req->rq_transno != 0), because it tries to release
3413 the failed open handle.
3418 Description: Set cat log bitmap only after create log success.
3419 Details : in some rare cases, the cat log bitmap is set too early. and it
3420 should be set only after create log success.
3424 Description: Accessing a block bevice can re-enable I/O when Lustre is
3425 tearing down a device.
3426 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3431 Description: service threads may hog cpus when there are a lot of requests
3432 Details : Insert cond_resched to give other threads a chance to use some CPU
3437 Description: the cat log was not initialized in recovery
3438 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3439 unlink log on mds will not be initialized until mds post
3440 recovery. And also in mds post recovery, the unlink log will
3441 initialization will be done asynchronausly, so there will be race
3442 between add unlink log and unlink log initialization.
3446 Description: brw_stats were being printed incorrectly
3447 Details : brw_stats were being printed as log2 but all of them were not
3448 recorded as log2. Also remove some code duplication arising from
3449 filter_tally_{read,write}.
3453 Frequency : rare, only in recovery.
3454 Description: ASSERTION(req->rq_type != LI_POISON) failed
3455 Details : imp_lock should be held while iterating over imp_sending_list for
3456 prevent destroy request after get timeout in ptlrpc_queue_wait.
3460 Description: replay-single.sh test 52 fails
3461 Details : A lock's skiplist need to be cleanup when it being unlinked
3462 from its resource list.
3466 Description: Short directio read returns full requested size rather than
3468 Details : Direct I/O operations should return actual amount of bytes
3469 transferred rather than requested size.
3471 Severity : enhancement
3473 Description: metadata RPC reduction (e.g. for rm performance)
3474 Details : decrease the amount of synchronous RPC between clients and servers
3475 by canceling conflicing lock before the operation on the client
3476 and packing thier handles into the main operation RPC to server.
3478 Severity : enhancement
3480 Description: Async OSC create to avoid the blocking unnecessarily.
3481 Details : If an OST has no remaining object, system will block on the
3482 creation when it needs to create a new object on this OST. Now,
3483 ways use pre-created objects when available, instead of blocking on
3484 an empty osc while others are not empty. If we must block, we
3485 block for the shortest possible period of time.
3489 Description: improve handling recoverable errors
3490 Details : if request processig with error which can be recoverable on server
3491 request should be resend, otherwise page released from cache and
3494 Severity : enhancement
3496 Description: refine locking for avoid write wrong info into lov_objid file
3497 Details : fix possible races with add new target and write/update data in
3500 --------------------------------------------------------------------------------
3502 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3507 Frequency : on some architectures
3509 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3510 Details : Uninitialized flags sometimes cause configuration commands to
3514 Frequency : patchless clients only
3516 Description: missing __iget() symbol export
3517 Details : The __iget() symbol export is missing. To avoid the need for
3518 this on patchless clients the deathrow inode reaper is turned
3519 off, and we depend on the VM to clean up old inodes. This
3520 dependency was during via the fix for bug 12181.
3522 --------------------------------------------------------------------------------
3524 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3526 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3527 INTEROPERATE with older versions automatically. In many cases a
3528 special upgrade step is needed. Please read the
3529 user documentation before upgrading any part of a live system.
3530 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3531 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3532 * WARNING: Lustre configuration and startup changes are required with
3533 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3537 Severity : enhancement
3539 Description: MountConf
3540 Details : Lustre configuration is now managed via mkfs and mount
3541 commands instead of lmc and lconf. New obd types (MGS, MGC)
3542 are added for dynamic configuration management. See
3543 https://mail.clusterfs.com/wikis/lustre/MountConf for
3546 Severity : enhancement
3548 Description: dynamic OST addition
3549 Details : OSTs can now be added to a live filesystem
3551 Severity : enhancement
3553 Description: startup order invariance
3554 Details : MDTs and OSTs can be started in any order. Clients only
3555 require the MDT to complete startup.
3557 Severity : enhancement
3559 Description: parallel, asynchronous orphan cleanup
3560 Details : orphan cleanup is now performed in separate threads for each
3561 OST, allowing parallel non-blocking operation.
3563 Severity : enhancement
3565 Description: optimized stripe assignment
3566 Details : stripe assignments are now made based on ost space available,
3567 ost previous usage, and OSS previous usage, in order to try
3568 to optimize storage space and networking resources.
3570 Severity : enhancement
3572 Description: Permanently set tunables
3573 Details : All writable /proc/fs/lustre tunables can now be permanently
3574 set on a per-server basis, at mkfs time or on a live system.
3576 Severity : enhancement
3578 Description: Lustre message v2
3579 Details : Add lustre message format v2.
3581 Severity : enhancement
3583 Description: client OST exclusion list
3584 Details : Clients can be started with a list of OSTs that should be
3585 declared "inactive" for known non-responsive OSTs.
3589 Description: SPEC SFS validation failure on NFS v2 over lustre.
3590 Details : Changes the blocksize for regular files to be 2x RPC size,
3591 and not depend on stripe size.
3593 Severity : enhancement
3595 Description: Multiple MD RPCs in flight.
3596 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3597 To avoid overly-overloading MDS, introduce a limit on number of
3598 MDS RPCs in flight for a single client and add /proc controls
3599 to adjust this limit.
3601 Severity : enhancement
3603 Description: client read/write statistics
3604 Details : Add client read/write call usage stats for performance
3605 analysis of user processes.
3606 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3607 file access. extents_stats shows chunk size distribution.
3608 extents_stats_per_process show chunk size distribution per
3611 Severity : enhancement
3613 Description: mds statistics
3614 Details : Add detailed mds operations statistics in
3615 /proc/fs/lustre/mds/*/stats.
3619 Description: Failure of copying files with lustre special EAs.
3620 Details : Client side always return success for setxattr call for lustre
3621 special xattr (currently only "trusted.lov").
3626 Description: Refcount LNET uuids
3627 Details : The global LNET uuid list grew linearly with every startup;
3628 refcount repeated list entries instead of always adding to
3631 Severity : enhancement
3633 Description: Dynamic service threads
3634 Details : Within a small range, start extra service threads
3635 automatically when the request queue builds up.
3638 Frequency : mixed-endian client/server environments
3640 Description: mixed-endian crashes
3641 Details : The new msg_v2 system had some failures in mixed-endian
3644 Severity : enhancement
3646 Description: Easy OST removal
3647 Details : OSTs can be permanently deactivated with e.g. 'lctl
3648 conf_param lustre-OST0001.osc.active=0'
3650 Severity : enhancement
3652 Description: MGS proc entries
3653 Details : Added basic proc entries for the MGS showing what filesystems
3656 Severity : enhancement
3658 Description: provide MGS failover
3659 Details : Added config lock reacquisition after MGS server failover.
3661 Severity : enhancement
3663 Description: add Linux 2.4 support
3664 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3668 Description: a large application tries to do I/O to the same resource and dies
3669 in the middle of it.
3670 Details : Check the req->rq_arrival time after the call to
3671 ost_brw_lock_get(), but before we do anything about
3672 processing it & sending the BULK transfer request. This
3673 should help move old stale pending locks off the queue as
3674 quickly as obd_timeout.
3677 Frequency : when an incorrect nid is specified during startup
3679 Description: ptlrpc connect to non-existant node causes kernel crash
3680 Details : LNET can't be re-entered from an event callback, which
3681 happened when we expire a message after the export has been
3682 cleaned up. Instead, hand the zombie cleanup off to another
3685 Severity : enhancement
3687 Description: plain/inodebits lock performance improvement
3688 Details : Grouping plain/inodebits in granted list by their request modes
3689 and bits policy, thus improving the performance of search through
3693 Frequency : only if OST filesystem is corrupted
3695 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3696 Details : for a short time RPCs with bulk IO are in the replay list,
3697 but replay of bulk IOs is unimplemented. If the OST filesystem
3698 is corrupted due to disk cache incoherency and then replay is
3699 started it is possible to trip an assertion. Avoid putting
3700 committed RPCs into the replay list at all to avoid this issue.
3703 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3705 Description: System hang on startup
3706 Details : This bug allowed the liblustre (e.g. catamount) client to
3707 return to the app before handling all startup RPCs. This
3708 could leave the node unresponsive to lustre network traffic
3709 and manifested as a server ptllnd timeout.
3711 Severity : enhancement
3713 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3714 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3715 boolean to control whether to print peer debug info when a
3716 client's RPC times out.
3719 Frequency : only for kernels with patches from Lustre below 1.4.3
3721 Description: Remove old rdonly API
3722 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3725 Frequency : only for devices with external journals
3727 Description: Set external device read-only also
3728 Details : During a commanded failover stop, we set the disk device
3729 read-only while the server shuts down. We now also set any
3730 external journal device read-only at the same time.
3733 Frequency : when upgrading from 1.4 while trying to change parameters
3735 Description: The wrong (new) MDC name was used when setting parameters for
3736 upgraded MDT's. Also allows changing of OSC (and MDC)
3737 parameters if --writeconf is specified at tunefs upgrade time.
3740 Frequency : when setting specific ost indicies
3742 Description: QOS code breaks on skipped indicies
3743 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3744 created with --index need not be sequential.
3746 Severity : enhancement
3748 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3749 Details : The uninit_groups feature works in conjunction with the kernel
3750 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3751 up the pass1 processing of e2fsck. This is a read-only feature
3752 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3753 mount filesystems that have had this feature enabled.
3755 Severity : enhancement
3757 Description: Improve multi-block allocation algorithm to avoid fragmentation
3758 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3759 allocation locality and avoid filesystem fragmentation.
3761 ------------------------------------------------------------------------------
3763 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3765 * Support for kernels:
3766 2.4.21-47.0.1.EL (RHEL 3)
3767 2.6.5-7.283 (SLES 9)
3768 2.6.9-42.0.10.EL (RHEL 4)
3769 2.6.12.6 vanilla (kernel.org)
3770 2.6.16.27-0.9 (SLES 10)
3771 * Recommended e2fsprogs version: 1.39.cfs5
3773 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3777 Frequency : occasional, depends on client load and configuration
3778 Bugzilla : 12181, 12203
3779 Description: data loss for recently-modified files
3781 Details : In some cases it is possible that recently written or created
3782 files may not be written to disk in a timely manner (this should
3783 normally be within 30s unless client IO load is very high).
3784 The problem appears as zero-length files or files that are a
3785 multiple of 1MB in size after a client crash or client eviction
3786 that are missing data at the end of the file.
3788 This problem is more likely to be hit on clients where files are
3789 repeatedly created and unlinked in the same directory, clients
3790 have a large amount of RAM, have many CPUs, the filesystem has
3791 many OSTs, the clients are rebooted frequently, and/or the files
3792 are not accessed by other nodes after being written.
3794 The presence of the problem can be detected by looking at
3795 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3796 smaller than the second (nr_unused) then dirty files will not
3797 be flushed automatically to disk. "sync; sleep 10" should be
3798 run several times on the node before unmounting it to update
3799 Lustre (this is also safe to run on nodes without this problem).
3801 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3802 that can cause this same problem, so customers using that kernel
3803 also need to update the kernel in addition to Lustre. In order
3804 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3806 It is normal that files written just before a client crash (less
3807 than 30s) may not yet have been flushed to disk, even for local
3811 Frequency : frequent on thin XT3 nodes
3813 Description: UUID collision on thin XT3 Linux nodes
3814 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3815 randomly, since we relied on an insufficiently-seeded PRNG.
3820 Description: OSS hangs after "All ost request buffers busy"
3821 Details : A deadlock between quota and journal operations caused OSS
3822 hangs after printing "All ost request buffers busy."
3825 Frequency : always on liblustre builds
3827 Description: Cleanup compiler warnings on liblustre
3830 Frequency : always on liblustre builds on XT3
3832 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3836 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3837 > 32000 subdirectories
3838 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3839 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3840 subdirectory count crosses 32000. This will aid e2fsck to
3841 correctly handle more than 32000 subdirectories.
3844 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3846 Description: System hang on startup
3847 Details : This bug allowed the liblustre (e.g. catamount) client to
3848 return to the app before handling all startup RPCs. This
3849 could leave the node unresponsive to lustre network traffic
3850 and manifested as a server ptllnd timeout.
3852 Severity : enhancement
3854 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3855 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3856 boolean to control whether to print peer debug info when a
3857 client's RPC times out.
3862 Description: make O_SYNC working on 2.6 kernels
3863 Details : 2.6 kernels use different method for mark pages for write,
3864 so need add a code to lustre for O_SYNC work.
3869 Description: Failure to close file and release space on NFS
3870 Details : Put inode details into lock acquired in ll_intent_file_open.
3871 Use mdc_intent_lock in ll_intent_open to properly
3872 detect all kind of errors unhandled by mdc_enqueue.
3877 Description: proc file read during shutdown sometimes raced obd removal,
3879 Details : Add lock to prevent obd access after proc file removal.
3882 Frequency : Only for files larger than 4GB on 32-bit clients.
3884 Description: improperly doing page alignment of locks
3885 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3886 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3891 Description: RPCs being resent when they shouldn't be
3892 Details : Some RPCs that should not be resent are being resent. This
3893 can cause inconsistencies in the RPC state machine. Do not
3894 resend such requests.
3897 Frequency : rare, only with NFS export
3899 Description: Crash on NFS re-export node
3900 Details : under very unusual load conditions an assertion is hit in
3901 ll_intent_file_open()
3904 Frequency : only if OST filesystem is corrupted
3906 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3907 Details : for a short time RPCs with bulk IO are in the replay list,
3908 but replay of bulk IOs is unimplemented. If the OST filesystem
3909 is corrupted due to disk cache incoherency and then replay is
3910 started it is possible to trip an assertion. Avoid putting
3911 committed RPCs into the replay list at all to avoid this issue.
3916 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3917 Details : Large single O_DIRECT read and write calls can fail to allocate
3918 a sufficiently large buffer to process the request. In case of
3919 allocation failure the allocation is retried with a smaller
3920 buffer and broken into smaller requests.
3922 Severity : enhancement
3924 Description: Add -o localflock option to simulate old noflock behaviour.
3925 Details : This will achieve local-only flock/fcntl locks coherentness.
3930 Description: versioning check is incomplete
3931 Details : Checking the version difference of client vs. server, report
3932 error if the gap is too big.
