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 5.14.0-503.31.1.el9 (RHEL9.5)
7 5.14.0-427.42.1.el9 (RHEL9.4)
8 5.14.0-362.24.1.el9 (RHEL9.3)
9 4.18.0-553.46.1.el8 (RHEL8.10)
10 4.18.0-513.24.1.el8 (RHEL8.9)
11 4.18.0-477.27.1.el8 (RHEL8.8)
12 * Other server kernels known to build and work at some point (others may also work):
13 3.10.0-862.14.4.el7 (RHEL7.5)
14 3.10.0-957.27.2.el7 (RHEL7.6)
15 3.10.0-1062.18.1.el7 (RHEL7.7)
16 3.10.0-1127.19.1.el7 (RHEL7.8)
17 3.10.0-1160.88.1.el7 (RHEL7.9)
18 4.18.0-80.11.2.el8 (RHEL8.0)
19 4.18.0-147.8.1.el8 (RHEL8.1)
20 4.18.0-193.28.1.el8 (RHEL8.2)
21 4.18.0-240.22.1.el8 (RHEL8.3)
22 4.18.0-305.25.1.el8 (RHEL8.4)
23 4.18.0-348.23.1.el8 (RHEL8.5)
24 4.18.0-372.32.1.el8 (RHEL8.6)
25 4.18.0-425.10.1.el8 (RHEL8.7)
26 5.14.0-162.23.1.el9 (RHEL9.1)
27 5.14.0-284.30.1.el9 (RHEL9.2)
28 4.4.120-92.70 (SLES12 SP2)
29 4.4.180-94.100 (SLES12 SP3)
30 5.14.21-150500.55.65 (SLES15 SP5)
31 4.4.0-131 (Ubuntu 16.04)
32 4.15.0-32 (Ubuntu 18.04)
33 5.4.0-48 (Ubuntu 20.04)
34 6.8.0-38 (Ubuntu 24.04)
35 6.10.0-15 (Ubuntu 24.04)
36 vanilla linux 5.4.0 (ZFS + ldiskfs)
37 vanilla linux 5.4.21 (ZFS + ldiskfs)
38 vanilla linux 5.4.136 (ZFS + ldiskfs)
39 vanilla linux 6.1.36 (ZFS + ldiskfs)
40 vanilla linux 6.6.13 (ZFS + ldiskfs)
41 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
42 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
43 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
44 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
45 * Client primary kernels built and tested during release cycle:
46 5.14.0-503.31.1.el9 (RHEL9.5)
47 5.14.0-427.42.1.el9 (RHEL9.4)
48 5.14.0-362.24.1.el9 (RHEL9.3)
49 4.18.0-553.46.1.el8 (RHEL8.10)
50 4.18.0-513.24.1.el8 (RHEL8.9)
51 4.18.0-477.27.1.el8 (RHEL8.8)
52 5.15.0-88 (Ubuntu 22.04)
53 6.8.0-35 (Ubuntu 24.04)
54 6.4.0-150600.23.38 (SLES15 SP6)
55 5.14.21-150500.55.73 (SLES15 SP5)
56 * Other clients known to build on these kernels at some point (others may also work):
57 3.10.0-862.14.4.el7 (RHEL7.5)
58 3.10.0-957.27.2.el7 (RHEL7.6)
59 4.14.0-49.13.1.el7a (RHEL7.5)
60 4.14.0-115.2.2.el7a (RHEL7.6)
61 3.10.0-1062.18.1.el7 (RHEL7.7)
62 3.10.0-1127.19.1.el7 (RHEL7.8)
63 3.10.0-1160.88.1.el7 (RHEL7.9)
64 4.18.0-80.11.2.el8 (RHEL8.0)
65 4.18.0-147.8.1.el8 (RHEL8.1)
66 4.18.0-193.28.1.el8 (RHEL8.2)
67 4.18.0-240.22.1.el8 (RHEL8.3)
68 4.18.0-305.25.1.el8 (RHEL8.4)
69 4.18.0-348.23.1.el8 (RHEL8.5)
70 4.18.0-372.32.1.el8 (RHEL8.6)
71 4.18.0-425.10.1.el8 (RHEL8.7)
72 5.14.0-70.30.1.el9 (RHEL9.0)
73 5.14.0-162.23.1.el9 (RHEL9.1)
74 5.14.0-284.30.1.el9 (RHEL9.2)
75 4.4.120-92.70 (SLES12 SP2)
76 4.4.180-94.100 (SLES12 SP3)
77 4.12.14-95.48 (SLES12 SP4)
78 4.12.14-122.133 (SLES12 SP5)
79 4.12.14-197.75 (SLES15 SP1)
80 5.3.18-24.96 (SLES15 SP2)
81 5.3.18-150300.59.93 (SLES15 SP3)
82 5.14.21-150400.24.100 (SLES15 SP4)
83 4.4.0-131 (Ubuntu 16.04)
84 4.15.0-48 (Ubuntu 18.04)
85 5.4.0-37 (Ubuntu 20.04)
86 5.8.0-53 (Ubuntu 20.04.2 HWE)
87 5.11.0-31 (Ubuntu 20.04.3 HWE)
88 5.11.0 (vanilla kernel.org)
89 5.10.0 (openEuler 22.03 LTS)
90 * Recommended e2fsprogs version: 1.47.1-wc1 or newer
91 * Recommended ZFS version: 2.2.6
92 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
93 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
94 stack overflow. For more information, please refer to bugzilla 17630.
95 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
96 the re-exporting nodes to be mounted with "32bitapi" mount option
98 --------------------------------------------------------------------------------
103 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
104 for currently supported client and server kernel versions.
105 * Server primary kernels built and tested during release cycle:
106 5.14.0-427.31.1.el9 (RHEL9.4)
107 5.14.0-362.24.1.el9 (RHEL9.3)
108 4.18.0-553.16.1.el8 (RHEL8.10)
109 4.18.0-513.24.1.el8 (RHEL8.9)
110 4.18.0-477.27.1.el8 (RHEL8.8)
111 * Other server kernels known to build and work at some point (others may also work):
112 3.10.0-862.14.4.el7 (RHEL7.5)
113 3.10.0-957.27.2.el7 (RHEL7.6)
114 3.10.0-1062.18.1.el7 (RHEL7.7)
115 3.10.0-1127.19.1.el7 (RHEL7.8)
116 3.10.0-1160.88.1.el7 (RHEL7.9)
117 4.18.0-80.11.2.el8 (RHEL8.0)
118 4.18.0-147.8.1.el8 (RHEL8.1)
119 4.18.0-193.28.1.el8 (RHEL8.2)
120 4.18.0-240.22.1.el8 (RHEL8.3)
121 4.18.0-305.25.1.el8 (RHEL8.4)
122 4.18.0-348.23.1.el8 (RHEL8.5)
123 4.18.0-372.32.1.el8 (RHEL8.6)
124 4.18.0-425.10.1.el8 (RHEL8.7)
125 5.14.0-162.23.1.el9 (RHEL9.1)
126 5.14.0-284.30.1.el9 (RHEL9.2)
127 4.4.120-92.70 (SLES12 SP2)
128 4.4.180-94.100 (SLES12 SP3)
129 5.14.21-150500.55.65 (SLES15 SP5)
130 4.4.0-131 (Ubuntu 16.04)
131 4.15.0-32 (Ubuntu 18.04)
132 5.4.0-48 (Ubuntu 20.04)
133 6.8.0-38 (Ubuntu 24.04)
134 6.10.0-15 (Ubuntu 24.04)
135 vanilla linux 5.4.0 (ZFS + ldiskfs)
136 vanilla linux 5.4.21 (ZFS + ldiskfs)
137 vanilla linux 5.4.136 (ZFS + ldiskfs)
138 vanilla linux 6.1.36 (ZFS + ldiskfs)
139 vanilla linux 6.6.13 (ZFS + ldiskfs)
140 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
141 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
142 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
143 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
144 * Client primary kernels built and tested during release cycle:
145 5.14.0-427.31.1.el9 (RHEL9.4)
146 5.14.0-362.24.1.el9 (RHEL9.3)
147 4.18.0-553.16.1.el8 (RHEL8.10)
148 4.18.0-513.24.1.el8 (RHEL8.9)
149 4.18.0-477.27.1.el8 (RHEL8.8)
150 5.4.0-37 (Ubuntu 20.04)
151 6.4.0-150600.23.17 (SLES15 SP6)
152 5.14.21-150500.55.73 (SLES15 SP5)
153 * Other clients known to build on these kernels at some point (others may also work):
154 3.10.0-862.14.4.el7 (RHEL7.5)
155 3.10.0-957.27.2.el7 (RHEL7.6)
156 4.14.0-49.13.1.el7a (RHEL7.5)
157 4.14.0-115.2.2.el7a (RHEL7.6)
158 3.10.0-1062.18.1.el7 (RHEL7.7)
159 3.10.0-1127.19.1.el7 (RHEL7.8)
160 3.10.0-1160.88.1.el7 (RHEL7.9)
161 4.18.0-80.11.2.el8 (RHEL8.0)
162 4.18.0-147.8.1.el8 (RHEL8.1)
163 4.18.0-193.28.1.el8 (RHEL8.2)
164 4.18.0-240.22.1.el8 (RHEL8.3)
165 4.18.0-305.25.1.el8 (RHEL8.4)
166 4.18.0-348.23.1.el8 (RHEL8.5)
167 4.18.0-372.32.1.el8 (RHEL8.6)
168 4.18.0-425.10.1.el8 (RHEL8.7)
169 5.14.0-70.30.1.el9 (RHEL9.0)
170 5.14.0-162.23.1.el9 (RHEL9.1)
171 5.14.0-284.30.1.el9 (RHEL9.2)
172 4.4.120-92.70 (SLES12 SP2)
173 4.4.180-94.100 (SLES12 SP3)
174 4.12.14-95.48 (SLES12 SP4)
175 4.12.14-122.133 (SLES12 SP5)
176 4.12.14-197.75 (SLES15 SP1)
177 5.3.18-24.96 (SLES15 SP2)
178 5.3.18-150300.59.93 (SLES15 SP3)
179 5.14.21-150400.24.100 (SLES15 SP4)
180 4.4.0-131 (Ubuntu 16.04)
181 4.15.0-48 (Ubuntu 18.04)
182 5.8.0-53 (Ubuntu 20.04.2 HWE)
183 5.11.0-31 (Ubuntu 20.04.3 HWE)
184 5.11.0 (vanilla kernel.org)
185 5.10.0 (openEuler 22.03 LTS)
186 * Recommended e2fsprogs version: 1.47.1-wc1 or newer
187 * Recommended ZFS version: 2.1.15
188 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
189 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
190 stack overflow. For more information, please refer to bugzilla 17630.
191 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
192 the re-exporting nodes to be mounted with "32bitapi" mount option
194 --------------------------------------------------------------------------------
199 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
200 for currently supported client and server kernel versions.
201 * Server primary kernels built and tested during release cycle:
202 4.18.0-348.2.1.el8 (RHEL8.5)
203 * Other server kernels known to build and work at some point (others may also work):
204 3.10.0-862.14.4.el7 (RHEL7.5)
205 3.10.0-957.27.2.el7 (RHEL7.6)
206 3.10.0-1062.18.1.el7 (RHEL7.7)
207 3.10.0-1127.19.1.el7 (RHEL7.8)
208 3.10.0-1160.49.1.el7 (RHEL7.9)
209 4.18.0-80.11.2.el8 (RHEL8.0)
210 4.18.0-147.8.1.el8 (RHEL8.1)
211 4.18.0-193.28.1.el8 (RHEL8.2)
212 4.18.0-240.22.1.el8 (RHEL8.3)
213 4.18.0-305.25.1.el8 (RHEL8.4)
214 4.4.120-92.70 (SLES12 SP2)
215 4.4.180-94.100 (SLES12 SP3)
216 4.4.0-131 (Ubuntu 16.04)
217 4.15.0-32 (Ubuntu 18.04)
218 5.4.0-48 (Ubuntu 20.04)
219 vanilla linux 5.4.0 (ZFS + ldiskfs)
220 vanilla linux 5.4.21 (ZFS + ldiskfs)
221 vanilla linux 5.4.136 (ZFS + ldiskfs)
222 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
223 * Client primary kernels built and tested during release cycle:
224 4.18.0-348.2.1.el8 (RHEL8.5)
225 5.4.0-37 (Ubuntu 20.04)
226 5.3.18-59.27 (SLES15 SP3)
227 * Other clients known to build on these kernels at some point (others may also work):
228 3.10.0-862.14.4.el7 (RHEL7.5)
229 3.10.0-957.27.2.el7 (RHEL7.6)
230 4.14.0-49.13.1.el7a (RHEL7.5)
231 4.14.0-115.2.2.el7a (RHEL7.6)
232 3.10.0-1062.18.1.el7 (RHEL7.7)
233 3.10.0-1127.19.1.el7 (RHEL7.8)
234 3.10.0-1160.49.1.el7 (RHEL7.9)
235 4.18.0-80.11.2.el8 (RHEL8.0)
236 4.18.0-147.8.1.el8 (RHEL8.1)
237 4.18.0-193.28.1.el8 (RHEL8.2)
238 4.18.0-240.22.1.el8 (RHEL8.3)
239 4.18.0-305.25.1.el8 (RHEL8.4)
240 4.4.120-92.70 (SLES12 SP2)
241 4.4.180-94.100 (SLES12 SP3)
242 4.12.14-95.48 (SLES12 SP4)
243 4.12.14-122.91 (SLES12 SP5)
244 4.12.14-197.75 (SLES15 SP1)
245 5.3.18-24.96 (SLES15 SP2)
246 4.4.0-131 (Ubuntu 16.04)
247 4.15.0-48 (Ubuntu 18.04)
248 5.8.0-53 (Ubuntu 20.04.2 HWE)
249 5.11.0-31 (Ubuntu 20.04.3 HWE)
250 5.11.0 (vanilla kernel.org)
251 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
252 * Recommended ZFS version: 2.0.7
253 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
254 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
255 stack overflow. For more information, please refer to bugzilla 17630.
256 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
257 the re-exporting nodes to be mounted with "32bitapi" mount option
259 --------------------------------------------------------------------------------
263 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
264 for currently supported client and server kernel versions.
265 * Server primary kernels built and tested during release cycle:
266 3.10.0-1062.18.1.el7 (RHEL7.7)
267 3.10.0-1127.19.1.el7 (RHEL7.8)
268 3.10.0-1160.6.1.el7 (RHEL7.9)
269 * Other server kernels known to build and work at some point (others may also work):
270 3.10.0-862.14.4.el7 (RHEL7.5)
271 3.10.0-957.27.2.el7 (RHEL7.6)
272 4.18.0-80.11.2.el8 (RHEL8.0)
273 4.18.0-147.8.1.el8 (RHEL8.1)
274 4.18.0-193.28.1.el8 (RHEL8.2)
275 4.18.0-240.1.1.el8 (RHEL8.3)
276 4.4.120-92.70 (SLES12 SP2)
277 4.4.180-94.100 (SLES12 SP3)
278 4.4.0-131 (Ubuntu 16.04)
279 4.15.0-32 (Ubuntu 18.04)
280 5.4.0-48 (Ubuntu 20.04)
281 vanilla linux 5.4.0 (ZFS + ldiskfs)
282 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
283 * Client primary kernels built and tested during release cycle:
284 3.10.0-1062.18.1.el7 (RHEL7.7)
285 3.10.0-1127.19.1.el7 (RHEL7.8)
286 3.10.0-1160.6.1.el7 (RHEL7.9)
287 4.12.14-95.48 (SLES12 SP4)
288 4.12.14-122.20 (SLES12 SP5)
289 4.12.14-197.75 (SLES15 SP1)
290 5.3.18-24.24 (SLES15 SP2)
291 4.15.0-48 (Ubuntu 18.04)
292 * Other clients known to build on these kernels at some point (others may also work):
293 3.10.0-862.14.4.el7 (RHEL7.5)
294 3.10.0-957.27.2.el7 (RHEL7.6)
295 4.14.0-49.13.1.el7a (RHEL7.5)
296 4.14.0-115.2.2.el7a (RHEL7.6)
297 4.18.0-80.11.2.el8 (RHEL8.0)
298 4.18.0-147.8.1.el8 (RHEL8.1)
299 4.18.0-193.28.1.el8 (RHEL8.2)
300 4.18.0-240.1.1.el8 (RHEL8.3)
301 4.4.120-92.70 (SLES12 SP2)
302 4.4.180-94.100 (SLES12 SP3)
303 4.4.0-131 (Ubuntu 16.04)
304 5.4.0-37 (Ubuntu 20.04)
305 5.4.0 (vanilla kernel.org)
306 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
307 * Recommended ZFS version: 2.0.0
308 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
309 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
310 stack overflow. For more information, please refer to bugzilla 17630.
311 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
312 the re-exporting nodes to be mounted with "32bitapi" mount option
314 --------------------------------------------------------------------------------
318 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
319 for currently supported client and server kernel versions.
320 * Server primary kernels built and tested during release cycle:
321 3.10.0-1062.1.1.el7 (RHEL7.7)
322 * Other server kernels known to build and work at some point (others may also work):
323 3.10.0-862.14.4.el7 (RHEL7.5)
324 3.10.0-957.27.2.el7 (RHEL7.6)
325 4.18.0-80.7.1.el8 (RHEL8.0)
326 4.4.120-92.70 (SLES12 SP2)
327 4.4.180-94.100 (SLES12 SP3)
328 4.4.0-131 (Ubuntu 16.04)
329 4.15.0-32 (Ubuntu 18.04)
330 vanilla linux 4.6.7 (ZFS only)
331 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
332 * Client primary unpatched kernels built and tested during release cycle:
333 3.10.0-1062.1.1.el7 (RHEL7.7)
334 4.12.14-95.29 (SLES12 SP4)
335 4.15.0-48 (Ubuntu 18.04)
336 * Other clients known to build on unpatched kernels at some point (others may also work):
337 3.10.0-862.14.4.el7 (RHEL7.5)
338 3.10.0-957.27.2.el7 (RHEL7.6)
339 4.14.0-49.13.1.el7a (RHEL7.5)
340 4.14.0-115.2.2.el7a (RHEL7.6)
341 4.18.0-80.7.1.el8 (RHEL8.0)
342 4.4.120-92.70 (SLES12 SP2)
343 4.4.180-94.100 (SLES12 SP3)
344 4.4.0-131 (Ubuntu 16.04)
345 vanilla linux 4.15.0 (ZFS only)
346 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
347 * Recommended ZFS / SPL version: 0.7.13
348 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
349 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
350 stack overflow. For more information, please refer to bugzilla 17630.
351 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
352 the re-exporting nodes to be mounted with "32bitapi" mount option
353 * Note for Ubuntu 14.04.5:
354 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
355 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
357 --------------------------------------------------------------------------------
361 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
362 for currently supported client and server kernel versions.
363 * Server known to build on the following kernels (others may also work):
364 2.6.32-696.18.7.el6 (RHEL6.9)
365 3.10.0-862.14.4.el7 (RHEL7.5)
366 3.10.0-957.el7 (RHEL7.6)
367 4.4.120-92.70 (SLES12 SP2)
368 4.4.162-94.69 (SLES12 SP3)
369 4.4.0-131 (Ubuntu 16.04)
370 4.15.0-32 (Ubuntu 18.04)
371 vanilla linux 4.6.7 (ZFS only)
372 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
373 * Client known to build on unpatched kernels (others may also work):
374 2.6.32-696.18.7.el6 (RHEL6.9)
375 2.6.32-754.6.3.el6 (RHEL6.10)
376 3.10.0-862.14.4.el7 (RHEL7.5)
377 3.10.0-957.el7 (RHEL7.6)
378 4.14.0-49.13.1.el7a (RHEL7.5)
379 4.14.0-115.2.2.el7a (RHEL7.6)
380 4.4.120-92.70 (SLES12 SP2)
381 4.4.162-94.69 (SLES12 SP3)
382 4.4.0-131 (Ubuntu 16.04)
383 4.15.0-32 (Ubuntu 18.04)
384 vanilla linux 4.15.0 (ZFS only)
385 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
386 * Recommended ZFS / SPL version: 0.7.9
387 * Tested with ZFS / SPL version: 0.7.9
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
393 * Note for Ubuntu 14.04.5:
394 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
395 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
397 --------------------------------------------------------------------------------
399 2018-04-03 Intel Corporation
401 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
402 for currently supported client and server kernel versions.
403 * Server known to build on patched kernels:
404 2.6.32-431.29.2.el6 (RHEL6.5)
405 2.6.32-504.30.3.el6 (RHEL6.6)
406 2.6.32-573.26.1.el6 (RHEL6.7)
407 2.6.32-642.15.1.el6 (RHEL6.8)
408 2.6.32-696.18.7.el6 (RHEL6.9)
409 3.10.0-693.21.1.el7 (RHEL7.4)
410 3.0.101-0.47.71 (SLES11 SP3)
411 3.0.101-107 (SLES11 SP4)
412 3.12.74-60.64.40 (SLES12 SP1)
413 4.4.114-92.64 (SLES12 SP2)
414 4.4.114-94.11 (SLES12 SP3)
415 3.13.0-101 (Ubuntu 14.04, ZFS only)
416 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
417 4.4.0-116 (Ubuntu 16.04)
418 vanilla linux 4.6.7 (ZFS only)
419 * Client known to build on unpatched kernels:
420 2.6.32-431.29.2.el6 (RHEL6.5)
421 2.6.32-504.30.3.el6 (RHEL6.6)
422 2.6.32-573.26.1.el6 (RHEL6.7)
423 2.6.32-642.15.1.el6 (RHEL6.8)
424 2.6.32-696.18.7.el6 (RHEL6.9)
425 3.10.0-693.21.1.el7 (RHEL7.4)
426 3.0.101-0.47.71 (SLES11 SP3)
427 3.0.101-107 (SLES11 SP4)
428 3.12.74-60.64.40 (SLES12 SP1)
429 4.4.114-92.64 (SLES12 SP2)
430 4.4.114-94.11 (SLES12 SP3)
431 3.13.0-101 (Ubuntu 14.04)
432 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
433 4.4.0-116 (Ubuntu 16.04)
435 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
436 * Recommended ZFS / SPL version: 0.7.6
437 * Tested with ZFS / SPL version: 0.7.6
438 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
439 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
440 stack overflow. For more information, please refer to bugzilla 17630.
441 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
442 the re-exporting nodes to be mounted with "32bitapi" mount option
443 * Note for Ubuntu 14.04.5:
444 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
445 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
447 --------------------------------------------------------------------------------
449 2017-07-05 Intel Corporation
451 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
452 for currently supported client and server kernel versions.
453 * Server known to build on patched kernels:
454 2.6.32-431.29.2.el6 (RHEL6.5)
455 2.6.32-504.30.3.el6 (RHEL6.6)
456 2.6.32-573.26.1.el6 (RHEL6.7)
457 2.6.32-642.15.1.el6 (RHEL6.8)
458 2.6.32-696.3.1.el6 (RHEL6.9)
459 3.10.0-514.21.1.el7 (RHEL7.3)
460 3.0.101-0.47.71 (SLES11 SP3)
461 3.0.101-100 (SLES11 SP4)
462 3.12.74-60.64.40 (SLES12 SP1)
463 4.4.59-92.17 (SLES12 SP2)
464 vanilla linux 4.6.7 (ZFS only)
465 * Client known to build on unpatched kernels:
466 2.6.32-431.29.2.el6 (RHEL6.5)
467 2.6.32-504.30.3.el6 (RHEL6.6)
468 2.6.32-573.26.1.el6 (RHEL6.7)
469 2.6.32-642.15.1.el6 (RHEL6.8)
470 2.6.32-696.3.1.el6 (RHEL6.9)
471 3.10.0-514.21.1.el7 (RHEL7.3)
472 3.0.101-0.47.71 (SLES11 SP3)
473 3.0.101-100 (SLES11 SP4)
474 3.12.74-60.64.40 (SLES12 SP1)
475 4.4.59-92.17 (SLES12 SP2)
477 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
478 * Recommended ZFS / SPL version: 0.6.5.9
479 * Tested with ZFS / SPL version: 0.6.5.9
480 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
481 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
482 stack overflow. For more information, please refer to bugzilla 17630.
483 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
484 the re-exporting nodes to be mounted with "32bitapi" mount option
486 --------------------------------------------------------------------------------
488 12-07-2016 Intel Corporation
490 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
491 for currently supported client and server kernel versions.
492 * Server known to build on patched kernels:
493 2.6.32-431.29.2.el6 (RHEL6.5)
494 2.6.32-504.30.3.el6 (RHEL6.6)
495 2.6.32-573.26.1.el6 (RHEL6.7)
496 2.6.32-642.11.1.el6 (RHEL6.8)
497 3.10.0-514.el7 (RHEL7.3)
498 3.0.101-0.47.71 (SLES11 SP3)
499 3.0.101-80 (SLES11 SP4)
500 3.12.62-60.64.8 (SLES12 SP1)
501 vanilla linux 4.5.7 (ZFS only)
502 * Client known to build on unpatched kernels:
503 2.6.32-431.29.2.el6 (RHEL6.5)
504 2.6.32-504.30.3.el6 (RHEL6.6)
505 2.6.32-573.26.1.el6 (RHEL6.7)
506 2.6.32-642.11.1.el6 (RHEL6.8)
507 3.10.0-514.el7 (RHEL7.3)
508 3.0.101-0.47.71 (SLES11 SP3)
509 3.0.101-80 (SLES11 SP4)
510 3.12.62-60.64.8 (SLES12 SP1)
512 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
513 * Recommended ZFS / SPL version: 0.6.5.8
514 * Tested with ZFS / SPL version: 0.6.5.8
515 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
516 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
517 stack overflow. For more information, please refer to bugzilla 17630.