3937 Description: add support PG_writeback bit
3938 Details : add support for PG_writeback bit for Lustre, for more carefull
3939 work with page cache in 2.6 kernel. This also fix some deadlocks
3940 and remove hack for work O_SYNC with 2.6 kernel.
3942 Severity : enhancement
3944 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3945 Details : The uninit_groups feature works in conjunction with the kernel
3946 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3947 up the pass1 processing of e2fsck. This is a read-only feature
3948 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3949 mount filesystems that have had this feature enabled.
3951 Severity : enhancement
3953 Description: Improve multi-block allocation algorithm to avoid fragmentation
3954 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3955 allocation locality and avoid filesystem fragmentation.
3957 ------------------------------------------------------------------------------
3959 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3961 * Support for kernels:
3962 2.6.9-42.0.3.EL (RHEL 4)
3963 2.6.5-7.276 (SLES 9)
3964 2.4.21-47.0.1.EL (RHEL 3)
3965 2.6.12.6 vanilla (kernel.org)
3966 2.6.16.21-0.8 (SLES10)
3967 * Recommended e2fsprogs version: 1.39.cfs2-0
3969 * The backwards-compatible /proc/sys/portals symlink has been removed
3970 in this release. Before upgrading, please ensure that you change
3971 any configuration scripts or /etc/sysctl.conf files that access
3972 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3973 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3974 in advance of the upgrade on any system running Lustre 1.4.6 or
3975 newer, since /proc/sys/lnet was added in that version.
3976 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3980 Frequency : only when quota is used
3982 Description: avoid scanning export list for quota master
3983 Details : Change the algorithms to avoid scanning export list in order
3984 to improve the efficiency.
3987 Frequency : MDS failover only, very rarely
3989 Description: "went back in time" messages on mds failover
3990 Details : The greatest transno may be lost when the current operation
3991 finishes with an error (transno==0) and the client's last_rcvd
3992 record is over-written. Save the greatest transno in the
3993 mds_last_transno for this case.
3996 Frequency : always for specific kernels and striping counts
3998 Description: client may get "Matching packet too big" without ACL support
3999 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4000 when trying to access files in certain configurations. The
4001 clients should in fact be denied when mounting because they do
4002 not understand ACLs.
4005 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4007 Description: many clients connecting with IO in progress causes connect timeouts
4008 Details : Avoid synchronous journal commits to avoid delays caused by many
4009 clients connecting/disconnecting when bulk IO is in progress.
4010 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4011 OST_IO_PORTAL to avoid delays behind potentially many pending
4015 Frequency : occasionally with multiple writers to a single file
4017 Description: shared writes to file may result in wrong size reported by stat()
4018 Details : Allow growing of kms when extent lock is cancelled
4021 Frequency : always with random mmap IO to multi-striped file
4023 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4024 Details : Only if the hole is at the end of OST object so that kms is too
4025 small. Fix is to increase kms accordingly in ll_nopage.
4028 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4030 Description: writes to a missing object would leak memory on the OST
4031 Details : If there is an inconsistency between the MDS and OST filesystems,
4032 such that the MDS references an object that doesn't exist, writes
4033 to that object will leak memory due to incorrect cleanup in the
4034 error handling path, eventually running out of memory on the OST.
4039 Description: Creating too long symlink causes lustre errors
4040 Details : Check symlink and name lengths before sending requests to MDS.
4043 Frequency : only if flock is enabled (not on by default)
4045 Description: posix locks not released on fd closure on 2.6.9+
4046 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4047 kernels, this caused such locks not to be released on fd close and
4048 then assertions on fs unmount about still used locks.
4051 Frequency : MDS failover only, very rarely
4053 Description: clients may get ASSERTION(granted_lock != NULL)
4054 Details : When request was taking a long time, and a client was resending
4055 a getattr by name lock request. The were multiple lock requests
4056 with the same client lock handle and
4057 mds_getattr_name->fixup_handle_for_resent_request found one of the
4058 lock handles but later failed with ASSERTION(granted_lock != NULL).
4063 Description: handle->h_buffer_credits > 0, assertion failure
4064 Details : h_buffer_credits is zero after truncate, causing assertion
4065 failure. This patch extends the transaction or creates a new
4069 Frequency : NFS re-export or patchless client
4070 Bugzilla : 11179, 10796
4071 Description: Crash on NFS re-export node (__d_move)
4072 Details : We do not want to hash the dentry if we don't have a lock.
4073 But if this dentry is later used in d_move, we'd hit uninitialised
4074 list head d_hash, so we just do this to init d_hash field but
4075 leave dentry unhashed.
4078 Frequency : NFS re-export or patchless client
4080 Description: NFS exports has problem with symbolic link
4081 Details : lustre client didn't properly install dentry when re-exported
4082 to NFS or running patchless client.
4085 Frequency : NFS re-export or patchless client
4087 Description: Various nfs/patchless fixes.
4088 Details : fixes reuse disconected alias for lookup process - this fixes
4089 warning "find_exported_dentry: npd != pd",
4090 fix permission error with open files at nfs.
4091 fix apply umask when do revalidate.
4094 Frequency : occasional
4096 Description: Crash on NFS re-export node
4097 Details : calling clear_page() on the wrong pointer triggered oops in
4098 generic_mapping_read().
4101 Frequency : rarely, using O_DIRECT IO
4103 Description: unaligned directio crashes client with LASSERT
4104 Details : check for unaligned buffers before trying any requests.
4107 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4109 Description: stale data returned from RAID cache
4110 Details : If only a small amount of IO is done to the RAID device before
4111 reading it again it is possible to get stale data from the RAID
4112 cache instead of reading it from disk.
4115 Frequency : always for sles10 kernel
4117 Description: sles10 support
4118 Details : ll_follow_link: compile fixes and using of nd_set_link
4119 under newer kernels.
4122 Frequency : depends on arch, kernel and compiler version, always on sles10
4125 Description: recursive or deep enough symlinks cause stack overflow
4126 Details : getting rid of large stack-allocated variable in
4130 Frequency : depends on hardware
4132 Description: lustre write performance loss in the SLES10 kernel
4133 Details : the performance loss is caused by using of write barriers in the
4134 ext3 code. The SLES10 kernel turns barrier support on by
4135 default. The fix is to undo that change for ldiskfs.
4137 ------------------------------------------------------------------------------
4139 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4141 * Support for kernels:
4142 2.6.9-42.0.3EL (RHEL 4)
4143 2.6.5-7.276 (SLES 9)
4144 2.4.21-47.0.1.EL (RHEL 3)
4145 2.6.12.6 vanilla (kernel.org)
4149 Frequency : quota enabled and large files being deleted
4151 Description: releasing more than 4GB of quota at once hangs OST
4152 Details : If a user deletes more than 4GB of files on a single OST it
4153 will cause the OST to spin in an infinite loop. Release
4154 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4159 Description: statfs data retrieved from /proc may be stale or zero
4160 Details : When reading per-device statfs data from /proc, in the
4161 {kbytes,files}_{total,free,avail} files, it may appear
4162 as zero or be out of date.
4165 Frequency : systems with MD RAID1 external journal devices
4167 Description: lconf's call to blkid is confused by RAID1 journal devices
4168 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4169 of returning all block devices that match the journal UUID.
4172 Frequency : always, for aggregate stripe size over 4GB
4174 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4175 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4176 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4179 Frequency : always if "lfs find" used on a local file/directory
4181 Description: "lfs find" segfaults if used on a local file/directory
4182 Details : The case where a directory component was not specified wasn't
4183 handled correctly. Handle this properly.
4186 Frequency : always on ppc64
4188 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4189 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4190 perform endianness conversion.
4193 Frequency : rarely (truncate to non-zero file size after write under load)
4194 Bugzilla : 10730, 10687
4195 Description: Files padded with zeros to next 4K multiple
4196 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4197 it is possible that files that are truncated to a non-zero size
4198 immediately after being written are filled with zero bytes beyond
4199 the truncated size. No file data is lost.
4201 Severity : enhancement
4203 Description: Allow recovery/failover for liblustre clients.
4204 Details : liblustre clients were unaware of failover configurations until
4207 Severity : enhancement
4209 Description: user file locks should fail when not mounting with flock option
4210 Details : Set up an error-returning stub in ll_file_operations.lock field
4211 to prevent incorrect behaviour when client is mounted without
4212 flock option. Also, set up properly f_op->flock field for
4216 Frequency : always on ia64
4218 Description: "lfs df" loops on printing out MDS statfs information
4219 Details : The obd_ioctl_data was not initialized and in some systems
4220 this caused a failure during the ioctl that did not return
4221 an error. Initialize the struct and return an error on failure.
4224 Frequency : SLES 9 only
4226 Description: Error of copying files with lustre special EAs as root
4227 Details : Client side always return success for setxattr call for lustre
4228 special xattr (currently only "trusted.lov").
4231 Frequency : rarely on clusters with both ia64+i386 clients
4233 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4234 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4235 clients are doing concurrent writes to the same file it is
4236 possible that the ia64 clients may LASSERT because the OST
4237 extent locks are not PAGE_SIZE aligned. Ensure that grown
4238 locks are always aligned on the request boundary.
4241 Frequency : specific use, occasional
4243 Description: Overwriting in use executable truncates on-disk binary image
4244 Details : If one node attempts to overwrite an executable in use by
4245 another node, we now correctly return ETXTBSY instead of
4246 truncating the file.
4248 Severity : enhancement
4250 Description: Async OSC create to avoid the blocking unnecessarily.
4251 Details : If an OST has no remaining object, system will block on the creating
4252 when need to create a new object on this OST. Now, ways use
4253 pre-created objects when available, instead of blocking on an
4254 empty osc while others are not empty. If we must block, we block
4255 for the shortest possible period of time.
4260 Description: chmod on Lustre root is propagated to other clients
4261 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4262 invalid by the follow_mount time.
4265 Frequency : liblustre clients only
4267 Description: Race in 'instant cancel' lock handling could lead to such locks
4268 never to be granted in case of SMP MDS
4269 Details : Do not destroy not yet granted but cbpending locks in
4273 Frequency : replay/resend of open
4275 Description: non null lock assetion failure in mds_intent_policy
4276 Details : Trying to replay/resend lockless open requests resulted in
4277 mds_open() returning 0 with no lock. Now it sets a flag if
4278 a lock is going to be returned.
4280 Severity : enhancement
4282 Description: Checksum enhancements
4283 Details : New checksum enhancements allow for resending RPCs that failed
4286 Severity : enhancement
4288 Description: Tunables on number of dirty pages in cacche
4289 Details : Allow to set limit on number of dirty pages cached.
4294 Description: client crash on unmount - lock still has references
4295 Details : In some error handling cases it was possible to leak a lock
4296 reference on a client while accessing a file. This was not
4297 harmful to the client during operation, but would cause the
4298 client to crash when the filesystem is unmounted.
4301 Frequency : specific case, rare
4303 Description: ETXTBSY on mds though file not in use by client
4304 Details : ETXTBSY is no longer incorrectly returned when attempting to
4305 chmod or chown a directory that the user previously tried to
4306 execute or a currently-executing binary.
4309 Frequency : extremely rare except on liblustre-based clients
4311 Description: Lustre space not freed when files are deleted
4312 Details : Clean up open-unlinked files after client eviction. Previously
4313 the unlink was skipped and the files remained as orphans.
4318 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4319 Details : In some cases it was possible to send a blocking callback to
4320 a client doing a glimpse, even though that client didn't get
4321 a lock granted. When the glimpse lock is cancelled on the OST
4322 the freed lock is left on the waiting list and corrupted the list.
4325 Frequency : all core dumps
4327 Description: Broke core dumps to lustre
4328 Details : Negative dentry may be unhashed if parent does not have UPDATE
4329 lock, but some callers, e.g. do_coredump, expect dentry to be
4330 hashed after successful create, hash it in ll_create_it.
4332 ------------------------------------------------------------------------------
4334 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4336 * Support for kernels:
4337 2.6.9-42.0.2.EL (RHEL 4)
4338 2.6.5-7.276 (SLES 9)
4339 2.4.21-40.EL (RHEL 3)
4340 2.6.12.6 vanilla (kernel.org)
4344 Frequency : always on RHEL 3
4346 Description: Number of open files grows over time
4347 Details : The number of open files grows over time, whether or not
4348 Lustre is started. This was due to a filp leak introduced
4349 by one of our kernel patches.
4351 ------------------------------------------------------------------------------
4353 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4355 * Support for kernels:
4356 2.6.9-42.EL (RHEL 4)
4357 2.6.5-7.276 (SLES 9)
4358 2.4.21-40.EL (RHEL 3)
4359 2.6.12.6 vanilla (kernel.org)
4364 Bugzilla : 5719, 9635, 9792, 9684
4365 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4366 Details : If a server is under heavy load and cannot reply to new
4367 connection requests before the client resends the (re)connect,
4368 the connection handling code can behave badly if two service
4369 threads are concurrently handing separate (re)connections from
4370 the same client. Add better locking to the connection handling
4371 code, and ensure that only a single connection will be processed
4372 for a given client UUID, even if the lock is dropped.
4374 Severity : enhancement
4376 Description: add TCP zero-copy support to kernel
4377 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4378 sends if the hardware supports scatter-gather and checksumming.
4379 This allows socklnd to do client-write and server-read more
4380 efficiently and reduce CPU utilization from skbuf copying.
4383 Frequency : only if NFS exporting from client
4385 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4386 Details : If setting attributes on a file created under NFS that had
4387 never been opened it would be possible to oops the client
4388 if the file had no objects.
4391 Frequency : always for liblustre
4393 Description: liblustre client does MDS+OSTs setattr RPC for each write
4394 Details : When doing a write from a liblustre client, the client
4395 incorrectly issued an RPC to the MDS and each OST the file was
4396 striped over in order to update the timestamps. When writing
4397 with small chunks and many clients this could overwhelm the MDS
4398 with RPCs. In all cases it would slow down the write because
4399 these RPCs are unnecessary.
4401 Severity : enhancement
4403 Description: allow number of MDS service threads to be changed at module load
4404 Details : It is now possible to change the number of MDS service threads
4405 running. Adding "options mds mds_num_threads={N}" to the MDS's
4406 /etc/modprobe.conf will set the number of threads for the next
4407 time Lustre is restarted (assuming the "mds" module is also
4408 reloaded at that time). The default number of threads will
4409 stay the same, 32 for most systems.
4414 Description: OST crash if filesystem is unformatted or corrupt
4415 Details : If an OST is started on a device that has never been formatted
4416 or if the filesystem is corrupt and cannot even mount then the
4417 error handling cleanup routines would dereference a NULL pointer.
4422 Description: NULL pointer deref in llap_from_page.
4423 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4424 incorrect) mapping if the page was truncated/reclaimed while it was
4425 searched for. Check for this condition and skip such pages when
4426 doing readahead. Introduce extra check to llap_from_page() to
4427 verify page->mapping->host is non-NULL (so page is not anonymous).
4430 Frequency : Sometimes when using sys_sendfile
4432 Description: "page not covered by a lock" warnings from ll_readpage
4433 Details : sendfile called ll_readpage without right page locks present.
4434 Now we introduced ll_file_sendfile that does necessary locking
4435 around call to generic_file_sendfile() much like we do in
4439 Frequency : with certain MDS communication failures at client mount time
4441 Description: NULL pointer deref after failed client mount
4442 Details : a client connection request may delayed by the network layer
4443 and not be sent until after the PTLRPC layer has timed out the
4444 request. If the client fails the mount immediately it will try
4445 to clean up before the network times out the request. Add a
4446 reference from the request import to the obd device and delay
4447 the cleanup until the network drops the request.
4450 Frequency : occasionally during client (re)connect
4452 Description: assertion failure during client (re)connect
4453 Details : processing a client connection request may be delayed by the
4454 client or server longer than the client connect timeout. This
4455 causes the client to resend the connection request. If the
4456 original connection request is replied in this interval, the
4457 client may trip an assertion failure in ptlrpc_connect_interpret()
4458 which thought it would be the only running connect process.
4461 Frequency : only with obd_echo servers and clients that are rebooted
4463 Description: kernel BUG accessing uninitialized data structure
4464 Details : When running an obd_echo server it did not start the ping_evictor
4465 thread, and when a client was evicted an uninitialized data
4466 structure was accessed. Start the ping_evictor in the RPC
4467 service startup instead of the OBD startup.
4469 Severity : enhancement
4470 Bugzilla : 10193 (patchless)
4471 Description: Remove dependency on various unexported kernel interfaces.
4472 Details : No longer need reparent_to_init, exit_mm, exit_files,
4473 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4476 Frequency : rare (only users of deprecated and unsupported LDAP config)
4478 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4479 Details : LDAP apparently contains 'lustreName' attributes instead of
4480 'name'. A simple remapping of the name is sufficient.
4483 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4485 Description: waiting_locks_callback trips kernel BUG if client is evicted
4486 Details : Running with the dump_on_timeout debug flag turned on makes
4487 it possible that the waiting_locks_callback() can try to dump
4488 the Lustre kernel debug logs from an interrupt handler. Defer
4489 this log dumping to the expired_lock_main() thread.
4491 Severity : enhancement
4493 Description: Support NFS exporting on 2.6 kernels.
4494 Details : Implement non-rawops metadata methods for NFS server to use without
4495 changing NFS server code.
4498 Frequency : very rare (synthetic metadata workload only)
4500 Description: two racing renames might cause an MDS thread to deadlock
4501 Details : Running the "racer" program may cause one MDS thread to rename
4502 a file from being the source of a rename to being the target of
4503 a rename at exactly the same time that another thread is doing
4504 so, and the second thread has already enqueued these locks after
4505 doing a lookup of the target and is trying to relock them in
4506 order. Ensure that we don't try to re-lock the same resource.
4509 Frequency : only very large systems with liblustre clients
4511 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4512 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4513 all OSTs in parallel. This allows the eviction of stale liblustre
4514 clients to proceed much faster than if they were done in series,
4515 and also offers similar improvements for other set_info RPCs.
4520 Description: excessive CPU usage during initial read phase on client
4521 Details : During the initial read phase on a client, it would agressively
4522 retry readahead on the file, consuming too much CPU and impacting
4523 performance (since 1.4.5.8). Improve the readahead algorithm
4524 to avoid this, and also improve some other common cases (read
4525 of small files in particular, where "small" is files smaller than
4526 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4531 Description: MDS crash when receiving packet with unknown intent.
4532 Details : Do not LBUG in unknown intent case, just return -EFAULT
4534 Severity : enhancement
4535 Bugzilla : 9293, 9385
4536 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4537 Details : Do not serialize getattr (non-intent version) and statfs.
4540 Frequency : occasional, when OST network is overloaded/intermittent
4542 Description: client evicted by OST after bulk IO timeout
4543 Details : If a client sends a bulk IO request (read or write) the OST
4544 may evict the client if it is unresposive to its data GET/PUT
4545 request. This is incorrect if the network is overloaded (takes
4546 too long to transfer the RPC data) or dropped the OST GET/PUT
4547 request. There is no need to evict the client at all, since
4548 the pinger and/or lock callbacks will handle this, and the
4549 client can restart the bulk request.
4552 Frequency : Always when mmapping file with no objects
4554 Description: client crashes when mmapping file with no objects
4555 Details : Check that we actually have objects in a file before doing any
4556 operations on objects in ll_vm_open, ll_vm_close and
4562 Description: Request leak when working with deleted CWD
4563 Details : Introduce advanced request refcount tracking for requests
4564 referenced from lustre intent.
4566 Severity : Enhancement
4568 Description: Cache open file handles on client.
4569 Details : MDS now will return special lock along with openhandle, if
4570 requested and client is allowed to hold openhandle, even if unused,
4571 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4572 closing files for every read/write openration.
4574 Severity : Enhancement
4576 Description: Cache open negative dentries on client when possible.
4577 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4578 negative dentries on lock revocation.
4583 Description: Remounting a client read-only wasn't possible with a zconf mount
4584 Details : It wasn't possible to remount a client read-only with llmount.
4586 Severity : enhancement
4587 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4588 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4589 MPICH is included to add Lustre-specific ADIO interfaces.
4590 This is based closely on the UFS ADIO layer and only differs
4591 in file creation, in order to allow the OST striping to be set.
4592 This is user-contributed code and not supported by CFS.
4597 Description: extended inode attributes (immutable, append-only) work improperly
4598 when 2.4 and 2.6 kernels are used on client/server or vice versa
4599 Details : Introduce kernel-independent values for these flags.
4601 Severity : enhancement
4604 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4605 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4606 granularity. Now, Fractional MB granularity is be supported,
4607 this is very useful for low memory system.
4609 Severity : enhancement
4611 Description: Getattr by fid
4612 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4613 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4614 have locks and avoids getattr rpc requests for it on every stat.
4617 Frequency : Always, for filesystems larger than 2TB
4619 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4620 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4621 filesystems larger than 2TB to be mounted with mballoc enabled.
4624 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4626 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4627 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4628 filesystems larger than 2TB is mounted then the kernel will
4629 silently corrupt the start of the filesystem. CONFIG_LBD is
4630 enabled for all CFS-supported kernels, but the possibility of
4631 this happening with a modified kernel config exists.
4633 Severity : enhancement
4635 Description: add client O_DIRECT support for 2.6 kernels
4636 Details : It is now possible to do O_DIRECT reads and writes to files
4637 in the Lustre client mountpoint on 2.6 kernel clients.
4639 Severity : enhancement
4641 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4642 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4643 parallel, not waiting for response from every OST before sending
4644 a rpc to the next OST.
4649 Description: setattr vs write race when updating file timestamps
4650 Details : Client processes that update a file timestamp into the past
4651 right after writing to the file (e.g. tar) it is possible that
4652 the updated file modification time can be reset to the current
4653 time due to a race between processing the setattr and write RPC.
4655 Severity : enhancement
4657 Description: Bring 'lfs find' closer in line with regular Linux find.
4658 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4659 -print0 options and obtains all the needed info through the lustre
4662 Severity : enhancement
4664 Description: support up to 1024 configured devices on one node
4665 Details : change obd_dev array from statically allocated to dynamically
4666 allocated structs as they are first used to reduce memory usage
4671 Description: Flush dirty partially truncated pages during truncate
4672 Details : Immediatelly flush partially truncated pages in filter_setattr,
4673 this way we completely avoid having any pages in page cache on OST
4674 and can retire ugly workarounds during writes to flush such pages.
4679 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4680 Details : protect inode from truncation within vfs_unlink() context
4681 just take a reference before calling vfs_unlink() and release it
4682 when parent's i_sem is free.
4687 Description: last_id value checked outside lock on OST caused LASSERT failure
4688 Details : If there were multiple MDS->OST object precreate requests in
4689 flight, it was possible that the OST's last object id was checked
4690 outside a lock and incorrectly tripped an assertion. Move checks
4691 inside locks, and discard old precreate requests.
4694 Frequency : always, if extents are used on OSTs
4696 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4697 Details : OSTs using the extents format would not zero the high 16 bits of
4698 the index physical block number. This is not a problem for any
4699 OST filesystems smaller than 16TB, and no kernels support ext3
4700 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4701 new/modified files) and can be fixed for existing filesystems
4702 with e2fsprogs-1.39-cfs1.
4707 Description: import connection selection may be incorrect if timer wraps
4708 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4709 import connections to be ignored if the 32-bit jiffies counter
4710 wraps. Use a 64-bit jiffies counter.
4713 Frequency : very large clusters immediately after boot
4715 Description: LNET request buffers exhausted under heavy short-term load
4716 Details : If a large number of client requests are generated on a service
4717 that has previously never seen so many requests it is possible
4718 that the request buffer growth cannot keep up with the spike in
4719 demand. Instead of dropping incoming requests, they are held in
4720 the LND until the RPC service can accept more requests.
4723 Frequency : Sometimes during replay
4725 Description: Assertion failure in ll_local_open after replay.
4726 Details : If replay happened on an open request reply before we were able
4727 to set replay handler, reply will become not swabbed tripping the
4728 assertion in ll_local_open. Now we set the handler right after
4729 recognising of open request
4732 Frequency : very rare
4734 Description: kernel reports "badness in vsnprintf"
4735 Details : Reading from the "recovery_status" /proc file in small chunks
4736 may cause a negative length in lprocfs_obd_rd_recovery_status()
4737 call to vsnprintf() (which is otherwise harmless). Exit early
4738 if there is no more space in the output buffer.
4740 Severity : enhancement
4742 Description: clear OBD RPC statistics by writing to them
4743 Details : It is now possible to clear the OBD RPC statistics by writing
4744 to the "stats" file.
4749 Description: Client mtime is not the same on different clients after utimes
4750 Details : In some cases, the client was using the utimes() syscall on
4751 a file cached on another node. The clients now validate the
4752 ctime from the MDS + OSTs to determine which one is right.
4757 Description: Inability to activate failout mode
4758 Details : lconf script incorrectly assumed that in pythong string's numeric
4759 value is used in comparisons.
4762 Frequency : always with multiple stripes per file
4764 Description: Inefficient object allocation for mutli-stripe files
4765 Details : When selecting which OSTs to stripe files over, for files with
4766 a stripe count that divides evenly into the number of OSTs,
4767 the MDS is always picking the same starting OST for each file.
4768 Return the OST selection heuristic to the original design.
4773 Description: mount failures may take full timeout to return an error
4774 Details : Under some heavy load conditions it is possible that a
4775 failed mount can wait for the full obd_timeout interval,
4776 possibly several minutes, before reporting an error.
4777 Instead return an error as soon as the status is known.
4779 Frequency : quota enabled and large files being deleted
4781 Description: releasing more than 4GB of quota at once hangs OST
4782 Details : If a user deletes more than 4GB of files on a single OST it
4783 will cause the OST to spin in an infinite loop. Release
4784 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4789 Description: statfs data retrieved from /proc may be stale or zero
4790 Details : When reading per-device statfs data from /proc, in the
4791 {kbytes,files}_{total,free,avail} files, it may appear
4792 as zero or be out of date.
4795 Frequency : systems with MD RAID1 external journal devices
4797 Description: lconf's call to blkid is confused by RAID1 journal devices
4798 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4799 of returning all block devices that match the journal UUID.
4802 Frequency : always, for aggregate stripe size over 4GB
4804 Description: assertion fails when trying to use 4GB stripe size
4805 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4806 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4809 Frequency : always on ppc64
4811 Description: the first write on an ext3 filesystem with mballoc got stuck
4812 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4813 perform endianness conversion.
4815 ------------------------------------------------------------------------------
4817 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4819 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4820 INTEROPERATE with older versions automatically. Please read the
4821 user documentation before upgrading any part of a live system.
4822 * WARNING: Lustre networking configuration changes are required with
4823 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4826 * Support for kernels:
4827 2.6.9-22.0.2.EL (RHEL 4)
4828 2.6.5-7.244 (SLES 9)
4829 2.6.12.6 vanilla (kernel.org)
4832 Severity : enhancement
4833 Bugzilla : 7981/8208
4834 Description: Introduced Lustre Networking (LNET)
4835 Details : LNET is new networking infrastructure for Lustre, it includes
4836 a reorganized network configuration mode (see the user
4837 documentation for full details) as well as support for routing
4838 between different network fabrics. Lustre Networking Devices
4839 (LNDs) for the supported network fabrics have also been
4840 created for this new infrastructure.
4842 Severity : enhancement
4843 Description: Introduced Access control lists
4844 Details : clients can set ACLs on files and directories in order to have
4845 more fine-grained permissions than the standard Unix UGO+RWX.
4846 The MDS must be started with the "-o acl" mount option.
4848 Severity : enhancement
4849 Description: Introduced filesystem quotas
4850 Details : Administrators may now establish per-user quotas on the
4853 Severity : enhancement
4855 Description: Configuration change for the XT3
4856 The PTLLND is now used to run Lustre over Portals on the XT3
4857 The configure option(s) --with-cray-portals are no longer used.
4858 Rather --with-portals=<path-to-portals-includes> is used to
4859 enable building on the XT3. In addition to enable XT3 specific
4860 features the option --enable-cray-xt3 must be used.
4865 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4866 Details : A race between allocating a new llcd and re-getting the llcd_lock
4867 allowed another thread to grab newly-allocated llcd.
4869 Severity : enhancement
4871 Description: 2.6 OST async journal commit and locking fix to improve performance
4872 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4873 2.6 kernels are now async as they already were in 2.4 kernels so
4874 that they can commit concurrently with the network bulk transfer.
4875 For block-allocated files the filter allocation semaphore is held
4876 to avoid filesystem fragmentation during allocation. BKL lock
4877 removed for 2.6 xattr operations where it is no longer needed.