518 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
519 the re-exporting nodes to be mounted with "32bitapi" mount option
521 --------------------------------------------------------------------------------
523 02-29-2016 Intel Corporation
525 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
526 for currently supported client and server kernel versions.
527 * Server known to build on patched kernels:
528 2.6.32-431.29.2.el6 (RHEL6.5)
529 2.6.32-504.30.3.el6 (RHEL6.6)
530 2.6.32-573.12.1.el6 (RHEL6.7)
531 3.10.0-327.3.1.el7 (RHEL7.2)
532 3.0.101-0.47.71 (SLES11 SP3)
533 3.0.101-68 (SLES11 SP4)
534 vanilla linux 4.2.1 (ZFS only)
535 * Client known to build on unpatched kernels:
536 2.6.32-431.29.2.el6 (RHEL6.5)
537 2.6.32-504.30.3.el6 (RHEL6.6)
538 2.6.32-573.12.1.el6 (RHEL6.7)
539 3.10.0-327.3.1.el7 (RHEL7.2)
540 3.0.101-0.47.71 (SLES11 SP3)
541 3.0.101-68 (SLES11 SP4)
544 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
545 * Recommended ZFS / SPL version: 0.6.4.2
546 * Tested with ZFS / SPL version: 0.6.4.2
547 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
548 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
549 stack overflow. For more information, please refer to bugzilla 17630.
550 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
551 the re-exporting nodes to be mounted with "32bitapi" mount option
553 --------------------------------------------------------------------------------
555 03-10-2015 Intel Corporation
557 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
558 for currently supported client and server kernel versions.
559 * Server known to build on patched kernels:
560 2.6.32-431.29.2.el6 (RHEL6.5)
561 2.6.32-504.8.1.el6 (RHEL6.6)
562 3.0.101-0.46 (SLES11 SP3)
563 * Client known to build on unpatched kernels:
564 2.6.32-431.29.2.el6 (RHEL6.5)
565 2.6.32-504.8.1.el6 (RHEL6.6)
566 3.10.0-123.20.1.el7 (RHEL7)
567 3.0.101-0.46 (SLES11 SP3)
568 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
569 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
570 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
571 stack overflow. For more information, please refer to bugzilla 17630.
572 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
573 the re-exporting nodes to be mounted with "32bitapi" mount option
575 Severity : enhancement
577 Description: control OST-index in IDIF via ROCOMPAT flag.
578 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
579 last_rcvd file. For new formatted OST device, it will be auto set;
580 for the case of upgrading from old OST device, you can enable it
581 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
582 enabled, for new created OST-object, its IDIF-in-LMA will contain
583 the OST-index; for the existing OST-object, the OSD will convert
584 old format IDIF as new format IDIF with OST-index stored in the
585 LMA EA when accessing such OST-object or via OI scrub. Once such
586 flag is enabled, it cannot be reverted back, so the system cannot
587 be downgraded to the orignal incompatible version.
589 --------------------------------------------------------------------------------
591 07-30-2014 Intel Corporation
593 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
594 for currently supported client and server kernel versions.
595 * Server known to build on patched kernels:
596 2.6.32-431.20.3.el6 (RHEL6.5)
597 3.0.101-0.31 (SLES11 SP3)
598 * Client known to build on unpatched kernels:
599 2.6.32-431.20.3.el6 (RHEL6.5)
600 3.0.101-0.31 (SLES11 SP3)
603 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
604 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
605 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
606 stack overflow. For more information, please refer to bugzilla 17630.
607 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
608 the re-exporting nodes to be mounted with "32bitapi" mount option
610 --------------------------------------------------------------------------------
612 10-31-2013 Intel Corporation
614 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
615 * Server builds for kernels:
616 2.6.32-358.18.1.el6 (RHEL6.4)
617 * Client builds for unpatched kernels:
618 2.6.32-358.18.1.el6 (RHEL6.4)
619 3.0.92-0.8 (SLES11 SP3)
621 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
622 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
623 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
624 stack overflow. For more information, please refer to bugzilla 17630.
625 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
626 the re-exporting nodes to be mounted with "32bitapi" mount option
628 --------------------------------------------------------------------------------
630 05-30-2013 Intel Corporation
632 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
633 * Server builds for kernels:
634 2.6.32-358.6.2.el6 (RHEL6.4)
635 * Client builds for unpatched kernels:
636 2.6.32-358.6.2.el6 (RHEL6.4)
637 3.0.74-0.6.6 (SLES11 SP2)
639 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
640 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
641 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
642 stack overflow. For more information, please refer to bugzilla 17630.
643 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
644 the re-exporting nodes to be mounted with "32bitapi" mount option
646 --------------------------------------------------------------------------------
648 9-30-2012 Whamcloud, Inc.
650 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
651 * Server support for kernels:
652 2.6.32-279.5.1.el6 (RHEL6)
653 * Client support for unpatched kernels:
654 2.6.32-279.5.1.el6 (RHEL6)
656 2.6.18-238.19.1.el5 (RHEL5)
657 2.6.16.60-0.69.1 (SLES10)
658 * Recommended e2fsprogs version: 1.41.90.wc4
659 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
660 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
661 stack overflow. For more information, please refer to bugzilla 17630.
662 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
663 the re-exporting nodes to be mounted with "32bitapi" mount option
665 Severity : enhancement
667 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
669 Severity : enhancement
671 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
673 Severity : enhancement
675 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
677 Severity : enhancement
679 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
681 Severity : enhancement
683 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
685 --------------------------------------------------------------------------------
687 04-30-2012 Whamcloud, Inc.
689 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
690 * Server support for kernels:
691 2.6.32-220.4.2.el6 (RHEL6)
692 * Client support for unpatched kernels:
693 2.6.32-220.4.2.el6 (RHEL6)
695 2.6.18-238.19.1.el5 (RHEL5)
696 2.6.16.60-0.69.1 (SLES10)
697 * Recommended e2fsprogs version: 1.41.90.wc4
698 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
699 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
700 stack overflow. For more information, please refer to bugzilla 17630.
701 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
702 the re-exporting nodes to be mounted with "32bitapi" mount option
704 Severity : enhancement
706 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
708 Severity : enhancement
710 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
712 Severity : enhancement
714 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
716 Severity : enhancement
718 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
720 --------------------------------------------------------------------------------
722 09-30-2011 Whamcloud, Inc.
724 * Server support for kernels:
725 2.6.18-238.19.1.el5 (RHEL5)
726 2.6.32-131.6.1.el6 (RHEL6)
727 * Client support for unpatched kernels:
728 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
729 2.6.18-238.19.1.el5 (RHEL5)
730 2.6.32-131.6.1.el6 (RHEL6)
731 2.6.29.4-167.fc11 (FC11)
732 2.6.16.60-0.69.1 (SLES10)
734 * Recommended e2fsprogs version: 1.41.90.wc3
735 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
736 Lustre filesystem with 4K stack may cause a stack overflow. For more
737 information, please refer to bugzilla 17630.
738 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
739 mounted with "32bitapi" mount option
741 Severity : enhancement
742 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
744 Severity : enhancement
745 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
747 Severity : enhancement
748 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
751 Frequency : only with SLES10
753 Description: Use OFED "KMP" provided by Novell
754 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
755 Lustre is now built against this package. That means you need to
756 install the ofed-kmp package from Novell for the patchless client.
757 Note that the ofed-kmp that Novell ships may not exactly match the
758 kernel version but should still be compatible.
760 Severity : enhancement
762 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
763 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
765 --------------------------------------------------------------------------------
767 2010-07-15 Oracle, Inc.
769 * Server support for kernels:
770 2.6.18-164.11.1.el5 (RHEL 5)
771 2.6.18-164.11.1.0.1.el5 (OEL 5)
772 * Client support for unpatched kernels:
773 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
774 2.6.18-164.11.1.el5 (RHEL 5),
775 2.6.18-164.11.1.0.1.el5 (OEL 5)
776 2.6.16.60-0.42.8 (SLES 10),
778 2.6.29.4-167.fc11 (FC11)
779 * Recommended e2fsprogs version: 1.41.10-sun2
781 Severity : enhancement
783 Description: Distribute lfs_migrate script with lustre
784 Details : The lfs_migrate mentioned in the manual is now included in the
785 lustre tree in lustre/scripts and will distributed in lustre RPMs.
787 Severity : enhancement
789 Description: Failover nid list added to import proc file.
791 Severity : enhancement
793 Description: "som_preview" mount option required to use SOM
794 Details : The clients will need to use som_preview mount option
798 Description: Lustre does not do 1MB IOs to HW RAID
799 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
801 Severity : enhancement
803 Description: lfs quota output is not very convenient for awk/sed-parsing
804 Details : Some positions in lfs quota output table could be empty or
805 non-empty which made it hard to parse it with scripts, now
806 a dash is put instead of space where there is not supposed
811 Description: Update to OFED 1.5.1
812 Details : OFED 1.5.1 will be used in environments that don't use
813 vendor supplied OFED.
817 Description: Remove lnet. prefix from lctl params display
818 Details : Also renamed lnet.memused to lnet_memused to
823 Description: Use conf_param -d to remove permanant settings.
827 Description: Add list_param -R
828 Details : lctl list_param now supports recursive mode.
831 Bugzilla : 21829, 22312
832 Description: Fix llobdstat.
833 Details : 1. Now updating the write_bytes counter
834 2. Port script fixes from 1.8.
838 Description: decrease the usage of memory on clients.
839 Details : 1. On clients, recycle dentries and inodes unused.
840 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
845 Description: llobdstat fix and enhancement
846 Details : add a counter to set a limit to how many samples will be returned
847 fix a wildcard in the path to limit to obdfilter stats only
851 Description: lfs check only allowed for root
852 Details : Code cleanups and added a regresion test of non-root use of
857 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
858 Details : Fix hr->hr_index overload.
862 Description: Fix FIEMAP functionality.
866 Description: FC11 patchless client support added
870 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
871 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
875 Description: New acc-sm script and various test-framework enhancments.
879 Description: Previously landed patch was reverted.
883 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
887 Description: Add ldiskfs maxdirsize mount option.
891 Description: Fix mkfs.lustre support for 16TB LUNs.
895 Description: Print mmp_check_interval
896 Details : print mmp_check_interval and make it possible to abort
897 mount operation in case it takes too long.
901 Description: Add cascading_rw to lustre mpi tests.
905 Description: Store SOM data in EA.
906 Details : Also several SOM related cleanups.
910 Description: Add refcount to md_open_data
914 Description: Use refcount on lov_request_set to prevent memory corruption.
918 Description: Statahead: use parent dir's i_mutex for synchronization.
922 Description: loadgen improvements.
926 Description: Enable peer health detection in ksocklnd and ko2iblnd.
930 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
934 Description: Fix race on obd_devcie:md_stats
938 Description: Fix SLES 11 client builds.
939 Details : Resolved uninitalized pointer gcc warnings.
943 Description: Allow non-root access for "lfs check".
944 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
948 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
950 Severity : enhancement
952 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
953 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
957 Description: Sanity test_27p fails.
958 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
959 Sanity tests 27 were cleaned up and improved.
963 Description: Unify req format on client/servers
964 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
965 buffers and swabbers approach..
970 Description: drop join file stuff.
972 Severity : enhancement
974 Description: Add 2.6.27 and SLES11 (patchless client) support
978 Description: Fix in ptlrpc_expire_one_request() to print the signed time
981 Severity : enhancement
983 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
986 Severity : enhancement
988 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
990 Severity : enhancement
992 Description: correctly handle big reply message.
993 Details : send LNet event if reply is bigger than buffer and adjust this buffer
998 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
1000 Severity : enhancement
1002 Description: Add support for OEL5.
1006 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
1008 Severity : enhancement
1010 Description: "weak-modules" support
1011 Details : Implement "weak-modules" support which enables kernel modules
1012 to be used with any kernel that implements the same kABI. In
1013 order to achieve this modules are now installed in
1014 /lib/modules/$(uname -r)/updates/kernel on all distributions.
1018 Description: client could not reconnect to OST because of an active request.
1019 Details : The client now retries to reconnect to the same server, if a
1020 connect request failed with EBUSY or -EAGAIN. Test case 35b
1021 conf-sanity.sh has been created.
1025 Description: Don't allow make backward step on assignin osc next id.
1026 Details : race between allocation next id and ll_sync thread can be cause
1027 of set wrong osc next id and can be kill valid ost objects.