4882 Description: lconf incorrectly determined whether two IP networks could talk
4883 Details : In some more complicated routing and multiple-network
4884 configurations, lconf will avoid trying to make a network
4885 connection to a disjoint part of the IP space. It was doing the
4886 math incorrectly for one set of cases.
4891 Description: Fix for potential infinite loop processing records in an llog.
4892 Details : If an llog record is corrupted/zeroed, it is possible to loop
4893 forever in llog_process(). Validate the llog record length
4894 and skip the remainder of the block on error.
4897 Frequency : occasional (liblustre only)
4899 Description: liblustre could not open files whose last component is a symlink
4900 Details : sysio_path_walk() would incorrectly pass the open intent to
4901 intermediate path components.
4904 Frequency : rare (liblustre only with non-standard tuning)
4905 Bugzilla : 7201 (7350)
4906 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4907 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4908 after it has been released, only do this for referenced locks
4910 Severity : enhancement
4912 Description: specifying an (invalid) directory default stripe_size of -1
4913 would reset the directory default striping
4914 Details : stripe_size -1 was used internally to signal directory stripe
4915 removal, now use "all default" to signal dir stripe removal
4916 as a directory striping of "all default" is not useful
4919 Frequency : common for large clusters running liblustre clients
4921 Description: doing an ls when liblustre clients are running is slow
4922 Details : sending a glimpse AST to a liblustre client waits for every AST
4923 to time out, as liblustre clients will not respond. Since they
4924 cannot cache data we refresh the OST lock LVB from disk instead.
4926 Severity : enhancement
4928 Description: doing an ls at the same time as file IO can be slow
4929 Details : enqueue and other "small" requests can be blocked behind many
4930 large IO requests. Create a new OST IO portal for non-IO
4931 requests so they can be processed faster.
4934 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4936 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4937 created file with mode 000
4938 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4939 allow an ftruncate() as a non-root user to a file with mode 000.
4940 Lustre now respects this flag to disable mode checks when
4941 truncating a file owned by the user
4944 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4947 Description: Revoking locks from clients that went dead or catatonic might take
4949 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4950 cancellation of such locks instant on servers without waiting for
4951 any reply from clients. Clients drops these locks when cancel
4952 notification from server is received without replying.
4955 Frequency : liblustre-only, when liblustre client dies or becomes busy
4957 Description: Doing ls on Linux clients can take a long time with active
4959 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4960 granting such locks to it in the first place if possible. Locks
4961 are taken by proxy on the OST during the read or write and
4962 dropped immediately afterward. Add connect flags handling, do
4963 not grant locks to liblustre clients for glimpse ASTs.
4965 Severity : enhancement
4967 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4968 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4969 sequential reads and behaves badly with applications doing random
4970 reads. Improve it by reading ahead at least read region, and
4971 avoiding excessive large RPC for small reads.
4973 Severity : enhancement
4975 Description: Creating more than 1000 files for a single job may cause a load
4976 imbalance on the OSTs if there are also a large number of OSTs.
4977 Details : qos_prep_create() uses an OST index reseed value that is an
4978 even multiple of the number of available OSTs so that if the
4979 reseed happens in the middle of the object allocation it will
4980 still utilize the OSTs as uniformly as possible.
4985 Description: OST or MDS may oops in ping_evictor_main()
4986 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4987 but doesn't restart at beginning of obd_exports_timed list
4990 Severity : enhancement
4992 Description: improve by-nid export eviction on the MDS and OST
4993 Details : allow multiple exports with the same NID to be evicted at one
4994 time without re-searching the exports list.
4997 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4999 Description: MDS may oops in groups_free()
5000 Details : in rare race conditions a newly allocated group_info struct is
5001 freed again, and this can be NULL. The 2.4 compatibility code
5002 for groups_free() checked for a NULL pointer, but 2.6 did not.
5005 Frequency : common for liblustre clients doing little filesystem IO
5006 Bugzilla : 9352, 7313
5007 Description: server may evict liblustre clients accessing contended locks
5008 Details : if a client is granted a lock or receives a completion AST
5009 with a blocking AST already set it would not reply to the AST
5010 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5013 Frequency : lfs setstripe, only systems with more than 160 OSTs
5015 Description: unable to set striping with a starting offset beyond OST 160
5016 Details : llapi_create_file() incorrectly limited the starting stripe
5017 index to the maximum single-file stripe count.
5020 Frequency : LDAP users only
5022 Description: lconf did not handle in-kernel recovery with LDAP properly
5023 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5025 Severity : enhancement
5027 Description: bind OST threads to NUMA nodes to improve performance
5028 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5029 node and do their allocations there to localize memory access
5031 Severity : enhancement
5033 Description: llmount can determine client NID directly from Myrinet (GM)
5034 Details : the client NID code from gmnalnid was moved directly into
5035 llmount, removing the need to use this or specifying the
5036 client NID explicitly when mounting GM clients with zeroconf
5039 Frequency : if client is started with down MDS
5041 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5042 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5043 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5044 ptlrpc_import_delayed_req() because we didn't check if the
5045 request was interrupted, and we also didn't break out of the
5046 event loop if there was no timeout
5051 Description: data loss during non-page-aligned writes to a single file from
5052 both multiple nodes and multiple threads on one node at same time
5053 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5054 inconsistency led to false short-reads, that were cached and later
5055 used by ->prepare_write() to fill in partially written page,
5056 leading to data loss.
5059 Frequency : always, if lconf --abort_recovery used
5061 Description: lconf --abort_recovery fails with 'Operation not supported'
5062 Details : lconf was attempting to abort recovery on the MDT device and not
5065 Severity : enhancement
5067 Description: remove cleanup logs
5068 Details : replace lconf-generated cleanup logs with lustre internal
5069 cleanup routines. Eliminates the need for client-cleanup and
5072 Severity : enhancement
5074 Description: add support for EAs (user and system) on lustre filesystems
5075 Details : it is now possible to store extended attributes in the Lustre
5076 client filesystem, and with the user_xattr mount option it
5077 is possible to allow users to store EAs on their files also
5079 Severity : enhancement
5081 Description: Add possibility (config option) to show minimal available OST free
5083 Details : When compiled with --enable-mindf configure option, statfs(2)
5084 (and so, df) will return least minimal free space available from
5085 all OSTs as amount of free space on FS, instead of summary of
5086 free spaces of all OSTs.
5088 Severity : enhancement
5090 Description: do not expand extent locks acquired on OST-side
5091 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5092 by server: they are not cached anyway.
5095 Frequency : when mmap is used/binaries executed from Lustre
5097 Description: Unmmap pages before throwing them away from read cache.
5098 Details : llap_shrink cache now attempts to unmap pages before discarding
5099 them (if unmapping failed - do not discard). SLES9 kernel has
5100 extra checks that trigger if this unmapping is not done first.
5105 Description: lconf didn't resolve symlinks before checking to see whether a
5106 given mountpoint was already in use
5109 Frequency : when migrating failover services
5110 Bugzilla : 6395, 9514
5111 Description: When migrating a subset of services from a node (e.g. failback
5112 from a failover service node) the remaining services would
5113 time out and evict clients.
5114 Details : lconf --force (implied by --failover) sets the global obd_timeout
5115 to 5 seconds in order to quickly disconnect, but this caused
5116 other RPCs to time out too quickly. Do not change the global
5117 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5119 Severity : enhancement
5120 Frequency : if MDS is started with down OST
5121 Bugzilla : 9439,5706
5122 Description: Allow startup/shutdown of an MDS without depending on the
5123 availability of the OSTs.
5124 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5125 Add appropriate locking and lov-osc refcounts for safe
5126 cleaning. Add osc abort_inflight calls in case the
5127 synchronize never started.
5130 Frequency : occasional (Cray XT3 only)
5132 Description: root not authorized to access files in CRAY_PORTALS environment
5133 Details : The client process capabilities were not honoured on the MDS in
5134 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5135 been accessed by an authorized user then root was able to access
5136 the file on the local client also. The root user capabilities
5137 are now allowed on the MDS, as this environment has secure UID.
5140 Frequency : occasional
5142 Description: ldiskfs "too long searching" message happens too often
5143 Details : A debugging message (otherwise harmless) prints too often on
5144 the OST console. This has been reduced to only happen when
5145 there are fragmentation problems on the filesystem.
5150 Description: Division by zero in statfs when all OSCs are inactive
5151 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5152 lov_statfs divides by value returned by lov_get_stripecnt().
5156 Bugzilla : 9489, 3273
5157 Description: First write from each client to each OST was only 4kB in size,
5158 to initialize client writeback cache, which caused sub-optimal
5159 RPCs and poor layout on disk for the first written file.
5160 Details : Clients now request an initial cache grant at (re)connect time
5161 and so that they can start streaming writes to the cache right
5162 away and always do full-sized RPCs if there is enough data.
5163 If the OST is rebooted the client also re-establishes its grant
5164 so that client cached writes will be honoured under the grant.
5169 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5170 Details : Now I/O RPCs go to different portal number and (presumably) fast
5171 lock requests (and glimses) and other RPCs get their own service
5172 threads pool that should be able to service those RPCs
5175 Severity : enhancement
5177 Description: Ability to exchange lustre version between client and servers and
5178 issue warnings at client side if client is too old. Also for
5179 liblustre clients there is ability to refuse connection of too old
5181 Details : New 'version' field is added to connect data structure that is
5182 filled with version info. That info is later checked by server and
5186 Frequency : rare, liblustre only.
5187 Bugzilla : 9296, 9581
5188 Description: Two simultaneous writes from liblustre at offset within same page
5189 might proceed at the same time overwriting eachother with stale
5191 Details : I/O lock withing llu_file_prwv was released too early, before data
5192 actually was hitting the wire. Extended lock-holding time until
5193 server acknowledges receiving data.
5196 Frequency : extremely rare. Never observed in practice.
5198 Description: avoid generating lustre_handle cookie of 0.
5199 Details : class_handle_hash() generates handle cookies by incrementing
5200 global counter, and can hit 0 occasionaly (this is unlikely, but
5201 not impossible, because initial value of cookie counter is
5202 selected randonly). Value of 0 is used as a sentinel meaning
5203 "unassigned handle" --- avoid it. Also coalesce two critical
5204 sections in this function into one.
5206 Severity : enhancement
5208 Description: allow liblustre clients to delegate truncate locking to OST
5209 Details : To avoid overhead of locking, liblustre client instructs OST to
5210 take extent lock in ost_punch() on client's behalf. New connection
5211 flag is added to handle backward compatibility.
5213 Severity : enhancement
5214 Bugzilla : 4928, 7341, 9758
5215 Description: allow number of OST service threads to be specified
5216 Details : a module parameter allows the number of OST service threads
5217 to be specified via "options ost ost_num_threads={N}" in the
5218 OSS's /etc/modules.conf or /etc/modprobe.conf.
5222 Bugzilla : 6146, 9635, 9895
5223 Description: servers crash with bad pointer in target_handle_connect()
5224 Details : In rare cases when a client is reconnecting it was possible that
5225 the connection request was the last reference for that export.
5226 We would temporarily drop the export reference and get a new
5227 one, but this may have been the last reference and the export
5228 was just destroyed. Get new reference before dropping old one.
5230 Severity : enhancement
5231 Frequency : if client is started with failover MDS
5233 Description: Allow multiple MDS hostnames in the mount command
5234 Details : Try to read the configuration from all specified MDS
5235 hostnames during a client mount in case the "primary"
5238 Severity : enhancement
5240 Description: Stop sending data to evicted clients as soon as possible.
5241 Details : Check if the client we are about to send or are sending data to
5242 was evicted already. (Check is done every second of waiting,
5243 for which l_wait_event interface was extended to allow checking
5244 of exit condition at specified intervals).
5247 Frequency : rare, normally only when NFS exporting is done from client
5249 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5251 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5254 Frequency : always, if srand() is called before liblustre initialization
5256 Description: Liblustre uses system PRNG disturbing its usage by user application
5257 Details : Introduce internal to lustre fast and high-quality PRNG for
5258 lustre usage and make liblustre and some other places in generic
5259 lustre code to use it.
5261 Severity : enhancement
5262 Bugzilla : 9477, 9557, 9870
5263 Description: Verify that the MDS configuration logs are updated when xml is
5264 Details : Check if the .xml configuration logs are newer than the config
5265 logs stored on the MDS and report an error if this is the case.
5266 Request --write-conf, or allow starting with --old_conf.
5268 Severity : enhancement
5270 Description: Handle symlinks in the path when checking if Lustre is mounted.
5271 Details : Resolve intermediate symlinks when checking if a client has
5272 mounted a filesystem to avoid duplicate client mounts.
5277 Description: lconf can hit an error exception but still return success.
5278 Details : The lconf command catches the Command error exception at the top
5279 level script context and will exit with the associated exit
5280 status, but doesn't ensure that this exit status is non-zero.
5285 Description: failure of ptlrpc thread startup can cause oops
5286 Details : Starting a ptlrpc service thread can fail if there are a large
5287 number of threads or the server memory is very fragmented.
5288 Handle this without oopsing.
5291 Frequency : always, only if liblustre and non-default acceptor port was used
5293 Description: liblustre cannot connect to servers with non-default acceptor port
5294 Details : tcpnal_set_default_params() was not called and was therefore
5295 ignoring the environment varaible TCPNAL_PORT, as well as other
5296 TCPNAL_ environment variables
5301 Description: two objects could be created on the same OST for a single file
5302 Details : If an OST is down, in some cases it was possible to create two
5303 objects on a single OST for a single file. No problems other
5304 than potential performance impact and spurious error messages.
5308 Bugzilla : 5681, 9562
5309 Description: Client may oops in ll_unhash_aliases
5310 Details : Client dcache may become inconsistent in race condition.
5311 In some cases "getcwd" can fail if the current directory is
5317 Description: Inode refcounting problems in NFS export code
5318 Details : link_raw functions used to call d_instantiate without obtaining
5319 extra inode reference first.
5323 Bugzilla : 9942, 9903
5324 Description: Referencing freed requests leading to crash, memleaks with NFS.
5325 Details : We used to require that call to ll_revalidate_it was always
5326 followed by ll_lookup_it. Also with revalidate_special() it is
5327 possible to call ll_revalidate_it() twice for the same dentry
5328 even if first occurence returned success. This fix changes semantic
5329 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5330 reference on a request referred from the intent.
5331 ll_intent_release() then releases such a request.