1031 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1035 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1039 Description: Error handling in osc_statfs_interpret() has been improved.
1040 Details : Check in osc_statfs_interpret() for EBADR.
1044 Description: Do not send statfs() requests to OSTs disabled by administrator.
1045 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1047 Severity : enhancement
1049 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
1051 Severity : enhancement
1053 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1055 Severity : enhancement
1057 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1061 Description: Improved cleanup procedure for conf-sanity test_46a.
1065 Description: Do not start bulk transfer, if client tried to reconnect.
1066 Details : Do not start bulk transfer, if client tried to reconnect,
1067 but failed due to active rpcs. Check for exp_abort_active_req
1068 has been implemented.
1072 Description: Prevent inconsistences between linux and lustre mount structures.
1073 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1074 Make the sleep interruptible.
1076 Severity : enhancement
1078 Description: provide server to client comms path
1079 Details : server to client communications path via new LDLM_SET_INFO rpc
1083 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1084 Details : ldiskfs_truncate calls grab_cache_page which may start page
1085 allocation under an open transaction. This may lead to
1086 calling prune_icache with consequent lustre reentrance.
1090 Description: send correctly lsm on open replay
1091 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1096 Description: enable client interface failover
1097 Details : When a child reconnects from another NID, properly update export
1098 nid hash position and ldlm reverse import.
1100 Severity : enhancement
1102 Description: Communicate OST degraded/readonly state via statfs to MDS
1103 Details : Flags in the statfs returned from OSTs indicate whether the
1104 OST is in a degraded RAID state, or if the filesystem has
1105 turned read-only after a filesystem error is detected.
1110 Description: don't panic if EPROTO was hit when reading symlink
1111 Details : correctly handling request reference in error cases.
1113 Severity : enhancement
1115 Description: Add LustreNetLink, a kernel-userspace communcation path.
1117 Severity : enhancement
1119 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1124 Description: access to llog context before init.
1125 Details : move handling CATALOGS file at osc layer and forbid access to llog
1126 context before init.
1129 Frequency : rare, only when client is writing to OST on local node
1131 Description: Avoid deadlock for local client writes
1132 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1133 memory freeing context. This allows OST threads to set the
1134 PF_MEMALLOC flag on task structures in order to allocate memory
1135 from reserved pools and complete IO.
1136 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1137 so that the OST threads generate memory pressure and allow
1138 inactive pages to be reclaimed.
1140 Severity : enhancement
1142 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1147 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1148 Details : move ldlm namespace creation in setup phase to avoid grab
1149 _lprocfs_lock with cli_sem held.
1153 Description: Temporarily disable grant shrink.
1154 Details : Disable the feature for debugging.
1157 Frequency : only when formatting test filesystems
1159 Description: Unable to run several mkfs.lustre on loop devices at the same time
1160 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1161 formatting. The solution is to proper handle the error.
1163 Severity : enhancement
1165 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1167 Severity : enhancement
1169 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1174 Description: resolve race between obd_disconnect and class_disconnect_exports
1175 Details : if obd_disconnect will be called to already disconnected export he
1176 forget release one reference and osc module can't unloaded.
1178 Severity : enhancement
1180 Description: Allow tuning service thread via /proc
1181 Details : For each service a new
1182 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1183 created that can be used to set min/max thread counts, and get the
1184 current number of running threads.
1189 Descriptoin: don't return error if have partially created objects for file.
1190 Details : lov_update_create_set uses set->set_success as index for created
1191 objects, so if some requests will be failed, they will have hole
1192 at end of array and qos_shrink_lsm for allocate correct lsm.
1194 Severity : enhancement
1196 Description: Update OFED support to 1.4.1
1201 Description: Reduce small size read RPC
1202 Details : Set read-ahead limit for every file and only do read-ahead when
1203 available read-ahead pages are bigger than 1M to avoid small size
1206 Severity : enhancement
1208 Description: move AT tunable parameters for more consistent usage
1209 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1211 Severity : enhancement
1213 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1214 Details : allow skip disconnected ost for send statfs request and hide error
1220 Description: fix racy locking of mballoc block bitmaps causing BUG
1221 Details : The locking of the mballoc buddy bitmap and the in-memory
1222 block bitmap was using two different spin locks in some
1223 cases. This made it possible to incorrectly access the
1224 mballoc bitmap while another process was modifying it,
1225 causing a sanity assertion to fail. While no on-disk corruption
1226 was reported, there was some risk of this happening.
1229 Frequency : rare, on llog test 6
1231 Descriptoin: don't allow connect to already connected import
1232 Details : allowing connect to already connected import is hide connecting problem.
1237 Descriptoin: allow kill process which wait statahead result
1238 Details : for some reasons 'ls' can stick in waiting result from statahead,
1239 in this case need way for kill this process.
1241 Severity : enhancement
1243 Description: Add state history info file, enhance import info file
1244 Details : Track import connection state changes in a new osc/mdc proc file;
1245 add overview-type data to the osc/mdc import proc file.
1247 Severity : enhancement
1249 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1251 Severity : enhancement
1253 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1258 Descriptoin: OST grant too much space to client even there are not enough space.
1259 Details : Client will shrink its grant cache to OST if there are no write
1260 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1261 this grant cache if there are already not enough avaible space
1262 (left_space < total_clients * 32M).
1265 Frequency : start MDS on uncleanly shutdowned MDS device
1267 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1268 Details : stay in waiting mds<>ost recovery finished produce random bugs
1269 due race between two ll_sync thread for one lov target. send
1270 ACTIVATE event only if connect realy finished and import have
1274 Frequency : rare, connect and disconnect target at same time
1276 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1277 Details : don't call obd_disconnect under lov_lock. this long time
1278 operation and can block ptlrpcd which answer to connect request.
1283 Descriptoin: don't lose wakeup for imp_recovery_waitq
1284 Details : recover_import_no_retry or invalidate_import and import_close can
1285 both sleep on imp_recovery_waitq, but we was send only one wakeup
1289 Frequency : always with long access acl
1291 Descriptoin: mds can't pack reply with long acl.
1292 Details : mds don't control size of acl but they limited by reint/getattr
1295 Severity : enhancement
1297 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1299 Severity : enhancement
1301 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1304 Frequency : start MDS on uncleanly shutdowned MDS device
1306 Descriptoin: aborting recovery hang on MDS
1307 Details : don't throttle destroy RPCs for the MDT.
1310 Frequency : on remount
1312 Description: external journal device not working after the remount
1313 Details : clear dev_rdonly flag for external journal devices in
1319 Description: shutdown vs evict race
1320 Details : client_disconnect_export vs connect request race.
1321 if client will evicted at this time - we start invalidate
1322 thread without referece to import and import can be freed
1326 Frequency : rare, need acl's on inode.
1328 Description: client can't handle ost additional correctly
1329 Details : if ost was added after client connected to mds client can have
1330 hit lnet_try_match_md ... to big messages to wide striped files.
1331 in this case need teach client to handle config events about add
1332 lov target and update client max ea size at that event.
1334 Severity : enhancement
1336 Description: Changelogs
1337 Details : Changelogs are a lightweight mechanism to track filesystem
1338 metadata and namespace changes. The changelog is recorded
1339 permanently on the MDTs, and is periodically "consumed" / purged
1340 when records are no longer needed.
1342 Severity : enhancement
1344 Description: compact fld format with extents
1345 Details : Store range of seq rather than every seq in FLD. Seq
1346 controller update FLD rather than clients. In Case of CMD, mdt0
1347 has FLD, all other metadata server act as non persistent proxy
1348 for FLD queries and cache fld entries in fld cache.
1353 Description: don't skip ost target if they assigned to file
1354 Details : Drop slow OSCs if we can, but not for requested start idx.
1355 This means "if OSC is slow and it is not the requested
1356 start OST, then it can be skipped, otherwise skip it only
1357 if it is inactive/recovering/out-of-space.
1361 Description: more cleanup in mds_lov
1362 Details : not send LOV EA under replay, we can't know about they size at this
1363 time. Don't allow client connect to mds before any ost connected,
1364 for avoid problems with LOV EA size and returning EIO to client.
1366 Severity : enhancement
1368 Description: Interoperability at server side (Disk interoperability)
1370 Severity : enhancement
1372 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1374 Severity : enhancement
1376 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1378 Severity : enhancement
1380 Description: New client IO stack (CLIO).
1382 Severity : enhancement
1384 Description: Commit on sharing. Eliminate inter-client dependencies between
1385 uncommitted transactions by doing transaction commits.
1386 Thereby clients may recovery independently.
1389 Frequency : Create a symlink file with a very long name
1391 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1392 Details : If there is no extra space in the request for early cancels,
1393 ldlm_req_handles_avail() returns 0 instead of a negative value.
1395 Severity : enhancement
1397 Description: Add /proc entry for import status
1398 Details : The mdc, osc, and mgc import directories now have
1399 an import directory that contains useful import data for debugging
1400 connection problems.
1402 Severity : enhancement
1404 Description: Re-disable certain /proc logging
1405 Details : Enable and disable client's offset_stats, extents_stats and
1406 extents_stats_per_process stats logging on the fly.
1409 Frequency : Only on FC kernels 2.6.22+
1411 Description: oops in statahead
1412 Details : Do not drop reference count for the dentry from VFS when lookup,
1413 VFS will do that by itself.
1415 Severity : enhancement
1417 Description: Generic /proc file permissions
1418 Details : Set /Proc file permissions in a more generic way to enable non-
1419 root users operate on some /proc files.
1423 Description: Hitting mdc_commit_close() ASSERTION
1424 Details : Properly handle request reference release in
1425 ll_release_openhandle().
1429 Description: quota recovery deadlock during mds failover
1430 Details : This patch includes att18982, att18236, att18237 in bz14840.
1432 1. fix osts hang when mds does failover with quotaon
1433 2. prevent watchdog storm when osts threads wait for the
1438 Frequency : only patchless client
1439 Description: add workaround for race between add/remove dentry from hash
1441 Severity : enhancement
1443 Description: Allow OST glimpses to return PW locks
1447 Description: LBUG when llog conf file is full
1448 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1452 Description: Prevent import from entering FULL state when server in recovery
1456 Description: service mount cannot take device name with ":"
1457 Details : Only when device name contains ":/" will mount treat it as
1463 Description: replace ptlrpcd with the statahead thread to interpret the async
1464 statahead RPC callback
1468 Frequency : on recovery
1469 Description: I/O failures after umount during fail back
1470 Details : if client reconnected to restarted server we need join to recovery
1471 instead of find server handler is changed and process self
1472 eviction with cancel all locks.
1474 Severity : enhancement
1476 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1478 Severity : enhancement
1480 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1482 Severity : enhancement
1484 Description: Upcall on Lustre log has been dumped
1485 Details : Allow for a user mode script to be called once a Lustre log has
1486 been dumped. It passes the filename of the dumped log to the
1487 script, the location of the script can be specified via
1488 /proc/sys/lnet/debug_log_upcall.
1493 Description: avoid idr_remove called for id which is not allocated.
1494 Details : Move assigment s_dev for clustered nfs to end of initialization,
1495 to avoid problem with error handling.
1500 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1501 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1502 into CONN_USED_HASH and this prodice warning when put connection
1503 again in unused hash.
1506 Severity : enhancement
1508 Description: File striping can now be set to use an arbitrary pool of OSTs.
1510 Severity : enhancement
1512 Description: Export bytes_read/bytes_write count on OSC/OST.
1516 Description: Early reply size mismatch, MGC loses connection
1517 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1518 the connect flags are properly negotiated.
1523 Description: quotas are not honored with O_DIRECT
1524 Details : all writes with the flag O_DIRECT will use grants which leads to
1525 this problem. Now using OBD_BRW_SYNC to guard this.
1529 Description: add quota statistics
1530 Details : 1. sort out quota proc entries and proc code.
1531 2. add quota statistics
1533 Severity : enhancement
1535 Description: enable quota support for HEAD.
1539 Description: Properly propagate oinfo flags from lov to osc for statfs
1540 Details : restore missing copy oi_flags to lov requests.
1542 Severity : enhancement
1544 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1545 : l_getgroups(8), lst(8), routerstat(8)
1546 Details : included man pages for llobdstat(8), llstat(8),
1547 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1549 Severity : enhancement
1551 Description: configure's --enable-quota should check the
1552 : kernel .config for CONFIG_QUOTA
1553 Details : configure is terminated if --enable-quota is passed but
1554 : no quota support is in kernel
1558 Description: Remove portals compatibility
1559 Details : Remove portals compatibility, not interoperable with releases
1564 Description: Resolve device initialization race
1565 Details : Prevent proc handler from accessing devices added to the
1566 obd_devs array but yet be intialized.