5334 Frequency : rare, normally benchmark loads only
5336 Description: unlinked inodes were kept in memory on the client
5337 Details : If a client is repeatedly creating and unlinking files it
5338 can accumulate a lot of stale inodes in the inode slab cache.
5339 If there is no other client load running this can cause the
5340 client node to run out of memory. Instead flush old inodes
5341 from client cache that have the same inode number as a new inode.
5344 Frequency : SLES9 2.6.5 kernel and long filenames only
5345 Bugzilla : 9969, 10379
5346 Description: utime reports stale NFS file handle
5347 Details : SLES9 uses out-of-dentry names in some cases, which confused
5348 the lustre dentry revalidation. Change it to always use the
5352 Frequency : rare, unless heavy write-truncate concurrency is continuous
5353 Bugzilla : 4180, 6984, 7171, 9963, 9331
5354 Description: OST becomes very slow and/or deadlocked during object unlink
5355 Details : filter_destroy() was holding onto the parent directory lock
5356 while truncating+unlinking objects. For very large objects this
5357 may block other threads for a long time and slow overall OST
5358 responsiveness. It may also be possible to get a lock ordering
5359 deadlock in this case, or run out of journal credits because of
5360 the combined truncate+unlink. Solution is to do object truncate
5361 first in one transaction without parent lock, and then do the
5362 final unlink in a new transaction with the parent lock. This
5363 reduces the lock hold time dramatically.
5366 Frequency : rare, 2.4 kernels only
5368 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5369 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5370 before unmounting their filesystem in order to avoid deadlock.
5371 The kernel_locked() function in 2.4 kernels only checks whether
5372 the kernel lock is held, not whether it is this process that is
5373 holding it as 2.6 kernels do.
5378 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5379 Details : The client ptlrpc code may be trying to reconnect to a down
5380 server before a previous connection attempt has timed out.
5381 Increase the reconnect interval to be longer than the connection
5382 timeout interval to avoid sending duplicate connections to
5386 Frequency : echo_client brw_test command
5388 Description: fix echo_client to work with OST preallocated code
5389 Details : OST preallocation code (5137) didn't take echo_client IO path
5390 into account: echo_client calls filter methods outside of any
5391 OST thread and, hence, there is no per-thread preallocated
5392 pages and buffers to use. Solution: hijack pga pages for IO. As
5393 a byproduct, this avoids unnecessary data copying.
5397 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5398 Description: Client can oops in mdc_commit_close() after open replay
5399 Details : It was possible for the MDS to return an open request with no
5400 transaction number in mds_finish_transno() if the client was
5401 evicted, but without actually returning an error. Clients
5402 would later try to replay that open and may trip an assertion
5403 Simplify the client close codepath, and always return an error
5404 from the MDS in case the open is not successful.
5407 Frequency : rare, 2.6 OSTs only
5409 Description: OST may deadlock under high load on fragmented files
5410 Details : If there was a heavy load and highly-fragmented OST filesystems
5411 it was possible to have all the OST threads deadlock waiting on
5412 allocation of biovecs, because the biovecs were not released
5413 until the entire RPC IO was completed. Instead, release biovecs
5414 as soon as they are complete to ensure forward IO progress.
5416 Severity : enhancement
5418 Description: Support for specifying external journal device at mount
5419 Details : If an OST or MDS device is formatted with an external journal
5420 device, this device major/minor is stored in the ext3 superblock
5421 and may not be valid for failover. Allow detecting and
5422 specifying the external journal at mount time.
5427 Description: Mounting an MDS with pending unlinked files may cause oops
5428 Details : target_finish_recovery() calls mds_postrecov() which returned
5429 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5430 considers this an error and immediately begins cleaning up the
5431 lov, just after starting the mds_lov process
5433 Severity : enhancement
5435 Description: Implement 'lfs df' to report actual free space on per-OST basis
5436 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5437 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5438 usage of objects; '-h' to report in human readable format.
5440 ------------------------------------------------------------------------------
5442 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5449 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5450 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5451 at startup during orphan unlinking. Instead, skip buddy update
5452 before setup, as it will be regenerated after recovery is complete.
5457 Description: If an OST is inactive, its locks might reference stale inodes.
5458 Details : lov_change_cbdata() must iterate over all namespaces, even if
5459 they are inactive to clear inode references from the lock.
5461 Severity : enhancement
5462 Frequency : occasional, if non-standard max_dirty_mb used
5464 Description: Client will block write RPCs if not enough grant
5465 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5466 then the client will block writes while waiting for another RPC
5467 to complete instead of consuming its dirty limit. With change
5468 we get improved performance when max_dirty_mb is small.
5470 Severity : enhancement
5471 Bugzilla : 3389, 6253
5472 Description: Add support for supplementary groups on the MDS.
5473 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5474 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5475 lookups for user supplementary groups into a cache.
5479 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5480 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5481 for liblustre O_EXCL is enforced by MDS.
5485 Description: readdir never returns NULL in liblustre.
5486 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5487 dentry in struct dirent.
5491 Frequency : liblustre only, depends on application IO pattern
5492 Description: liblustre clients evicted if not contacting servers
5493 Details : Don't put liblustre clients into the ping_evictor list, so
5494 they will not be evicted by the pinger ever.
5496 Severity : enhancement
5498 Description: Add ability to evict clients by NID from MDS.
5499 Details : By echoing "nid:$NID" string into
5500 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5501 $NID would be instantly evicted from this MDS and from all active
5502 OSTs connected to it.
5506 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5507 Details : lookup_it_finish() used to query file size from OSTs that was not
5512 Description: service threads change working directory to that of init
5513 Details : Starting lustre service threads may pin the working directory
5514 of the parent thread, making that filesystem busy. Threads
5515 now change to the working directory of init to avoid this.
5519 Frequency : during shutdown only
5520 Description: shutdown with a failed MDS or OST can cause unmount to hang
5521 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5526 Frequency : 2.6 only
5527 Description: chmod/chown may include an extra supplementary group
5528 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5529 supplementary group and if none is specified this is used.
5532 Bugzilla : 5479 (6816)
5534 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5535 Details : If lookup is in progress on a file that is unlinked we might try
5536 to revalidate the inode and fail in revalidate after lookup is
5537 complete and ll_file_open() enqueues the open again but
5538 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5541 Frequency : always, if lconf --abort_recovery used
5543 Description: lconf --abort_recovery fails with 'Operation not supported'
5544 Details : lconf was attempting to abort recovery on the MDT device and not
5547 ------------------------------------------------------------------------------
5549 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5554 Frequency : rare (only unsupported configurations with a node running as an
5556 Bugzilla : 6514, 5137
5557 Description: Mounting a Lustre file system on a node running as an OST could
5559 Details : OSTs now preallocates memory needed to write out data at
5560 startup, instead of when needed, to avoid having to
5561 allocate memory in possibly low memory situations.
5562 Specifically, if the file system is mounted on on OST,
5563 memory pressure could force it to try to write out data,
5564 which it needed to allocate memory to do. Due to the low
5565 memory, it would be unable to do so and the node would
5566 become unresponsive.
5568 Severity : enhancement
5570 Description: Addition of lconf --service command line option
5571 Details : lconf now accepts a '--service <arg>' option, which is
5572 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5574 Severity : enhancement
5576 Description: Failover mode is now the default for OSTs.
5577 Details : By default, OSTs will now run in failover mode. To return to
5578 the old behaviour, add '--failout' to the lmc line for OSTs.
5580 Severity : enhancement
5582 Description: Health checks are now provided for MDS and OSTs
5583 Details : Additional detailed health check information on MSD and OSTs
5584 is now provided through the procfs health_check value.
5587 Frequency : occasional, depends on IO load
5589 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5590 after numerous create/delete cycles
5591 Details : The ext3 inode allocation policy would not allocate new inodes
5592 very well on the OSTs because there are no new directories
5593 being created. Instead we look for groups with free space if
5594 the parent directories are nearly full.
5599 Description: Network or server problems during mount may cause partially
5600 mounted clients instead of returning an error.
5601 Details : The config llog parsing code may overwrite the error return
5602 code during mount error handling, returning success instead
5608 Description: MDS can fail to allocate large reply buffers
5609 Details : After long uptimes the MDS can fail to allocate large reply
5610 buffers (e.g. zconf client mount config records) due to memory
5611 fragmentation or consumption by the buffer cache. Preallocate
5612 some large reply buffers so that these replies can be sent even
5613 under memory pressure.
5617 Frequency : rare (liblustre)
5618 Description: fsx running with liblustre complained that using truncate() to
5619 extend the file doesn't work. This patch corrects that issue.
5620 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5621 ATTR_SIZE and lsm use in llu_setattr_raw.
5625 Frequency : rare, only 2.6 kernels
5626 Description: Unusual file access patterns on the MDS may result in inode
5627 data being lost in very rare circumstances.
5628 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5629 code in the 2.6 kernel caused the inode and/or EA data not to
5630 be read from disk, causing single-file corruption.
5634 Frequency : rare, only 2.6 filesystems using extents
5635 Description: Heavy concurrent write and delete load may cause data corruption.
5636 Details : It was possible under high-load situations to have an extent
5637 metadata block in the block device cache from a just-unlinked
5638 file overwrite a newly-allocated data block. We now unmap any
5639 metadata buffers that alias just-allocated data blocks.
5643 Frequency : filesystems with default stripe_count larger than 77
5644 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5645 Details : lconf specifies an inode size of 4096 bytes when the default
5646 stripe_count is larger than 77. This conflicts with the default
5647 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5648 this case to avoid pinning too much memory for large EAs.
5650 ------------------------------------------------------------------------------
5652 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5657 Frequency : rare (extremely heavy IO load with hundreds of clients)
5659 Description: Client is evicted, gets IO error writing to file
5660 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5661 caused clients to be evicted for AST timeouts. The fixes in
5662 bug 5192 mean we no longer need to have such short AST timeouts
5663 so ldlm_timeout has been increased.
5666 Frequency : occasional during --force or --failover shutdown under load
5667 Bugzilla : 5949, 4834
5668 Description: Server oops/LBUG if stopped with --force or --failover under load
5669 Details : a collection of import/export refcount and cleanup ordering
5670 issues fixed for safer force cleanup
5673 Frequency : only filesystems larger than 120 OSTs
5674 Bugzilla : 5990, 6223
5675 Description: lfs getstripe would oops on a very large filesystem
5676 Details : lov_getconfig used kfree on vmalloc'd memory
5679 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5680 Bugzilla : 6242, 6243
5681 Description: reading from files that had been truncated to a non-zero size
5682 but never opened returned no data
5683 Details : ll_file_read() reads zeros from no-object files to EOF
5688 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5689 Details : There's little harm in aborting MDS/OSS recovery and letting it
5690 try again, so I removed the LASSERT and return an error instead.
5692 Severity : enhancement
5694 Description: New debugging infrastructure for tracking down data corruption
5695 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5696 (b) cover more of the client-side code path, and (c) try to narrow
5697 down where problems occurred
5701 Bugzilla : 3819, 4364, 4397, 6313
5702 Description: Racing close and eviction MDS could cause assertion in mds_close
5703 Details : It was possible to get multiple mfd references during close and
5704 client eviction, leading to one thread referencing a freed mfd.
5706 Severity: : enhancement
5707 Bugzilla : 3262, 6359
5708 Description: Attempts to reconnect to servers are now more aggressive.
5709 Details : This builds on the enhanced upcall-less recovery that was added
5710 in 1.4.2. When trying to reconnect to servers, clients will
5711 now try each server in the failover group every 10 seconds. By
5712 default, clients would previously try one server every 25 seconds.
5717 Description: After recovery, certain operations trigger a failed
5718 assertion on a client.
5719 Details : Failing over an mds, using lconf -d --failover, while a
5720 client was doing a readdir() call would cause the client to
5721 LBUG after recovery completed and the readdir() was resent.
5723 Severity : enhancement
5725 Description: Default groups are now added by lconf
5726 Details : You can now run lconf --group <servicename> without having to
5727 manually add groups with lmc.
5730 Frequency : occasional
5732 Description: Nodes with an elan id of 0 trigger a failed assertion
5735 Frequency : always when accessing e.g. tty/console device nodes
5737 Description: tty and some other devices nodes cannot be used on lustre
5738 Details : file's private_data field is used by device data and lustre
5739 values in there got lost. New field was added to struct file to
5740 store fs-specific private data.
5743 Frequency : when exporting Lustre via NFS
5745 Description: NFSD failed occasionally when looking up a path component
5746 Details : NFSD is looking up ".." which was broken in ext3 directories
5747 that had grown large enough to become hashed.
5750 Frequency : Clusters with multiple interfaces not on the same subnet
5752 Description: Nodes will repeatedly try to reconnect to an interface which it
5753 cannot reach and report an error to the log.
5754 Details : Extra peer list entries will be created by lconf with some peers
5755 unreachable. lconf now validates the peer before adding it.
5758 Frequency : Only if a default stripe is set on the filesystem root.
5760 Description: Setting a default stripe on the filesystem root prevented the
5761 filesystem from being remounted.
5762 Details : The client was sending extra request flags in the root getattr
5763 request and did not allocate a reply buffer for the dir EA.
5766 Frequency : occasional, higher if lots of files are accessed by one client
5767 Bugzilla : 6159, 6097
5768 Description: Client trips assertion regarding lsm mismatch/magic
5769 Details : While revalidating inodes the VFS looks up inodes with ifind()
5770 and in rare cases can find an inode that is being freed.
5771 The ll_test_inode() code will free the lsm during ifind()
5772 when it finds an existing inode and then the VFS later attaches
5773 this free lsm to a new inode.
5777 Bugzilla : 6422, 7030
5778 Description: MDS deadlock between mkdir and client eviction
5779 Details : Creating a new file via mkdir or mknod (starting a transaction
5780 and getting the ns lock) can deadlock with client eviction
5781 (gets ns lock and trying to finish a synchronous transaction).
5784 Frequency : occasional
5785 Description: While starting a server, the fsfilt_ext3 module could not be
5787 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5788 kernels. Previously, lconf would still use the ext3 name
5789 when trying to load modules. Now, it will correctly use
5790 ext3 on 2.4 and ldiskfs on 2.6.
5792 Severity : enhancement
5793 Description: The default stripe count has been changed to 1
5794 Details : The interpretation of the default stripe count (0, to lfs
5795 or lmc) has been changed to mean striping across a single
5796 OST, rather than all available. For general usage we have
5797 found a stripe count of 1 or 2 works best.