1568 Severity : enhancement
1570 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1572 Severity : enhancement
1574 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1579 Description: Using wrong pointer in osc_brw_prep_request
1580 Details : Access to array[-1] can produce panic if kernel compiled with
1581 CONFIG_PAGE_ALLOC enabled
1585 Description: Client runs out of low memory
1586 Details : Consider only lowmem when counting initial number of llap pages
1590 Description: *optional* service tags registration
1591 Details : if the "service tags" package is installed on a Lustre node
1592 When the filesystem is mounted, a local-node service tag will
1593 be created. See http://inventory.sun.com/ for more information
1594 about the Service Tags asset management system.
1598 Description: Kernel BUG tries to release flock
1599 Details : Lustre does not destroy flock lock before last reference goes
1600 away. So always drop flock locks when client is evicted and
1601 perform unlock regardless of successfulness of speaking to MDS.
1605 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1609 Description: sanity test 65a fails if stripecount of -1 is set
1610 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1615 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1616 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1619 Severity : enhancement
1621 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1625 Description: do not process already freed flock
1626 Details : flock can possibly be freed by another thread before it reaches
1627 to ldlm_flock_completion_ast.
1631 Description: LBUG during stress test
1632 Details : Need properly lock accesses the flock deadlock detection list.
1636 Description: oops in page fault handler
1637 Details : kernel page fault handler can return two special 'pages' in error
1638 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1642 Description: timeout with invalidate import.
1643 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1644 should be handled by ptlrpcd. This produce long age waiting and
1645 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1647 Severity : enhancement
1649 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1653 Description: enable MGS and MDT services start separately
1654 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1655 a co-located MGS without starting the MGS, which is a complement
1656 to 'nosvc' mount option.
1660 Frequency : after recovery
1661 Description: precreate to many object's after del orphan.
1662 Details : del orphan st in oscc last_id == next_id and this triger growing
1663 count of precreated objects. Set flag LOW to skip increase count
1664 of precreated objects.
1668 Frequency : rare, on clear nid stats
1669 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1670 Details : when clean nid stats sometimes try destroy live entry,
1671 and this produce panic in free.
1675 Description: Stack overflow during MDS log replay
1676 ease stack pressure by using a thread dealing llog_process.
1680 Description: wait until IO finished before start new when do lock cancel.
1681 Details : VM protocol want old IO finished before start new, in this case
1682 need wait until PG_writeback is cleared until check dirty flag and
1683 call writepages in lock cancel callback.
1685 Severity : enhancement
1687 Description: using special macro for print time and cleanup in includes.
1691 Description: mds_mfd_close() ASSERTION(rc == 0)
1692 Details : In mds_mfd_close(), we need protect inode's writecount change
1693 within its orphan write semaphore to prevent possible races.
1697 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1701 Frequency : rare, on shutdown ost
1702 Description: don't hit live lock with umount ost.
1703 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1704 use shrink_dcache_sb instead.
1708 Description: don't panic with use echo client
1709 Details : echo client pass NULL as client nid pointer and this produce null
1710 pointer dereference.
1714 Description: fix build on ppc32
1715 Details : compile code with -m64 flag produce wrong object file for ppc32.
1719 Description: add message levels for liblustreapi
1723 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1724 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1725 will result in an -EAGAIN and in the exit call path we call
1726 alloc_rr() which will with increasing aggressiveness attempt to
1727 aquire precreated objects on the minimum number of required OSCs.
1731 Description: Use old size assignment to avoid deadlock
1732 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1733 the scheduling while holding a spinlock. We do not need locking
1734 for size in ll_update_inode() because size is only updated from
1735 the MDS for directories or files without objects, so there is no
1736 other place to do the update, and concurrent access to such inodes
1737 are protected by the inode lock.
1741 Description: resolve "_IOWR redefined" build error on SLES10
1745 Description: dump the memory debugging after all modules are unloaded to
1746 suppress false negative in conf_sanity test 39
1748 Severity : enhancement
1750 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1755 Description: fix SLES kernel versioning
1756 Details : the kernel version for our SLES 10 kernel did not include a "-"
1757 before the "smp" at the end. while this was not a problem in
1758 general, it did mean that software trying to use the kernel
1759 version to try to detect a vendor specific kernel would fail.
1760 this was most evident by the OFED build scripts.
1764 Description: Don't update lov_desc members until making sure they are valid
1765 Details : When updating lov_desc members via proc fs, need fix their
1766 validities before doing the real update.
1770 Description: don't put request into delay list while invalidate in flight.
1771 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1772 import in flight. this produce timeout for invalidate and sometimes
1773 can cause stale data.
1777 Frequency : on ppc only
1778 Description: not convert ost objects for directory because it's not exist.
1779 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1783 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1785 Severity : enhancement
1787 Description: Update to support 2.6.22.14 vanilla kernel.
1791 Frequency : rare, on recovery
1792 Description: read procfs can produce deadlock in some situation
1793 Details : Holding lprocfs lock which send rpc can produce block for destroy
1794 obd objects and this also block reconnect with -EALREADY. This
1795 isn't fix all lprocfs bugs - but make it rare.
1797 Severity : enhancement
1799 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1802 Frequency : frequent on X2 node
1804 Description: mdc_set_open_replay_data LBUG
1805 Details : Set replay data for requests that are eligible for replay.
1809 Description: lustre_mgs: operation 101 on unconnected MGS
1810 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1811 MGC, and late on MGC cannot successfully connect to MGS and a lot
1812 of the error messages complaining that MGS is not connected.
1816 Frequency : on network error
1817 Description: panic with double free request if network error
1818 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1819 but it's true only for synchronus enqueue, for async enqueue
1820 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1823 Severity : enhancement
1825 Description: client-side metadata stat-ahead during readdir(directory readahead)
1826 Details : perform client-side metadata stat-ahead when the client detects
1827 readdir and sequential stat of dir entries therein
1830 Frequency : on start mds
1832 Description: Implement get_info(last_id) in obdfilter.
1835 Frequency : occasional
1837 Description: Correctly check stale fid, not start epoch if ost not support SOM
1838 Details : open with flag O_CREATE need set old fid in op_fid3 because
1839 op_fid2 was overwritten with new generated fid, but mds can answer
1840 with one of these two fids and both is not stale. Setattr
1841 incorrectly started an epoch and assume will be called
1842 done_writing, but without SOM done_writing ever being called.
1845 Frequency : rare, depends on device drivers and load
1847 Description: MDS or OSS nodes crash due to stack overflow
1848 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1849 In some cases, in conjunction with device drivers that use a lot
1850 of stack the MDS (or possibly OSS) service threads could overflow
1851 the stack. One change which was identified to consume additional
1852 stack has been reworked to avoid the extra stack usage.
1855 Frequency : occasional
1857 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1858 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1859 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1860 oscc_has_objects and see OSC has no objests and call
1861 oscc_internal_create to resend create request.
1863 Severity : enhancement
1865 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1867 Severity : enhancement
1869 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1872 Frequency : very rare
1874 Description: Fix replay if there is an un-replied request and open
1875 Details : In some cases, older replay request will revert the
1876 mcd->mcd_last_xid on MDS which is used to record the client's
1877 latest sent request.
1879 Severity : enhancement
1881 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1883 Severity : enhancement
1885 Description: Add rhel5 support to HEAD.
1887 Serverity : enhancement
1889 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1894 Description: Don't allow skipping OSTs if index has been specified.
1895 Details : Don't allow skipping OSTs if index has been specified, make
1896 locking in internal create lots better.
1900 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1901 Details : access to bitfield in structure is always rounded to long
1902 and this produce problem with not atomic change any bit.
1906 Description: Lustre make rpms failed.
1907 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1908 builds Lustre rpms from tarball.
1911 Frequency : rare on shutdown ost
1913 Description: If llog cancel was not send before clean_exports phase, this can
1914 produce deadlock in llog code.
1915 Details : If llog thread has last reference to obd and call class_import_put
1916 this produce deadlock because llog_cleanup_commit_master wait when
1917 last llog_commit_thread exited, but this never success because was
1918 called from llog_commit_thread.
1922 Description: allow userland application know is lost one of stripes.
1923 Details : fill lvb_blocks with error code on ost and return it to
1924 application if error flag found.
1928 Description: NULL lov_tgts causing MDS oops
1929 Details : more safe checks for NULL lov_tgts for avoid oops.
1931 Severity : enhancement
1933 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1937 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1938 Details : If an OST is down the MDS will hang indefinitely in
1939 obd_statfs_rqset() waiting for the statfs data. While for
1940 MDS QOS usage of statfs, it should not stuck in waiting.
1942 Severity : enhancement
1944 Description: remote_acl support
1945 Details : Support ACL-based permission check for remote user.
1946 Support setfacl/getfacl for remote user with the utils
1947 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1948 the system "{s,g}etfacl" utils.
1950 Severity : enhancement
1952 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1954 Severity : enhancement
1956 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1960 Description: make mgs_setparam() handle fsname containing dash
1961 Details : fsname containing a dash does not work with lctl conf_param
1963 Severity : enhancement
1965 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1967 Severity : enhancement
1969 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1973 Frequency : rare, at shutdown
1974 Description: access already free / zero obd_namespace.
1975 Details : if client_disconnect_export was called without force flag set,
1976 and exist connect request in flight, this can produce access to
1977 NULL pointer (or already free pointer) when connect_interpret
1978 store ocd flags in obd_namespace.
1982 Frequency : only at startup
1983 Description: not alloc memory with spinlock held.
1984 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1985 if any spinlock held.
1987 Severity : enhancement
1989 Description: make lustre randomly fail allocating memory
1990 Details : Make lustre randomly failed allocating memory for testing purpose.
1992 Severity : enhancement
1994 Description: lost problems with lov objid file
1995 Details : Fixes some scability and access to not inited memory problems
1996 in work with lov objdid file.
2001 Description: lfs find does not continue on file error
2002 Details : Continue other files processing when a file/dir is absent.
2006 Description: Inconsistent usage of lustre_pack_reply()
2007 Details : Standardize the usage of lustre_pack_reply() such that it
2008 always generate a CERROR on failure.
2011 Frequency : occasional
2013 Description: MDS hang or stay in waiting lock
2014 Details : If client receive lock with CBPENDING flag ldlm need send lock
2015 cancel as separate rpc, to avoid situation when cancel request
2016 can't processed due all i/o threads stay in wait lock.
2020 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2024 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2028 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2032 Description: Update to RHEL4 latest kernel.
2034 Severity : enhancement
2036 Description: Build SLES10 patchless client fails
2037 Details : The configure was broken by run ./configure with
2038 --with-linux-obj=.... argument for patchless client. When the
2039 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2040 can't search header adequately. Use absolute path such as
2041 -I($LINUX)/include instead.
2045 Description: interrupt oig_wait produce painc on resend.
2046 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2047 private set, and this produce situation when rq_ptlrpcd_data not
2048 copyed to new allocated request and triggered LBUG on assert
2049 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2050 ptlrpcd set if request is changed and can be safety changed to use
2055 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2056 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2057 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2058 these macros and replace them with their difinitions instead.
2062 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2063 Details : Update which_patch & target file for SLES10 latest kernel.
2066 Description: Remove client patches from SLES 10 kernel.
2067 Details : This causes SLES 10 clients to behave as patchless clients
2068 even on a Lustre-patched (server) kernel.
2070 Severity : enhancement
2072 Description: self-adjustable client's lru lists
2073 Details : use adaptive algorithm for managing client cached locks lru
2074 lists according to current server load, other client's work
2075 pattern, memory activities, etc. Both, server and client
2076 side namespaces provide number of proc tunables for controlling
2079 Severity : enhancement
2081 Description: light-weight GSS support
2082 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2083 and reduce performance overhead.
2085 Severity : enhancement
2087 Description: Linux keyring support
2088 Details : Support using service of Linux keyring for Lustre GSS internal
2089 context refresh/cache mechanism.
2093 Description: Fix errors in lfs documentation
2094 Details : Fixes man pages
2098 Description: don't use GFP_* in generic Lustre code.
2099 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2103 Description: obdclass is limited by single OBD_ALLOC(idarray)
2104 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2108 Description: warnings with build patchless client with vanila 2.6.19 and up
2109 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2110 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2114 Frequency : Only for SLES
2115 Description: sanity_quota fail test_1
2116 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2117 file, which makes TSTID[2] inunique.
2121 Frequency : Only for SLES9
2122 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2123 Details : Update target/ChangeLog/which_patch .
2127 Frequency : Only for test_10 in sanity-quota.sh
2128 Description: a bug in quota test code
2129 Details : checking a test flag in a wrong place
2133 Frequency : Only for test_13 in sanity-quota.sh
2134 Description: a bug in quota test script
2135 Details : shouldn't check overfull allocated quota
2139 Description: jbd statistics
2140 Details : Port older jbd statistics patch for sles10
2142 Severity : enhancement
2144 Description: Add printing inode info into message about error in writepage.
2149 Description: reply_lock_interpret crash due to race with it and lock cancel.
2150 Details : Do not replay locks that are being cancelled. Do not reference
2151 locks by their address during replay, just by their handle.