5799 Severity : enhancement
5800 Description: Add support for compiling against Cray portals.
5801 Details : Conditional compiling for some areas that are different
5805 Frequency : occasional
5806 Bugzilla : 6409, 6834
5807 Description: Creating files with an explicit stripe count may lead to
5808 a failed assertion on the MDS
5809 Details : If some OSTs are full or unavailable, creating files may
5810 trigger a failed assertion on the MDS. Now, Lustre will
5811 try to use other servers or return an error to the
5815 Frequency : occasional
5817 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5818 caused lock timeout during readahead (since 1.4.2).
5819 Details : Processes doing readahead might match a lock that hasn't been
5820 granted yet if there are overlapping and conflicting lock
5821 requests. The readahead process waits on ungranted lock
5822 (original lock is CBPENDING), while OST waits for that process
5823 to cancel CBPENDING read lock and eventually evicts client.
5825 Severity : enhancement
5827 Description: Initial enabling of flock support for clients
5828 Details : Implements fcntl advisory locking and file status functions.
5829 This feature is provided as an optional mount flag (default
5830 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5831 locking are implemented yet, and those that are are not guaranteed
5832 to be completely correct in production environments.
5833 mount -t lustre -o [flock|noflock] ...
5836 Frequency : occasional
5838 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5839 assertion in the ext3 JBD (journaling) layer.
5840 Details : The b_committed_data struct is protected by the big kernel lock
5841 in 2.4 kernels, serializing journal_commit_transaction() and
5842 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5843 access to this struct. In 2.6 kernels there is finer grained
5844 locking to improve SMP performance of the JBD layer.
5848 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5853 Description: merge and cleanup kernel patches.
5856 -----------------------------------------------------------------------------
5858 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5860 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5861 versions of Lustre. Please update all servers and clients to
5862 version 1.4.2 or later at the same time. You must also run
5863 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5864 to update the configuration logs.
5866 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5867 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5868 - swab llog records of type '0' so we get proper header size/idx (5861)
5869 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5870 - fix rename of one directory over another leaking an inode (5953)
5871 - avoid SetPageDirty on 2.6 (5981)
5872 - don't re-add just-being-destroyed locks to the waiting list (5653)
5873 - when creating new directories, inherit the parent's custom
5874 striping settings if present parent (3048)
5875 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5876 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5877 - don't hold client locks on temporary worklist from l_lru (5666)
5878 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5879 - automatically evict dead clients (5921)
5880 - Update file size properly in create+truncate+fstat case (6196)
5881 - Do not unhash mountpoint dentries, do not allow removal of
5883 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5884 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5885 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5886 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5887 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5888 - Correctly handle reads of files with no objects (6243)
5889 - lctl recover will also mark a device active if deactivate used (5933)
5891 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5892 - allow --write-conf on an MDS with different nettype than client (5619)
5893 - don't write config llogs to MDS for mounts not from that MDS (5617)
5894 - lconf should create multiple TCP connections from a client (5201)
5895 - init scripts are now turned off by default; run chkconfig --on
5896 lustre and chkconfig --on lustrefs to use them
5897 - upcalls are no longer needed for clients to recover to failover
5899 - add --abort-recovery option to lconf to abort recovery on device
5901 - add support for an arbitrary number of OSTs (3026)
5902 - Quota support protocol changes.
5903 - forward compatibility changes to wire structs (6007)
5904 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5905 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5906 - improved "lustre status" script
5907 - initialize blocksize for non-regular files (6062)
5908 - added --disable-server and --disable-client configure options (5782)
5909 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5910 - Vanilla 2.4.29 support
5911 - increase maximum number of obd devices to 520 (6242)
5912 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5913 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5914 SLES 9 2.6.5 kernels
5915 - limit stripes per file to 160 (the maximum EA size) (6093)
5917 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5920 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5921 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5922 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5923 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5924 - small ext3 extents cleanups and fixes (5733)
5925 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5926 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5927 - avoid needless client->OST connect, fix handle mismatch (5317)
5928 - fix DLM error path that led to out-of-sync client, long delays (5779)
5929 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5930 - fix several locking issues related to i_size (5492,5624,5654,5672)
5931 - don't move pending lock onto export if it is already evicted (5683)
5932 - fix kernel oops when creating .foo in unlinked directory (5548)
5933 - fix deadlock in obdfilter statistics vs. object create (5811)
5934 - use time_{before,after} to avoid timer jiffies wrap (5882)
5935 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5936 - Do not leak request if server was not able to process it (5154)
5937 - If mds_open unable to find parent dir, make that negative lookup(5154)
5938 - don't create new directories with extent-mapping (5909, 5936)
5940 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5941 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5942 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5943 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5944 - support for RHEL4 kernel on i686 (5773)
5945 - provide error messages when incompatible logs are encountered (5898)
5947 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5948 * version 1.4.0.10 (1.4.1 release candidate 1)
5950 - don't keep a lock reference when lock is not granted (4238)
5951 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5952 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5953 - drop import inflight refcount on signal_completed_replay error (5255)
5954 - unlock page after async write error during send (3677)
5955 - handle missing objects in filter_preprw_read properly (5265)
5956 - no transno return for symlink open, don't save no-trasno open (3440)
5957 - don't try to complete elan receive that already failed (4012)
5958 - free RPC server reply state on error (5406)
5959 - clean up thread from ptlrpc_start_thread() on error (5160)
5960 - readahead could read extra page into cache that wasn't ejected (5388)
5961 - prevent races in class_attach/setup/cleanup/detach (5260)
5962 - don't dereference de->d_inode after l_dput of de (5458)
5963 - use "int" for stripe value returned from lock_to_stripe (5544)
5964 - mballoc allocation and error-checking fixes in 2.6 (5504)
5965 - block device patches to fix I/O request sizes in 2.6 (5482)
5966 - look up hostnames for IB nals (5602)
5967 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5968 - don't start multiple acceptors for the same port (5277)
5969 - fix incorrect LASSERT in mds_getattr_name (5635)
5970 - export a proc file for general "ping" checking (5628)
5971 - fix "lfs check" to not block when the MDS is down (5628)
5973 - service request history (4965)
5974 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5975 - create an "evict_client" /proc entry on OSTs, like the MDS has
5976 - fix mount usage message, return errors per mount(8) (5168)
5977 - change grep [] to grep "[]" in tests so they work in more UMLs
5978 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5979 - remove ancient LOV_MAGIC_V0 EA support (5047)
5980 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5981 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5982 - allow readahead allocations to fail when low on memory (5383)
5983 - mmap locking landed again, after considerable improvement (2828)
5984 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5986 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5989 - send OST transaction number in read/write reply to free req (4966)
5990 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5991 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5992 - the watchdog thread now runs as interruptible (5246)
5993 - flock/lockf fixes (but it's still disabled, pending 5135)
5994 - don't use EXT3 constants in llite code (5094)
5995 - memory shortage at startup could cause assertion (5176)
5997 - reorganization of lov code
5998 - single portals codebase
6000 - add extents/mballoc support (5025)
6001 - direct I/O reads in the obdfilter (4048)
6002 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6004 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6007 - send OST transaction number in read/write reply to free req (4966)
6008 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6009 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6010 - the watchdog thread now runs as interruptible (5246)
6011 - handle missing objects in filter_preprw_read properly (5265)
6012 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6013 - drop import inflight refcount on signal_completed_replay error (5255)
6014 - unlock page after async write error during send (3677)
6015 - return original error code on reconstructed replies (3761)
6016 - no transno return for symlink open, don't save no-trasno open (3440)
6018 - add pid to ldlm debugging output (4922)
6019 - bump the watchdog timeouts -- we can't handle 30sec yet
6020 - extra debugging for orphan dentry/inode bug (5259)
6022 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6025 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6026 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6027 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6028 - fix NULL dereference in /proc/sys/portals/routes (4827)
6029 - allow failed mdc_close() operations to be interrupted (4561)
6030 - stop precreate on OST before MDS would time out on it (4778)
6031 - don't send partial-page writes before EOF from client (4410)
6032 - discard client grant for sub-page writes on large-page clients (4520)
6033 - don't free dentries not owned by NFS code, check generation (4806)
6034 - fix lsm leak if mds_create_objects() fails (4801)
6035 - limit debug_daemon file size, always print CERROR messages (4789)
6036 - use transno after validating reply (3892)
6037 - process timed out requests if import state changes (3754)
6038 - update mtime on OST during writes, return in glimpse (4829)
6039 - add mkfsoptions to LDAP (4679)
6040 - use ->max_readahead method instead of zapping global ra (5039)
6041 - don't interrupt __l_wait_event() during strace
6043 - add software watchdogs to catch hung threads quickly (4941)
6044 - make lustrefs init script start after nfs is mounted
6045 - fix CWARN/ERROR duplication (4930)
6046 - return async write errors to application if possible (2248)
6047 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6048 - print NAL number in %x format (4645)
6049 - update barely-supported suse-2.4.21-171 series (4842)
6050 - support for sles 9 %post scripts
6051 - support for building 2.6 kernel-source packages
6052 - support for sles km_* packages
6054 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6057 - ignore -ENOENT errors in osc_destroy (3639)
6058 - notify osc create thread that OSC is being cleaned up (4600)
6059 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6060 - reconstruct ost_handle() like mds_handle() (4657)
6061 - create a new thread to do import eviction to avoid deadlock (3969)
6062 - let lconf resolve symlinked-to devices (4629)
6063 - don't unlink "objects" from directory with default EA (4554)
6064 - hold socknal file ref over connect in case target is down (4394)
6065 - allow more than 32000 subdirectories in a single directory (3244)
6066 - fix blocks count for O_DIRECT writes (3751)
6067 - OST returns ENOSPC from object create when no space left (4539)
6068 - don't send truncate RPC if file size isn't changing (4410)
6069 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6070 - bind to privileged port in socknal and tcpnal (3689)
6072 - rate limit CERROR/CWARN console message to avoid overload (4519)
6073 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6074 - basic mmap support (3918)
6075 - kernel patch series update from b1_4 (4711)
6077 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6080 - avoid crash during MDS cleanup with OST shut down (2775)
6081 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6082 - don't use bad inodes on the MDS (3744)
6083 - dynamic object preallocation to improve recovery speed (4236)
6084 - don't hold spinlock over lock dumping or change debug flags (4401)
6085 - don't zero obd_dev when it is force cleaned (3651)
6086 - print grants to console if they go negative (4431)
6087 - "lctl deactivate" will stop automatic recovery attempts (3406)
6088 - look for existing locks in ldlm_handle_enqueue() (3764)
6089 - don't resolve lock handle twice in recovery avoiding race (4401)
6090 - revalidate should check working dir is a directory (4134)
6092 - don't always mark "slow" obdfilter messages as errors (4418)
6094 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6097 - don't close LustreDB during write_conf until it is done (3860)
6098 - fix typo in lconf for_each_profile (3821)
6099 - allow dumping logs from multiple threads at one time (3820)
6100 - don't allow multiple threads in OSC recovery (3812)
6101 - fix debug_size parameters (3864)
6102 - fix mds_postrecov to initialize import for llog ctxt (3121)
6103 - replace config semaphore with spinlock (3306)
6104 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6105 - don't allow enqueue to complete on a destroyed export (3822)
6106 - down write_lock before checking llog header bitmap (3825)
6107 - recover from lock replay timeout (3764)
6108 - up llog sem before sending rpc (3652)
6109 - reduce ns lock hold times when setting kms (3267)
6110 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6111 - fix NULL deref and obd_dev leak on setup error (3312)
6112 - replace some LBUG about llog ops with error handling (3841)
6113 - don't match INVALID dentries from d_lookup and spin (3784)
6114 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6115 - fix invalid assertion in ptlrpc_set_wait (3880)
6117 - add libwrap support for the TCP acceptor (3996)
6118 - add /proc/sys/portals/routes for non-root route listing (3994)
6119 - allow setting MDS UUID in .xml (2580)
6120 - print the stack of a process that LBUGs (4228)
6122 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6125 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6126 - make sure to unset replay flag from failed open requests (3440)
6127 - if default stripe count is 0, use OST count for inode size (3636)
6128 - update parent mtime/ctime on client for create/unlink (2611)
6129 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6130 - free recovery state on server during a forced cleanup (3571)
6131 - unregister_reply for resent reqs (3063)
6132 - loop back devices mounting and status check on 2.6 (3563)
6133 - fix resource-creation race that can provoke i_size == 0 (3513)
6134 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6135 - more debugging for page-accounting assertion (3746)
6136 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6137 - avoid deadlock after precreation failure (3758)
6138 - fix race and lock order deadlock in orphan handling (3450, 3750)
6139 - add validity checks when grabbing inodes from l_ast_data (3599)
6141 - add /proc/.../recovery_status to obdfilter (3428)
6142 - lightweight CDEBUG infrastructure, debug daemon (3668)
6143 - change default OSC RPC parameters to be better on small clusters
6144 - turn off OST read cache for files smaller than 32MB
6145 - install man pages and include them in rpms (3100)
6146 - add new init script for (un)mounting lustre filesystems (2593)
6147 - run chkconfig in %post for init scripts (3701)
6148 - drop scimac NAL (unmaintained)
6150 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6153 - clean kiobufs before and after use (3485)
6154 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6155 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6156 - change init script to fail more clearly if not run as root (1528)
6157 - allow clients to reconnect during replay (1742)
6158 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6159 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6160 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6161 - fix cond_resched (3554)
6162 - only evict unfinished clients after recovery (3515)
6163 - allow bulk resend, prevent data loss (3570)
6164 - dynamic ptlrpc request buffer allocation (2102)
6165 - don't allow unlinking open directory if it isn't empty (2904)
6166 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6167 - remove extraneous obd dereference causing LASSERT failure (3334)
6168 - don't use get_cycles() when creating temp. files on the mds (3156)
6169 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6170 - handle EEXIST for set-stripe, set proper directory name (3336)
6172 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6173 - fix ksocknal_fmb_callback() error messages (2918)
6175 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6178 - don't copy lvb into (possibly NULL) reply on error (2983)
6179 - don't deref dentry after dput, don't free lvb on error (2922)
6180 - use the kms to determine writeback rpc length (2947)
6181 - increment oti_logcookies when osc is inactive (2948)
6182 - update client's i_blocks count via lvb messages (2543)
6183 - handle intent open/close of special files properly (1557)
6184 - mount MDS with errors=remount-ro, like obdfilter (2009)
6185 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6186 - don't use cancelling-locks' kms values (2947)
6187 - use highest lock extent for kms, not last one (2925)
6188 - don't dereference ERR_PTR() dentry in error handling path (3107)
6189 - fix thread race in portals_debug_dumplog() (3122)
6190 - create lprocfs device entries at setup instead of at attach (1519)
6191 - common AST error handler, don't evict client on completion race (3145)
6192 - zero nameidata in detach_mnt in 2.6 (3118)
6193 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6194 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6195 - initialize RPC timeout timer earlier for 2.6 (3219)
6196 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6197 - print nal/nid for unknown nid (3258)
6198 - additional checks for oscc recovery before doing precreate (3284)
6199 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6200 - don't crash in mdc_close for bad permissions on open (3285)
6201 - zero i_rdev for non-device files (3147)
6202 - clear page->private before handing to FS, better assertion (3119)
6203 - tune the read pipeline (3236)
6204 - fix incorrect decref of invalidated dentry (2350)
6205 - provide read-ahead stats and refine rpc in flight stats (3328)
6206 - don't hold journal transaction open across create RPC (3313)
6207 - update atime on MDS at close time (3265)
6208 - close LDAP connection when recovering to avoid server load (3315)
6209 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6210 - don't leak open file on MDS after open resend (3325)
6211 - serialize filter_precreate and filter_destroy_precreated (3329)
6212 - loop device shouldn't call sync_dev() for nul device (3092)
6213 - clear page cache after eviction (2766)
6214 - resynchronize MDS->OST in background (2824)
6215 - refuse to mount the same filesystem twice on same mountpoint (3394)
6216 - allow llmount to create routes for mounting behind routers (3320)
6217 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6218 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6219 - fix rare problem with rename on htree directories (3417)
6221 - allow default OST striping configuration per directory (1414)
6222 - fix compilation for qswnal for 2.6 kernels (3125)
6223 - increase maximum number of MDS request buffers for large systems
6224 - change liblustreapi to be useful for external progs like lfsck (3098)
6225 - increase local configuration timeout for slow disks (3353)
6226 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6228 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6231 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6232 - don't overwrite extent policy data in reply if lock was blocked (2901)
6233 - drop filter export grants atomically with removal from device (2663)
6234 - del obd_self_export from work_list in class_disconnect_exports (2908)
6235 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6236 - swab reply message in mdc_close, other PPC fixes (2464)
6237 - fix destroying of named logs (2325)
6238 - overwrite old logs when running lconf --write_conf (2264)
6239 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6240 - fix race in target_handle_connect (2898)
6241 - mds_reint_create() should take same inode create lock (2926)
6242 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6243 - don't close files for self_export to avoid uninitialized obd (2936)
6244 - allow MDS with the same name as client node (2939)
6245 - hold dentry reference for closed log files for unlink (2325)
6246 - reserve space for all logs during transactions (2059)
6247 - don't evict page beyond end of stripe extent (2925)
6248 - don't oops on a deleted current working directory (2399)
6249 - handle hard links to targets without a parent properly (2517)
6250 - don't dereference NULL lock when racing during eviction (2867)
6251 - don't grow lock extents when lots of conflicting locks (2919)
6253 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6256 - account for cache space usage on clients to avoid data loss (974)
6257 - lfsck support in lustre kernel code (2349)
6258 - reduce journal credits needed for BRW writes (2370)
6259 - orphan handling to avoid losing space on client/server crashes
6260 - ptlrpcd can be blocked, stopping ALL progress (2477)
6261 - use lock value blocks to assist in proper KMS, faster stat (1021)
6262 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6263 - recovery for initial connections (2355)
6264 - fixes for mds_cleanup_orphans (1934)
6265 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6266 - block all file creations until orphan recovery completes (1901)
6267 - client remove rq_connection from request struct (2423)
6268 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6269 - recovery timer race (2670)
6270 - mdc_close recovey bug (2532)
6271 - ptlrpc cleanup bug (2710)
6272 - mds timeout on local locks (2588)
6273 - namespace lock held during RPCs (2431)
6274 - handle interrupted sync write properly (2503)
6275 - don't try to handle a message that hasn't been replied to (2699)
6276 - client assert failure during cleanup after abort recovery (2701)
6277 - leak mdc device after failed mount (2712)
6278 - ptlrpc_check_set allows timedout requests to complete (2714)
6279 - wait for inflight reqs when ptlrpcd finishes (2710)
6280 - make sure unregistered services are removed from the srv_list
6281 - reset bulk XID's when resending them (caught by 1138 test)
6282 - unregister_bulk after timeout
6283 - fix lconf error (2694)
6284 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6285 - readahead locks pages, leaves pending causing memory pressure (2673)
6286 - increase OST request buffers to 4096 on large machines (2729)
6287 - fix up permission of existing directories in simple_mkdir (2661)
6288 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6289 - don't assign transno to errored transactions (2742)
6290 - don't delete objects on OST if given a bogus objid from MDS (2751)
6291 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6292 - if rq_no_resend, then timeout request after recovery (2432)
6293 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6294 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6295 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6296 - MDS recovery shouldn't delete objects that it has given out (2730)
6297 - if enqueue arrives after completion, don't clobber LVB (2819)
6298 - don't unlock pages twice when trigger_group_io returns error (2814)
6299 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6300 - don't write pages to disk if there was an error (1450)
6301 - don't ping imports that have recovery disabled (2676)
6302 - take buffered bytes into account when balancing socknal conn (2817)
6303 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6304 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6305 - remove little-endian swabbing of llog records (1987)
6306 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6307 - retry reposting request buffers if they fail (1191)
6308 - grow extent at grant time to avoid granting a revoked lock (2809)
6309 - lock revoke doesn't evict page if covered by a second lock (2765)
6310 - disable VM readahead to avoid reading outside lock extents (2805)
6312 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6313 - updated kernel patches for hp-2.4.20 kernel (2681)
6315 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6318 - fix truncated write corruption (2366)
6319 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6321 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6322 - ensure len(uuid) < 37 in lmc (1171)
6323 - fix ia64 OOPS in llog_test (2255)
6324 - zero end of page at obdfilter for partial page writes (2648)
6325 - don't leave stale dentries around after renames (bug 2428)
6326 - fix timeouts when evicting a client with a single lock held (2642)
6327 - set deadline for the initial HELLO message to drain (2634)
6328 - print out dotted-quad IP addresses in the socknal (2302)
6330 - additional debugging for MDS client eviction problem (2443)
6331 - fix mkfsoptions support for osts (2603, 2604)
6333 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6336 - add series for the vanilla 2.6.0 kernel
6337 - add series for the vanilla 2.4.24 kernel
6338 - add series for a cray x86/64 UL kernel drop
6339 - fix xattr patches for the vanilla 2.4.19 series
6341 - generate true UUIDs in lmc (1171)
6342 - have portals stack dumping break in UML (2466)
6343 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6344 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6345 - wake callback waiting threads on client eviction (2460)
6346 - Add --ptldebug and --subsystem to lmc (1719)
6347 - update assertion to allow safe interrupt allocation
6348 - set rq_no_resend for cancel requests (2432)
6349 - recalculate ptlrpcd timeout after resend (2494)
6350 - call vfs_rmdir when removing pending directories (2368)
6351 - fix renaming a file to itself (2429)
6352 - lmc creates a default one-stripe lov (2454)
6353 - expand procfs space to handle large clusters (2326)
6354 - increase UML stack to avoid overflow
6355 - update lconf's list of debug and subsystem masks
6356 - fix lfs find --obd (2510)
6357 - /proc tunable for disabling filter read caching (2591)
6358 - stop rpm packages from altering slapd.conf (2301)
6359 - disable nagle in the socknal under 0conf (2578)
6360 - choose mds inode size based on stripe count (2572)
6361 - fix kernel-source rpm problems (2516)
6363 - add --disable-doc to avoid pdf generation (2421)
6364 - update documentation, tests, type-os, comments
6365 - avoid format warnings on ia64
6366 - remove the TOE NAL
6367 - tiny code cleanups by removing unused fields
6369 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6372 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6373 - remove the most blatant lies from BUILDING file (2371)
6374 - change default debug level to reasonable production setting
6375 - reduce client side cache size to reduce cache flush time
6376 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6377 - make TCP zerocopy and pinger support enabled by default (2476)
6378 - sync writes completed after process exits caused crashes (2319)
6379 - maintain correct mount count on the MDS (2356)
6380 - backout 1557, because 2316 wasn't really fixed
6381 - better file I/O statistics gathering in /proc
6382 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6383 - another kernel patch to fix zero-copy TCP function export
6384 - don't take duplicate lock when processing re-sent getattr (2420)
6385 - lctl uses obd_self_export instead of creating new conn (2353)
6386 - MDS/OST recovery case which requires object creation asserted (2425)
6387 - move lfs from /usr/sbin to /usr/bin in packages
6388 - fix race between mds_client_add and mds_client_free (2417)
6389 - use kmalloc instead of slabs in portals (2430)
6390 - don't create duplicate records when a failover MDS is present (2442)
6391 - remove unnecessary mount age check (2332)
6392 - don't remove directory inodes from locks prematurely (2451)
6393 - don't break if MDS service name is the same as hostname (2103)
6394 - fix races in client write RPC generation when cache full (2482)
6396 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6399 - remove now-unused request->rq_obd (278)
6400 - if an allocation fails, print out how much memory we've used (1933)
6401 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6402 - add the "configurable stack size" patch to most series files (1256)
6403 - ability to write large log records, for 100+ OST configs (2306)
6404 - fix NULL deref when filter_prep fails (2314)
6405 - fix operator precedence error in filter_sync
6406 - dynamic allocation of socknal TX descriptors (2315)
6407 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6408 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6409 - ensure that CWARN messages in Portals always get to the syslog
6410 - __init/__exit are not for prototype decls (ldlm_init/exit)
6411 - x86-64 compile warning fixes
6412 - fix gateway LMC keyword conflict (2318)
6413 - fix MDS lock inversions in getattr/reint paths (1844)
6414 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6415 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6416 - DLM race condition prevented some lock evictions (2328)
6417 - ENOMEM detection and retry on socknal sends (2230)
6418 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6419 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6420 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6421 - better error messages when a client is rejected during recovery (1505)
6422 - avoid cancelling locks which were never granted, after failure (2330)
6423 - fix i_sem/journal inversion in mds_client_add (2333)
6424 - fix truncate/getattr lock cycle deadlock (2334)
6425 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6426 - fix two rare exit paths which could leak an l_lock() ref (2321)
6427 - fencepost error in MDS/OST orphan recovery (2226)
6428 - make log record alignment 8 bytes (1988)
6429 - lstripe now fails when requested offset > ost_count (2237)
6430 - ensure that all kernel series have a complete list.h (1607)
6431 - fix crashes in special-file operations (2316)
6432 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6434 - allow configurable automake binary, for testing new versions
6435 - small update to the lfs documentation
6437 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6439 * fix negative export reference count in fsfilt_sync (2312)
6441 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6442 * release candidate 0.9.1
6444 - orphans are moved into the PENDING directory for possible recovery
6445 - replayed opens now open by fid for orphan/rename safety (1042)
6446 - last close of an orphan inode generates a transno (683)
6447 - chdir() and mount() now pin the directory entry (1020)
6448 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6449 - discard very old requests without processing them (1502)
6450 - remove some common, well-understood CERRORs (1505)
6451 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6452 - clear "grant" flags in OST replies until OST grant code lands (1644)
6453 - fix read performance by not clobbering i_blksize on client (1598)
6454 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6455 - make LRU size a /proc tunable, clears locks when reduced (707)
6456 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6457 - server locks take references on exports now (1558)
6458 - build fixes for 2.4.20-rh trees (1663)
6459 - return an error from lov_create if all OSCs are inactive (1751)
6460 - fix import levels when a reconnect happens without a timeout (1597)
6461 - exit early from mds_open if we get a lookup error (1749)
6462 - partial page read at EOF wouldn't wait for disk before sending (1642)
6463 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6464 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6465 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6466 - write/truncate lock inversion (1639)
6467 - Don't auto-load obdclass, portals modules during cleanup (1495)
6468 - fix timestamps from jumping to "now" (1763)
6469 - extra journal assertions (1648)
6470 - add an extra multiunlink test (1771)
6471 - fix read_record/write_record API (1776)
6472 - fix leak of offset_extent, possible incorrect i_size later (1772)
6473 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6474 - Debugging for the kqswnal_get_idle_tx problems (1820)
6475 - Allow recovery to be attempted multiple times (1536)
6476 - Write out MDS last_rcvd file after it is first created (1600)
6477 - Fix tx_descriptor leak in failed transmit situations (1827)
6478 - ext3 journaling fixes for assertion failure after IO error (1871)
6479 - class_export_put() on freed export after completion AST error (1896)
6480 - Fix revalidate looping in VFS (1322)
6481 - Don't access a freed export during MDS_REINT timeout (1521)
6482 - Add open-unlink recovery support on the MDS (1673,1764)
6483 - Return an error if no MDS data was read from last_rcvd (1946)
6484 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6485 - Don't setup a disk device that doesn't match exported UUID (317)
6486 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6487 - avoid committing NULL handle in force close
6488 - local.sh is now a one-stripe LOV configuration
6489 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6490 - fix ext3 htree duplicate directory entry corruption (1516)
6491 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6492 - update to new LOV EA format (2097)
6493 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6495 2003-06-15 Phil Schwan <phil@clusterfs.com>
6498 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6499 - per-import/export recovery handling (958, 931, 959)
6500 - multiple last-rcvd slots, for serving multiple FSes (949)
6501 - connections are again shared between multiple imp/exports (963, 964)
6502 - "umount -f" would hang if any requests needed to be sent (393, 978)
6503 - avoid pinning large req buffer by copying for queued messages (989)
6504 - add "uuid" to "lctl device" command to help upcalls (991)
6505 - "open" RPCs with transnos would confuse recovery counters (1037)
6506 - do proper endian conversion of all wire messages (288, 340, 891)
6507 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6508 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6509 - fix ptlrpc_connection leak in target_handle_connect (1174)
6510 - fix import refcounting bug in OST and MDS cleanup (1134)
6511 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6512 - fix very unlikely obd_types race condition (501)
6513 - remove osc_open hack for echo_client (1187)
6514 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6515 - a failure in read_inode2 leads to deadlock (1139)
6516 - cancel ack-locks as soon as transaction is committed (1072)
6517 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6518 - make sure to commitrw after any preprw to avoid deadlock (1162)
6519 - failing to execute a file in a lustre FS would lock inode (1203)
6520 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6521 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6522 - obd_uuid2tgt would walk off the end of the list (1255)
6523 - on IA64 the osc would give portals incorrect bulk size (1258)
6524 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6525 - fix lock inversion caused by new llite matching code (1282)
6526 - limit the number of dirty pages on a client to 10MB (1286)
6527 - timed out locks were not being corrected cancelled (1289)
6528 - fix O_DIRECT above 4GB on IA-32 (1292)
6529 * major user-visible changes
6530 - fail out/fail over policy now controlled by the upcall (993)
6532 - add OBD_PING to check server availability and failure (954)
6533 - lustre messages are now sent in sending host order (288, 340, 891)
6534 - add eadatalen to MDS getattr reply (340)
6535 - OST read replies may contain second buffer, with per-page status (593)
6537 2003-03-11 Phil Schwan <phil@clusterfs.com>
6540 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6541 - failed open()s could cause deadlock; fixed (867, 869)
6542 - stop cancelling OST locks when files are closed (481)
6543 - overlapping XID spaces caused network corruption (851, 853)
6544 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6545 - setattr_raw added, to do single-RPC, server-side setattrs
6546 - lmc/lconf syntax change for OST UUIDs
6547 - fix crashy race condition between ptlrpc_free_req and osc_close
6548 - don't use request in mdc_enqueue if we hit a timeout (889)
6549 - don't set the inode i_size for regular files from the MDS (896)
6550 - handle out of order completion AST (842)
6551 - don't LBUG if a lock request times out after receiving AST (913)
6552 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6553 - if a bad lock AST arrives, send an error instead of dropping entirely
6554 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6555 - fix leak in bulk IO when only partially completed (899, 900, 926)
6556 - fix O_DIRECT for ia64 (55)
6557 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6558 - C-z after timeout could hang a process forever; fixed (977)
6560 - client-side I/O cache (678, 924, 929, 941, 970)
6562 - READPAGE and SETATTRs which don't take server-side locks get
6565 2003-02-11 Phil Schwan <phil@clusterfs.com>
6568 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6569 - Fix fsfilt_extN_readpage() to read a full page of directory
6570 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6571 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6572 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6573 - Most intent operations were removed, in favour of a new RPC mode
6574 that does a single RPC to the server and bypasses most of the VFS
6575 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6576 - Aggressively cancel local locks on DLM servers
6577 - mds_reint_unlink sends EA to the client if it's the last nlink.