2155 Frequency : When flocks are used.
2156 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2157 Details : Flock locks might destroy just granted lock if it could be merged
2158 with another existing flock, this is done in completion handler,
2159 so teach ldlm_cli_enquque_fini that this is a valid case for
2165 Description: Oops in read and write path when failing to allocate lock.
2166 Details : Check if lock allocation failed and return error back.
2170 Description: lstripe command fails for valid OST index
2171 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2172 instead of lov->desc.ld_active_tgt_count.
2176 Description: sanity.sh failed test 103
2177 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2182 Description: df doesn't work properly if diskfs blocksize != 4K
2183 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2187 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2188 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2190 Severity : enhancement
2192 Description: Adaptive timeouts
2193 Details : RPC timeouts adapt to changing server load and network
2194 conditions to reduce resend attempts and improve recovery time.
2198 Description: llapi_file_create() does not allow some changes
2199 Details : add llapi_file_open() that allows specifying the mode and
2200 open flags, and also returns an open file handle.
2204 Description: merge and cleanup kernel patches.
2205 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2209 Description: Add journal checksum support.(Kernel part)
2210 Details : The journal checksum feature adds two new flags i.e
2211 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2212 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2213 indicates that the commit block contains the checksum for
2214 the blocks described by the descriptor blocks. Now commit
2215 record can be sent to disk without waiting for descriptor
2216 blocks to be written to disk. This behavior is controlled
2217 using JBD2_FEATURE_ASYNC_COMMIT flag.
2221 Description: OSS needs mutliple precreate threads
2222 Details : Add ability to start more than one create thread per OSS.
2226 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2227 Details : execute lfs setstripe on client
2231 Description: mds_obd_create error creating tmp object
2232 Details : When the user sets quota on root, llog will be affected and can't
2233 create files and write files.
2236 Frequency : Always on ia64 patchless client, and possibly others.
2238 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2239 Details : This allows the patchless client to be loaded on architectures
2240 without this export.
2244 Description: RedHat Update kernel for RHEL5
2245 Details : Add the kernel config file for RHEL5.
2249 Description: RedHat Update kernel for RHEL5
2250 Details : Modify the kernel config file more closer RHEL5.
2254 Description: Build failure against Centos5 (RHEL5)
2255 Details : Define PAGE_SIZE when it isn't present.
2259 Description: test_7 on sanity-quota.sh can't be used on separate machines
2260 Details : add facet to handle it
2264 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2265 Details : with NFS, the anon dentry's parent was set to itself in
2266 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2267 corresponding dentry other than use rec->ur_name.
2269 Severity : enhancement
2271 Description: lfs setstripe enhancement
2272 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2276 Description: enable data checksumming by default
2277 Details : enable checksum by default, allow --disable-checksum
2278 configure option and "-o nochecksum" mount option. Checksums
2279 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2283 Description: lustre support for RHEL5
2284 Details : Add support for RHEL5.
2288 Description: Client eviction due to failover config
2289 Details : after a connection loss, the lustre client should attempt to
2290 reconnect to the last active server first before trying the
2291 other potential connections.
2295 Description: when mds and osts use different quota unit(32bit and 64bit),
2296 quota will be released repeatly.
2297 Details : void sending multiple quota reqs to mds, which will keep the
2298 status between the reqs.
2302 Description: osts not allocated evenly to files
2303 Details : change the condition to increase offset_idx
2306 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2308 Description: Data corruption for OSTs that are formatted larger than 2TB
2310 Details : When generating the bio request for lustre file writes the
2311 sector number would overflow a temporary variable before being
2312 used for the IO. The data reads correctly from Lustre (which
2313 will overflow in a similar manner) but other file data or
2314 filesystem metadata may be corrupted in some cases.
2318 Description: Tune the kernel for good SCSI performance.
2319 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2320 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2325 Description: rewrite ext2-derived code in obdclass/uuid.c
2326 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2327 from scratch preserving functionality.
2331 Description: "lfs find -obd UUID" prints directories
2332 Details : "lfs find -obd UUID" will return all directory names instead
2333 of just file names. It is incorrect because the directories
2334 do not reside on the OSTs.
2338 Description: use i_size_read and i_size_write in 2.6 port
2339 Details : replace inode->i_size access with i_size_read/write()
2342 Frequency : when removing large files
2344 Description: scheduling issue during removal of large Lustre files
2345 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2346 It causes scheduling issues when removing large files (17TB in the
2350 Frequency : only with liblustre clients on XT3
2352 Description: evictions taking too long
2353 Details : allow llrd to evict clients directly on OSTs
2356 Frequency : only on ppc
2358 Description: /proc/fs/lustre/devices broken on ppc
2359 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2360 We should make sure the type of 'index' is loff_t and then cast
2361 explicitly as needed below. Do not assign an explicitly cast
2365 Frequency : only for rhel5
2367 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2368 Details : Modify the target file & which_kernel.
2370 Severity : enhancement
2372 Description: omit set fsid for export NFS
2373 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2374 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2378 Frequency : Always, for kernels after 2.6.16
2379 Description: Fix warning idr_remove called for id=.. which is not allocated.
2380 Details : Last kernels save old s_dev before kill super and not allow
2381 to restore from callback - restore it before call kill_anon_super.
2385 Description: buffer overruns could theoretically occur
2386 Details : llapi_semantic_traverse() modifies the "path" argument by
2387 appending values to the end of the origin string, and a
2388 overrun may occur. Adding buffer overrun check in liblustreapi.
2392 Description: Fix error on 'ls .' at the top of the Lustre mount.
2393 Details : Don't revalidate dentry if it is a root dentry.
2397 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2398 Details : Modify vm-tunables-rhel4.patch.
2402 Description: Kernel config for 2.6.18-vanilla.
2403 Details : Modify targets/2.6-vanilla.target.in.
2404 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2405 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2406 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2407 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2411 Description: improve handling recoverable errors
2412 Details : if request processig with error which can be recoverable on server
2413 request should be resend, otherwise page released from cache and
2418 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2419 Details : Modify target file & which_patch.
2420 A flaw was found in the IA32 system call emulation provided
2421 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2422 value could be stored in the %RAX register, which could trigger an
2423 out-of-bounds system call table access. An untrusted local user
2424 could exploit this flaw to run code in the kernel
2425 (ie a root privilege escalation). (CVE-2007-4573).
2429 Description: O_DIRECT bypasses client statistics.
2430 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2431 read_ahead_stats were updated. copied the stats section from
2432 osc_send_oap_rpc() into async_internal().
2436 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2440 Description: change order of libsysio includes
2441 Details : '#include sysio.h' should always come before '#include xtio.h'
2445 Description: handle "serious error: objid * already exists" more gracefully
2446 Details : If LAST_ID value on disk is smaller than the objects existing in
2447 the O/0/d* directories, it indicates disk corruption and causes an
2448 LBUG(). If the object is 0-length, then we should use the existing
2449 object. This will help to avoid a full fsck in most cases.
2451 Severity : enhancement
2453 Description: adapt the lustre_config script to support the upgrade case
2454 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2455 server targets to 1.6 in parallel.
2459 Description: To avoid grant space > avaible space when the disk is almost
2460 full. Without this patch you might see the error "grant XXXX >
2461 available" or some LBUG about grant, when the disk is almost
2463 Details : In filter_check_grant, for non_grant cache write, we should
2464 check the left space by if (*left > ungranted + bytes), instead
2465 of (*left > ungranted), because only we are sure the left space
2466 is enough for another "bytes", then the ungrant space should be
2467 increase. In client, we should update cl_avail_grant only there
2468 is OBD_MD_FLGRANT in the reply.
2472 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2473 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2474 described in CVE-2007-4573. This problem could allow untrusted
2475 local users to gain root access.
2478 Frequency : when using O_DIRECT and quotas
2480 Description: Incorrect file ownership on O_DIRECT output files
2481 Details : block usage reported by 'lfs quota' does not take into account
2482 files that have been written with O_DIRECT.
2487 Description: touch file failed when fs is not full
2488 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2489 otherwise we can get ENOSP while fs is not full.
2493 Description: parallel lock callbacks
2494 Details : Instead of sending blocking and completion callbacks as separated
2495 requests, adding them to a set and sending in parallel.
2498 Frequency : only for Cray XT3
2499 Bugzilla : 12829/13455
2500 Description: Changing primary group doesn't change the group lustre assigns to
2502 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2503 overridden with the primary group provided by the group upcall,
2504 whereas the supplied fsgid can be trusted if it is in the list of
2505 supplementary groups returned by the group upcall.
2507 Severity : enhancement
2509 Description: Allow masking D_WARNING, D_ERROR messages from console
2510 Details : Console messages can now be disabled via lnet.printk.
2514 Description: User code with malformed file open parameter crashes client node
2515 Details : Before packing join_file req, all the related reference should be
2516 checked carefully in case some malformed flags cause fake
2517 join_file req on client.
2521 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2522 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2523 mode is equal to the granted mode. However, it should also check
2524 that LDLM_FL_CP_REQD is not set.
2528 Description: Heavy nfs access might result in deadlocks
2529 Details : After ELC code landed, it is now improper to enqueue any mds
2530 locks under och_sem, because enqueue might want to decide to
2531 cancel open locks for same inode we are holding och_sem for.
2535 Description: Client eviction while running blogbench
2536 Details : A lot of unlink operations with concurrent I/O can lead to a
2537 deadlock causing evictions. To address the problem, the number of
2538 oustanding OST_DESTROY requests is now throttled to
2539 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2540 callbacks are processed in priority.
2544 Description: enable ACLs on MDS by default
2545 Details : ACLs must be enabled on MDS by default.
2548 Frequency : PPC/PPC64 only
2550 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2551 Details : fix duplicated definitions between asm-ppc64/types.h and
2552 lustre_types.h on PPC.
2555 Frequency : PPC/PPC64 only
2557 Description: asm-ppc/segment.h does not exist
2558 Details : fix compile issue on PPC.
2562 Description: better handle error messages in extents code
2565 Frequency : RHEL4 only
2567 Description: mkfs is very slow on IA64/RHEL4
2568 Details : A performance regression has been discovered in the MPT Fusion
2569 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2570 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2571 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2573 Severity : enhancement
2575 Description: SNMP support enhancement
2576 Details : Adding total number of sampled request for an MDS node in snmp
2579 Severity : enhancement
2581 Description: Optimize ldlm waiting list processing for PR extent locks
2582 Details : When processing waiting list for read extent lock and meeting read
2583 lock that is same or wider to it that is not contended, skip
2584 processing rest of the list and immediatelly return current
2585 status of conflictness, since we are guaranteed there are no
2586 conflicting locks in the rest of the list.
2590 Description: Time out and refuse to reconnect
2591 Details : When the failover node is the primary node, it is possible
2592 to have two identical connections in imp_conn_list. We must
2593 compare not conn's pointers but NIDs, otherwise we can defeat
2594 connection throttling.
2598 Description: port llog fixes from b1_6 into HEAD
2599 Details : Port llog reference couting and some llog cleanups from b1_6
2600 (bug 10800) into HEAD, for protect from panic and access to already
2601 free llog structures.
2605 Description: Detect stride IO mode in read-ahead
2606 Details : When a client does stride read, read-ahead should detect that and
2607 read-ahead pages according to the detected stride pattern.
2611 Description: data checksumming impacts single node performance
2612 Details : add support for several checksum algorithm. Currently, only CRC32
2613 and Adler-32 are supported. The checksum type can be changed on
2614 the fly via /proc/fs/lustre/osc/*/checksum_type.
2618 Description: use adler32 for page checksums
2619 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2624 Description: build for x2 fails
2625 Details : fix compile issue on Cray systems.
2629 Description: Properly match for duplicate locks
2630 Details : Due to different lock order from skiplists code, we need to
2631 traverse entire list for now
2634 Frequency : only on PPC/SLES10
2636 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2637 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2638 be defined. Add a hack in mkfs_lustre.c to work around this compile
2643 Description: LASSERT on MDS when client holding flock lock dies
2644 Details : ldlm pool logic depends on number of granted locks equal to
2645 number of released locks which is not true for flock locks, so
2646 just exclude such locks from consideration.
2650 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2651 Details : Use fsfilt_sync() for both the whole filesystem sync and
2652 individual file sync to eliminate dangerous inode locking
2653 with I_LOCK that can lead to a deadlock.
2657 Description: performance in 1.6.3
2658 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2659 sure that 1MB requests are not fragmented by the block layer.
2661 Severity : enhancement
2663 Description: organize the server-side client stats on per-nid basis
2664 Details : Change the structure of stats under obdfilter and mds to
2672 The "uuid"s file would list the uuids of _active_ exports.
2673 And the clear entry is to clear all stats and stale nids.
2675 Severity : enhancement
2677 Description: eliminate client locks in face of contention
2678 Details : file contention detection and lockless i/o implementation
2679 for contended files.