6578 client uses that EA to unlink OST objects.
6579 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6580 - recursive symlinks were fixed (439)
6581 - fixed NULL deref in DEBUG_REQ
6582 - filter_update_lastobjid no longer calls sync, which annoyed extN
6583 - fixed multi-client small-writes to a single file problem (445)
6584 - fixed mtime updates during file writes (607)
6585 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6586 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6587 - fixed Linux OST size reporting problem (444, 656)
6588 - OST now updates object mtime with writes or setattr (607, 619)
6589 - client verifies file size before zeroing page past EOF (445)
6590 - OST now writes last allocated objid to disk with allocation (108)
6591 - LOV on echo now works (409)
6593 - mds_reint_unlink sends a new buffer, with the EA included. this
6594 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6595 set if a regular file was being unlinked, and it was the last link
6596 - use PtlGet from the target for bulk writes (315)
6597 - OST now updates object mtime with writes or setattr (607, 619)
6598 - LDLM now has a grant-time callback to revalidate locked items, if
6600 - Many MDS operations were reorganized to combat race conditions
6602 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6603 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6605 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6608 - Fully reactivate OST imports after reconnection (512, others)
6609 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6610 - More graceful error handling for truncating on dead OST (515)
6611 - Don't error out unless we're actually accessing dead stripes (474)
6612 - Fix garbage sizes when stripes are missing (410)
6613 - LRU counters were broken, causing constant lock purge (433, 432)
6614 - garbage on read from stripes with failed OSTs (441)
6615 - mark OSCs as active before reconnecting during recovery (438)
6616 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6617 - lfind did not preserve OST order in output (443)
6618 - symlinks cause hung clients, incorrect data (439)
6619 - stop dereferencing request after dropping refcount (457)
6620 - don't LASSERT(spin_is_locked) on non-SMP (455)
6621 - fixes for many rename() bugs
6622 - fstat didn't correctly synchronize attributes (399)
6623 - server must handle lock cancellation during blocking AST prep (487)
6624 - bulk descriptors were free()d too soon (511)
6625 - fix paths in lconf, which would load incorrect modules (451, 507)
6626 - fix confusing lconf 'host not found' error message (386)
6627 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6628 - fix race condition in mdc_blocking_ast() for inode access (526)
6629 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6630 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6631 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6632 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6633 - fix lstripe in several ways (kernel side) (527)
6634 - fix request leak in ldlm_cli_enqueue (262)
6635 - incorrect OSC was marked inactive after OST failure
6636 - call mds_fs_cleanup before unmounting filesystem (524)
6637 - fix races between taking ns_lock and ldlm_lock_change_resource
6638 - fix races updating LOV export open file list
6639 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6640 - fix recovery NULL deref in ldlm_cli_cancel_unused
6641 - fix some DLM races by using new hash table for lock handles (419)
6642 - permit the client to specify desired inodes, at replay
6643 - duplicate requests when we queue them for replay reintegration
6644 - fix last_rcvd offset calculation
6645 - sync after each recovered transaction, so we always make progress
6646 - never, not always, ERESTART requests without transnos
6647 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6649 - skip replay if the MDS says that the client is already connected
6650 - don't check for a recovery-enabled export to match lctl's UUID
6651 - don't INC_USE_COUNT for phantom exports
6652 - don't crash when cleaning up phantom exports (567)
6653 - don't double-finish or set replay data for errored mdc_open requests
6654 - abort requests when they time out, so we don't get old replies
6655 - send/receive replies for AST messages again
6656 - if the client says that it doesn't have the lock, cancel it on the
6658 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6659 mark it as destroyed, it will all work out when decref is called
6660 - fix module use counts (22, 581)
6662 - ASTs now expect a reply (server cancels lock on error reply)
6664 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6667 - fix many simultaneous client startup (392)
6668 - fix dentry->d_it clobbering
6669 - credentials weren't being shipped for readdir/getattr operations
6670 - remove invalid assertions triggered during some concurrent MD
6672 - proper Lustre versions added (336, 389)
6673 - fix memory leak for create error case (398)
6674 - fix LOV locking bug that would get cli/srv out of sync
6675 - fix echo client over LOV (409)
6676 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6677 - fix double-O_EXCL intent crash (424)
6678 - avoid sending multiple lock CANCELs (352)
6680 - MDS can do multi-client recovery (modulo bugs in new code)
6682 - many updates, edits, cleanups
6684 2002-11-18 Phil Schwan <phil@clusterfs.com>
6687 - fix null d_it dereference (346)
6688 - fix full OST/dbench hang (333)
6689 - fix permission problem with file removal (286)
6690 - fix removal of OSCs from LOV when they fail
6691 - fix NULL deref during bulk timeout (214)
6692 - fix problems related to multiple filesystems on one MDS (241)
6693 - fixed serious subtle metadata locking bugs
6694 - free locks on clients when inodes are removed due to memory
6696 - fix inode pointer in lock data (285)
6697 - partial support for multiple MDS on a single host (241)
6698 - data locks weren't cancelled at clear_inode time (290, 311)
6699 - intent locks could lead to unbounded lock growth (205)
6700 - added a maximum lock count, an LRU list, and a flusher
6701 - fix multiple rename (365)
6702 - properly abstracted the echo client
6703 - OSC locked 1 byte too many; fixed
6704 - rewrote brw callback code:
6705 - fixed recovery bugs related to LOVs (306)
6706 - fixed too-many-pages-in-one-write crash (191)
6707 - fixed (again) crash in sync_io_timeout (214)
6708 - probably fixed callback-related race (385)
6710 - Add capability to MDS protocol
6711 - LDLM cancellations and callbacks on different portals
6713 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6716 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6717 - defer open object destruction to close time (601981, 138)
6718 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6719 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6720 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6721 - don't call obd_setattr in ll_file_release for destroyed objects
6722 * protocol change to lustre_msg: move |version| and add |flags|
6723 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6724 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6725 * added replay of create, unlink, link and rename operations during
6726 MDS failover; recovery should be much more robust now
6727 * remove failed OSCs from LOVs (only lov_create uses this so far)
6728 * the lustre-HOWTO was brought (more) up to date (582544)
6730 2002-10-23 Phil Schwan <phil@clusterfs.com>
6733 - in-use dentries weren't being reused properly (617851)
6734 - prevent multiple LDLM setup (599178)
6735 - fix LOV size calculations for truncate (617853)
6736 - fix client handling of MDS intent errors (POSIX)
6737 - fix permission bug in lovstripe.c test (624321)
6738 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6739 - truncate past end of file could corrupt data
6740 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6741 - a race in recovery could return ETIMEDOUT to apps (623947)
6742 - building outside the source directory was fixed
6743 * the lustre-HOWTO was brought (more) up to date (582544)
6744 * major progress was made on recovery functionality
6746 2002-10-10 Phil Schwan <phil@clusterfs.com>
6749 - recovery deadlock fix
6750 - rm -rf causes LBUG fix (617817)
6751 - file open by multiple tasks fix (618962)
6752 - directory permissions bugs (602707 and 620007)
6753 - journal_stop fixed with locking (611313)
6754 - O_APPEND failures resolved (618273, perhaps 614459)
6755 - lconf PATH fix (619770)
6756 - IA64 build fix (621450)
6757 - RPC buffer sizes scale with amount of memory
6759 2002-10-01 Phil Schwan <phil@clusterfs.com>
6762 - locks would be cancelled without throwing away data pages,
6763 resulting in inconsistent data (605627)
6764 - inode attributes were not always being refreshed (605627, 612449)
6765 - lconf now continues to cleanup after lctl reports an error
6766 - MDS now enforces user permissions (602707)
6767 - lprocfs cleanup fixed, but not yet enabled (614157)
6768 - fixed infinite server hang, should a client not respond to an AST
6769 - avoid going into recovery if user calls readlink() with a buffer
6770 that's too small (613941)
6771 - AST RPCs no longer require replies (614867) -- this may be changed
6772 - don't crash server if client sends an IOV that's too big (611336)
6773 - fixed lock conversion deadlock (611892)
6774 - fixed the following of symlinks (614622)
6775 * recovery: the server can remove locks from a client that dies, other
6776 clients can make progress
6777 * more extN patch fixes
6778 * compile-time configurable ptlrpc buffer allocations
6780 - collaborative read cache document
6781 - Lustre Lite Performance CDR document-in-progress
6783 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6786 - fix typo in patch-2.4.18
6788 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6791 - clear ptlrpc request each time in handle_incoming_request()
6792 - unlink of files now destroys the object on the OST
6794 2002-09-19 Peter Braam <braam@clusterfs.com>
6796 * add hard link support
6797 * change obdfile creation method
6798 * kernel patch changed
6800 2002-09-19 Peter Braam <braam@clusterfs.com>
6803 - stack overflow bug in extN fixed
6805 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6807 * documentation updates
6808 - add man pages for config tools
6809 - update tests/README to describe testing with new config tools
6810 - finish metadata API descriptions
6811 * bug fixes and cleanups
6812 - statfs workaround for 16TB limit
6813 - LOV stripe allocation improved, can stripe on subset of OSTs
6814 - LOV file size/IO offset was wrong for files > 4GB in size
6815 - object EA data was being dropped, caused files to be unreadable
6816 - memory overflow with non-LOV OST caused memory corruption
6817 - fixed regression tests to work with new config tools, obdfilter
6818 - fixed bug when directory size became larger than 1 block
6819 - fixed bug (for single client case) when PWD was deleted
6820 - invalidate local directory pages when doing intent-based ops
6821 - avoid LDLM oops when lock callback contained bad data
6823 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6825 * documentation updates
6826 * bug fixes and cleanups
6827 - configuration tools
6830 - 64-bit compile warnings
6831 - 64-bit internal statfs data
6833 * test_brw on persistent OST devices
6835 * lprocfs (disabled)
6837 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6839 * documentation updates
6840 * bug fixes and cleanups
6841 * configuration tools
6843 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6846 * many small fixes to 0_5_4
6847 * io/network handling
6848 * thinkos in MDS operations
6850 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6853 * crucial basic fixes to 0.5.3
6854 * IOR, Iozone work over Elan
6857 2002-08-07 Phil Schwan <phil@clusterfs.com>
6858 * version 0_5_3, our first alpha
6859 * we use the new Portals iovs
6860 * documentation updates
6861 * bug fixes and cleanups
6862 * small changes in the DLM wire protocol
6864 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6865 * version 0_5_1 with some initial stability,
6866 * locking on MD and file I/O.
6867 * documentation updates
6868 * several bug fixes since 0.5.0
6869 * small changes in wire protocol
6871 2002-07-18 Phil Schwan <phil@clusterfs.com>
6873 * delivered as Lustre Light Alpha
6874 * fixed a crash after handling invalid MDS requests
6875 * fixed directory pages for architectures with non-4k pages sizes
6877 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6878 * release version v0_4_4
6879 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6881 * Fixes a number of interruption problems with OST operations.
6882 * Update documentation for portals header changes
6883 * Move all wire protocol structs/defines to lustre_idl.h
6884 * Fixes symlink length bug.
6885 * Add tcpdump to repository.
6887 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6888 * release version v0_4_3
6889 * Fixes statfs for inodes on extN.
6890 * Fixes bug in runtests which would delete /etc/hosts.
6891 * Use 64-bit object IDs wherever possible (not into VFS though)
6892 Remove ost_get_info, which is unused by lustre, and out of date.
6894 2002-07-03 Peter Braam <braam@clusterfs.com>
6895 * release version v0_4_2 Fixes a lookup error (type not passed)
6896 * move forward to head of Portals
6897 * move forward to latest Lustre kernel
6899 2002-06-25 Peter Braam <braam@clusterfs.com>
6900 * release version v0_4_1. Hopefully stable on single node use.