2683 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2686 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2688 Description: MDS may lose file striping (and hence file data) in some cases
2689 Details : If there are additional extended attributes stored on the MDS,
2690 in particular ACLs, SELinux, or user attributes (if user_xattr
2691 is specified for the client mount options) then there is a risk
2692 of attribute loss. Additionally, the Lustre file striping
2693 needs to be larger than default (e.g. striped over all OSTs),
2694 and an additional attribute must be stored initially in the
2695 inode and then increase in size enough to be moved to the
2696 external attribute block (e.g. ACL growing in size) for file
2701 Description: skiplist implementation simplification
2702 Details : skiplists are used to group compatible locks on granted list
2703 that was implemented as tracking first and last lock of each
2704 lock group the patch changes that to using doubly linked lists
2708 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2709 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2710 because LASSERT on both the data supplied by a client, and the
2711 data on disk is dangerous and incorrect.
2713 Severity : enhancement
2715 Description: Slow truncate/writes to huge files at high offsets.
2716 Details : Directly associate cached pages to lock that protect those pages,
2717 this allows us to quickly find what pages to write and remove
2718 once lock callback is received.
2722 Description: more ldlm soft lockups
2723 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2724 starve other threads from the resource lock for a long time in
2725 case of long waiting queue, so change the debug level from
2726 D_OTHER to the less frequently used D_INFO.
2728 Severity : enhancement
2730 Description: add -gid, -group, -uid, -user options to lfs find
2734 Description: Hung threads in invalidate_inode_pages2_range
2735 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2736 once one is completed. As a result, some RPCs are stuck in the
2737 queue and are never sent.
2741 Description: filter threads hungs on waiting journal commit
2742 Details : Cleanup filter group llog code, then only filter group llog will
2743 be only created in the MDS/OST syncing process.
2747 Description: Procfs and llog threads access destoryed import sometimes.
2748 Details : Sync the import destoryed process with procfs and llog threads by
2749 the import refcount and semaphore.
2751 Severity : enhancement
2753 Description: openlock cache of b1_6 port to HEAD
2758 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2759 Details : If insertion of an extent fails, then discard the inode
2760 preallocation and free data blocks else it can lead to duplicate
2765 Description: don't always update ctime in ext3_xattr_set_handle()
2766 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2767 In some cases the ctime should not be updated, for example for
2768 2.0->1.8 compatibility it is necessary to delete an xattr and it
2769 should not update the ctime.
2773 Bugzilla : 15713/16362
2774 Description: Assertion in iopen_connect_dentry in 1.6.3
2775 Details : looking up an inode via iopen with the wrong generation number can
2776 populate the dcache with a disconneced dentry while the inode
2777 number is in the process of being reallocated. This causes an
2778 assertion failure in iopen since the inode's dentry list contains
2779 both a connected and disconnected dentry.
2783 Description: assertion failure in ldlm_handle2lock()
2784 Details : fix a race between class_handle_unhash() and class_handle2object()
2785 introduced in lustre 1.6.5 by bug 13622.
2790 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2791 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2796 Description: oops in sd_iostats_seq_show()
2797 Details : unloading/reloading the scsi low level driver triggers a kernel
2798 bug when trying to access the sd iostat file.
2803 Description: Kernel panics during QLogic driver reload
2804 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2805 patch, causing memory corruption.
2810 Description: journal_dev option does not work in b1_6
2811 Details : pass mount option during pre-mount.
2813 Severity : enhancement
2815 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2816 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2817 extent information of a file. It can be used to map logical blocks
2818 in a file to physical blocks in the block device.
2822 Description: LDLM soft lockups - improvement
2823 Details : It is be possible to send the lock handle along with each read
2824 or write request because the client is already doing a lock match
2825 itself so there isn't any reason the OST should have to re-do that
2829 Frequency : only on Cray X2
2831 Description: X2 build failures
2832 Details : fix build failures on Cray X2.
2836 Description: xid & resent requests
2837 Details : Initialize RPC XID from clock at startup (randomly if clock is
2840 Severity : enhancement
2842 Description: Add lustre_start utility to start or stop multiple Lustre servers
2847 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2848 Details : In case of memory pressure, list_del() can be called twice on
2849 req->rq_history_list, causing a kernel oops.
2853 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2854 Details : fix stack overflow in the distributed lock manager by defering
2855 export eviction after a failed AST to the elt thread instead of
2856 handling it in the dlm interpret routine.
2860 Description: Convert some comments to new format.
2861 Details : Update documenting comments to match doxygen conventions.
2865 Description: Grammar fixes.
2866 Details : A couple of trivial spelling fixes.
2870 Description: OSD_COUNTERS-mandatory
2871 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2872 to simplify the code.
2876 Description: simplify cmm_device freeing logic.
2877 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2881 Description: Add lockdep support to dt_object_operations locking interface.
2882 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2883 parameter indicating lock ordering. Update mdd code to use new
2888 Description: Introduce failloc constants for lockless IO tests.
2889 Details : Add two new failloc constants to test lockless IO. Only one of
2890 them in implemented---another is checked in yet to be landed
2895 Description: Add lockdep support for inode mutex.
2896 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2897 in the situations where Lustre has to lock more than one inode
2902 Description: Add optional invariants checking support.
2903 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2904 configure switch. This macro is to be used for consistency and
2905 sanity checks that are too expensive to be left in `production'
2910 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2911 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2916 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2917 Details : Use spin_lock_nested() in (the only) situation where more than
2918 one ldlm_lock is locked simultaneously. Also, fix possible
2919 dead-lock in ldlm_lock_change_resource() by enforcing particular
2924 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2925 Details : Instead of specifying each ldlm_lock call-back through separate
2926 parameter, wrap them into struct ldlm_callback_suite.
2930 Description: Kill join_lru obd method and its callers.
2931 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2932 in memory---a requirement implemented through ->o_join_lru() obd
2933 method in HEAD. Remove this method and its users.
2937 Description: Add asynchronous ldlm ENQUEUE completion handler.
2938 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2939 case that stock ldlm_completion_ast() cannot handle as it waits
2940 until lock is granted. Introduce new ldlm_completion_ast_async()
2941 for this. Also comment ldlm_completion_ast().
2945 Description: ldlm_error <-> errno conversion.
2946 Details : Add functions to map (rather arbitrary) between LDLM error codes
2947 and standard errno values. CLIO needs this to prevent LDLM specific
2948 constants from escaping ldlm and osc.
2952 Description: Kill unused ldlm_handle2lock_ns() function.
2953 Details : Kill unused ldlm_handle2lock_ns() function.
2957 Description: Add lu_ref support to ldlm_lock
2958 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2959 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2960 and ldlm_resource. LDLM interface has to be changed, because code that
2961 releases a reference on a lock, has to "know" what reference this is.
2962 In the most frequent case
2964 lock = ldlm_handle2lock(handle);
2966 LDLM_LOCK_PUT(lock);
2968 no changes are required. When any other reference (received _not_ from
2969 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2970 instead of LDLM_LOCK_PUT().
2972 Arguably, changes are pervasive, and interface requires some discipline
2973 for proper use. On the other hand, it was very instrumental in finding
2974 a few leaked lock references.
2978 Description: Add ldlm_lock_addref_try().
2979 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2980 attempts to addref a lock that might be being canceled
2985 Description: Add ldlm_weigh_callback().
2986 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2987 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2988 hard-coded `number of pages' logic.
2992 Description: Add lockdep annotations to llog code.
2993 Details : Use appropriately tagged _nested() locking calls in the places
2994 where llog takes more than one ->lgh_lock lock.
2998 Description: Add loi_kms_set().
2999 Details : Wrap kms updates into a helper function.
3003 Description: Constify instances of struct lsm_operations.
3004 Details : Constify instances of struct lsm_operations.
3008 Description: lu_conf support.
3009 Details : On a server, a file system object is uniquely identified
3010 by a fid, which is sufficient to locate and load all object
3011 state (inode). On a client, on the other hand, more data are
3012 necessary instantiate an object. Change lu_object_find() and
3013 friends to take additional `lu_conf' argument describing object.
3014 Typically this includes layout information.
3018 Description: lu_context fixes.
3019 Details : Introduce new lu_context functions that are needed on the client
3020 side, where some system threads (ptlrpcd) are shared by multiple
3021 modules, and so cannot be stopped during module shutdown.
3025 Description: Add start and stop methods to lu_device_type_operations.
3026 Details : Introduce two new methods in lu_device_type_operations, that are
3027 invoked when first instance of a given type is created and last one
3028 is destroyed respectively. This is need by CLIO.
3032 Description: Add lu_ref support to struct lu_device.
3033 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
3034 track leaked references.
3038 Description: Introduce lu_kmem_descr.
3039 Details : lu_kmem_descr and its companion interface allow to create
3040 and destroy a number of kmem caches at once.
3044 Description: Fix lu_object finalization race.
3045 Details : Fix a race between lu_object_find() finding an object and its
3046 concurrent finalization. This race is (most likely) not possible
3047 on the server, but might happen on the client.
3051 Description: Introduce lu_ref interface.
3052 Details : lu_ref is a debugging module allowing to track references to
3053 a given object. It is quite cpu expensive, and has to be
3054 explicitly enabled with --enable-lu_ref. See usage description
3059 Description: Factor lu_site procfs stats into a separate function.
3060 Details : Separate lu_site stats printing code into a separate function
3061 that can be reused on a client.
3065 Description: Constify instances of struct {lu,dt,md}_device_operations.
3066 Details : Constify instances of struct {lu,dt,md}_device_operations.
3070 Description: Introduce struct md_site and move meta-data specific parts of
3071 struct lu_site here.
3072 Details : Move md-specific fields out of struct lu_site into special struct
3073 md_site, so that lu_site can be used on a client.
3077 Description: Kill mdd_lov_destroy().
3078 Details : Remove unused mdd code.
3082 Description: Add st_block checking to multistat.c.
3083 Details : Add st_block checking to multistat.c.
3087 Description: Add lu_ref support to struct obd_device.
3088 Details : Add lu_ref tracking to obd_device.
3092 Description: Kill obd_set_fail_loc().
3093 Details : Remove unused code.
3097 Description: Add special type for ptlrpc_request interpret functions.
3098 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3099 there. Actual usage will be in CLIO.
3103 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3104 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3105 doesn't work with lockdep.
3109 Description: Add rwv.c test program.
3110 Details : New testing program exercising readv(2) and writev(2) (Qian).
3114 Description: Add sendfile.c test program.
3115 Details : New testing program exercising sendfile(2) (Jay).
3119 Description: Ratelimit a message that can be very frequent.
3120 Details : Ratelimit a memory allocation failure message that can
3125 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3126 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3130 Description: Liblustre build fixes.
3131 Details : Liblustre build fixes.
3135 Description: libcfs: add cfs_{need,cond}_resched() interface.
3136 Details : libcfs: add cfs_{need,cond}_resched() definition and
3137 implementations for Linux, NT, and liblustre.
3139 Severity : enhancement
3141 Description: More exported tunables for mballoc
3142 Details : Add support for tunable preallocation window and new tunables for
3143 large/small requests
3147 Description: Detect corruption of block bitmap and checking for preallocations
3148 Details : Checks validity of on-disk block bitmap. Also it does better
3149 checking of number of applied preallocations. When corruption is
3150 found, it turns filesystem readonly to prevent further corruptions.
3154 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3155 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3156 window to 0. If it does hit cache miss, and read-pattern is still
3157 stride-io mode, does not reset the stride window, but also does not
3158 increase the stride window length in this case.
3162 Frequency : only for big-endian servers
3163 Description: Check if system is big-endian while mounting fs with extents feature
3164 Details : Mounting a filesystem with extents feature will fail on big-endian
3165 systems since ext3-based ldiskfs is not supported on big-endian
3166 systems. This can be over-riden with "bigendian_extents" mount option.
3168 Severity : enhancement
3170 Description: The root squash functionality
3171 Details : A security feature, which is to prevent users from being able
3172 to mount lustre on their desktop, run as root, and delete
3173 all of the files in the filesystem. The goal is accomplished by
3174 remapping user id (UID) and group id (GID) of the root user to
3175 a UID and GID specified by the system administartor via Lustre
3176 configuration management server (MGS). The functionality also
3177 allows to specify sets of clients for which the remapping does
3182 Description: Excessive recovery window
3183 Details : With AT enabled, the recovery window can be excessively long (6000+
3184 seconds). To address this problem, we no longer use
3185 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3186 timeout no longer depends on the service time, it is set to
3187 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3188 time via pb_service_time.
3192 Description: Watchdog triggered on MDS failover
3193 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3194 the OSTs know that the MDS "UUID" can be reused for the same export
3195 from a different NID, so we do not need to wait for the export to be
3199 Frequency : rare, only if using MMP with Linux RAID
3201 Description: MMP doesn't work with Linux RAID
3202 Details : While using HA for Lustre servers with Linux RAID, it is possible
3203 that MMP will not detect multiple mounts. To make this work we
3204 need to unplug the device queue in RAID when the MMP block is being
3205 written. Also while reading the MMP block, we should read it from
3206 disk and not the cached one.
3208 Severity : enhancement
3210 Description: open file using fid
3211 Details : A file can be opened using just its fid, like
3212 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3215 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3218 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3219 Details : add proper locking for creation of jbd_4k slab cache
3223 Description: MMP check in ext3_remount() fails without displaying any error
3224 Details : When multiple mount protection fails during remount, proper error
3227 Severity : enhancement
3229 Description: Allow stripe size to be up to 4G-64k
3230 Details : Fix math logic to allow large stripe sizes.
3234 Description: add check for >8TB ldiskfs filesystems
3235 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3236 Don't allow >8TB ldiskfs filesystems to be mounted without
3237 force_over_8tb mount option
3242 Descriptoin: OSS complains "too many blocks" and crashes
3243 Details : In filter_commitrw_write():
3244 1. lquota_chkquota();
3245 2. obdo_from_inode();
3246 3. lquota_pending_commit();
3247 "2." has changed uid and gid in this case, then caused
3248 lquota_chkquota() and lquota_pending_commit() to use
3249 different uid and gid. That is the root of the bug.
3253 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3254 Details : Detect and print out the 'stripe count = -1' case properly.
3258 Description: Out or order replies might be lost on replay
3259 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3260 request with tid smaller than one currently being added, add it
3261 to the start, not end of the list.
3263 --------------------------------------------------------------------------------
3265 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3267 * Support for kernels:
3268 2.6.5-7.283 (SLES 9),
3269 2.6.9-55.EL (RHEL 4),
3270 2.6.16.46-0.14 (SLES 10),
3271 2.6.18.8 vanilla (kernel.org)
3272 * Client support for unpatched kernels:
3273 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3274 2.6.16 - 2.6.22 vanilla (kernel.org)
3275 * Due to recently discovered recovery problems, we do not recommend
3276 using patchless RHEL 4 clients with this or any earlier release.
3277 * Recommended e2fsprogs version: 1.39.cfs8
3278 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3279 * Starting with this release, the ldiskfs backing filesystem required
3280 by Lustre is now in its own package, lustre-ldiskfs. This package
3281 should be installed. It is versioned separately from Lustre and
3282 may be released separately in future.
3286 Description: block reactivating mgc import until all deactivates complete
3287 Details : Fix race when failing back MDT/MGS to itself (testing)
3290 Frequency : at statup only
3292 Description: mds_lov_synchronize race leads to various problems
3293 Details : simultaneous MDT->OST connections at startup can cause the
3294 sync to abort, leaving the OSC in a bad state.
3296 Severity : enhancement
3298 Description: add optional extra BUILD_VERSION info
3299 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3300 to override the lustre version.
3303 Frequency : 2.6.18 servers only
3305 Description: ll_kern_mount() doesn't release the module reference
3306 Details : The ldiskfs module reference count never drops down to 0
3307 because ll_kern_mount() doesn't release the module reference.
3312 Description: server LBUG when using old ost_num_threads parameter
3313 Details : Accept the old ost_num_threads parameter but warn that it
3314 is deprecated, and fix an off-by-one error that caused an LBUG.
3319 Description: Transient SCSI error results in persistent IO issue
3320 Details : iobuf->dr_error is not reinitialized to 0 between two
3324 Frequency : sometimes when underlying device returns I/O errors
3326 Description: OSTs not going read-only during write failures
3327 Details : OSTs are not remounted read-only when the journal commit threads
3328 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3329 instead of the ext3 wrappers.
3333 Description: poor connect scaling with increasing client count
3334 Details : Don't run filter_grant_sanity_check for more than 100 exports
3335 to improve scaling for large numbers of clients.
3338 Frequency : SLES10 only
3340 Description: sanity-quota.sh quotacheck failed: rc = -22
3341 Details : Quotas cannot be enabled on SLES10.
3344 Frequency : liblustre clients only
3346 Description: getdirentries does not give error when run on compute nodes
3347 Details : getdirentries does not fail when the size specified as an argument
3348 is too small to contain at least one entry
3350 Severity : enhancement
3352 Description: Add LNET router traceability for debug purposes
3353 Details : If a checksum failure occurs with a router as part of the
3354 IO path, the NID of the last router that forwarded the bulk data
3355 is printed so it can be identified.
3360 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3361 Details : Due to a race condition, liblustre clients were occasionally
3362 evicted incorrectly.
3364 Severity : enhancement
3366 Description: lfs setstripe use optional parameters instead of postional
3369 Severity : enhancement
3371 Description: Nanosecond timestamp support for ldiskfs
3372 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3373 resolution timestamps. There is not yet support for this at
3374 the Lustre filesystem level.
3377 Frequency : during server recovery
3379 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3380 Details : request with rq_no_resend flag not awake l_wait_event if they get
3384 Frequency : nfs export on patchless client
3386 Description: connectathon hang when test nfs export over patchless client
3387 Details : Disconnected dentry cannot be found with lookup, so we do not need
3388 to unhash it or make it invalid
3391 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3392 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3393 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3398 Description: open req refcounting wrong on reconnect
3399 Details : If reconnect happened between getting open reply from server and
3400 call to mdc_set_replay_data in ll_file_open, we will schedule
3401 replay for unreferenced request that we are about to free.
3402 Subsequent close will crash in variety of ways.
3403 Check that request is still eligible for replay in
3404 mdc_set_replay_data().
3409 Description: disable writes to filesystem when reading health_check file
3410 Details : the default for reading the health_check proc file has changed
3411 to NOT do a journal transaction and write to disk, because this
3412 can cause reads of the /proc file to hang and block HA state
3413 checking on a healthy but otherwise heavily loaded system. It
3414 is possible to return to the previous behaviour during configure
3415 with --enable-health-write.
3417 Severity : enhancement
3419 Description: 64-bit inode version
3420 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3421 to the inode. This will be required for version-based recovery.
3426 Description: MDS fails to start if a duplicate client export is detected
3427 Details : in some rare cases it was possible for a client to connect to
3428 an MDS multiple times. Upon recovery the MDS would detect this
3429 and fail during startup. Handle this more gracefully.
3431 Severity : enhancement
3433 Description: Add -o localflock option to simulate old noflock
3435 Details : This will achieve local-only flock/fcntl locks
3441 Description: log_commit_thread vs filter_destroy race leads to crash
3442 Details : Take import reference before releasing llog record semaphore
3447 Description: Wrong request locking in request set processing
3448 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3449 imp_delayed_list, in this place should be used imp_lock.
3452 Frequency : when reconnection
3454 Description: Grant Leak when osc reconnect to OST
3455 Details : When osc reconnect ost, OST(filter) should check whether it
3456 should grant more space to client by comparing fed_grant and
3457 cl_avail_grant, and return the granted space to client instead
3458 of "new granted" space, because client will call osc_init_grant
3459 to update the client grant space info.
3462 Frequency : when client reconnect to OST
3464 Description: Grant Leak when osc do resend and replay bulk write
3465 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3466 bulk write request, because the grant info while be sync between
3467 OSC and OST when reconnect, and we should ignore the grant info
3468 these of resend/replay write req.
3473 Description: Grant space more than available left space sometimes.
3474 Details : When then OST is about to be full, if two bulk writing from
3475 different clients came to OST. Accord the avaliable space of the
3476 OST, the first req should be permitted, and the second one
3477 should be denied by ENOSPC. But if the seconde arrived before
3478 the first one is committed. The OST might wrongly permit second
3479 writing, which will cause grant space > available space.
3482 Frequency : when client is evicted
3484 Description: Grant might be wrongly erased when osc is evicted by OST
3485 Details : when the import is evicted by server, it will fork another
3486 thread ptlrpc_invalidate_import_thread to invalidate the
3487 import, where the grant will be set to 0. While the original
3488 thread will update the grant it got when connecting. So if
3489 the former happened latter, the grant will be wrongly errased
3490 because of this race.
3495 Description: Checking Stale with correct fid
3496 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3497 to check whether it is stale, because sometimes, we want the
3498 enqueue happened anyway, and op_data.fid2 will not be initialized.
3500 Severity : enhancement
3502 Description: update patchless client
3503 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3506 Frequency : only with 2.4 kernel
3508 Description: random memory corruption
3509 Details : size of struct ll_inode_info is to big for union inode.u and this
3510 can be cause of random memory corruption.
3515 Description: Memory leak in recovery
3516 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3517 It should also check obd_fail before fsfilt_start, otherwise if
3518 fsfilt_start return -EROFS,(failover mds during mds recovery).
3519 then the req will return with repmsg->transno = 0 and rc = EROFS.
3520 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3521 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3522 be freed no matter whether fsfilt_commit success or not.
3525 Frequency : only with huge count clients
3527 Description: Prevents from taking the superblock lock in llap_from_page for
3529 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3530 for llap_from_page call in ll_removepage prevents from taking the
3531 superblock lock for a soon died page.
3536 Description: Not check open intent error before release open handle
3537 Details : in some rare cases, the open intent error is not checked before
3538 release open handle, which may cause
3539 ASSERTION(open_req->rq_transno != 0), because it tries to release
3540 the failed open handle.
3545 Description: Set cat log bitmap only after create log success.
3546 Details : in some rare cases, the cat log bitmap is set too early. and it
3547 should be set only after create log success.
3551 Description: Accessing a block bevice can re-enable I/O when Lustre is
3552 tearing down a device.
3553 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3558 Description: service threads may hog cpus when there are a lot of requests
3559 Details : Insert cond_resched to give other threads a chance to use some CPU
3564 Description: the cat log was not initialized in recovery
3565 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3566 unlink log on mds will not be initialized until mds post
3567 recovery. And also in mds post recovery, the unlink log will
3568 initialization will be done asynchronausly, so there will be race
3569 between add unlink log and unlink log initialization.
3573 Description: brw_stats were being printed incorrectly
3574 Details : brw_stats were being printed as log2 but all of them were not
3575 recorded as log2. Also remove some code duplication arising from
3576 filter_tally_{read,write}.
3580 Frequency : rare, only in recovery.
3581 Description: ASSERTION(req->rq_type != LI_POISON) failed
3582 Details : imp_lock should be held while iterating over imp_sending_list for
3583 prevent destroy request after get timeout in ptlrpc_queue_wait.
3587 Description: replay-single.sh test 52 fails
3588 Details : A lock's skiplist need to be cleanup when it being unlinked
3589 from its resource list.
3593 Description: Short directio read returns full requested size rather than
3595 Details : Direct I/O operations should return actual amount of bytes
3596 transferred rather than requested size.
3598 Severity : enhancement
3600 Description: metadata RPC reduction (e.g. for rm performance)
3601 Details : decrease the amount of synchronous RPC between clients and servers
3602 by canceling conflicing lock before the operation on the client
3603 and packing thier handles into the main operation RPC to server.
3605 Severity : enhancement
3607 Description: Async OSC create to avoid the blocking unnecessarily.
3608 Details : If an OST has no remaining object, system will block on the
3609 creation when it needs to create a new object on this OST. Now,
3610 ways use pre-created objects when available, instead of blocking on
3611 an empty osc while others are not empty. If we must block, we
3612 block for the shortest possible period of time.
3616 Description: improve handling recoverable errors
3617 Details : if request processig with error which can be recoverable on server
3618 request should be resend, otherwise page released from cache and
3621 Severity : enhancement
3623 Description: refine locking for avoid write wrong info into lov_objid file
3624 Details : fix possible races with add new target and write/update data in
3627 --------------------------------------------------------------------------------
3629 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3634 Frequency : on some architectures
3636 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3637 Details : Uninitialized flags sometimes cause configuration commands to
3641 Frequency : patchless clients only
3643 Description: missing __iget() symbol export
3644 Details : The __iget() symbol export is missing. To avoid the need for
3645 this on patchless clients the deathrow inode reaper is turned
3646 off, and we depend on the VM to clean up old inodes. This
3647 dependency was during via the fix for bug 12181.
3649 --------------------------------------------------------------------------------
3651 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3653 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3654 INTEROPERATE with older versions automatically. In many cases a
3655 special upgrade step is needed. Please read the
3656 user documentation before upgrading any part of a live system.
3657 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3658 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3659 * WARNING: Lustre configuration and startup changes are required with
3660 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3664 Severity : enhancement
3666 Description: MountConf
3667 Details : Lustre configuration is now managed via mkfs and mount
3668 commands instead of lmc and lconf. New obd types (MGS, MGC)
3669 are added for dynamic configuration management. See
3670 https://mail.clusterfs.com/wikis/lustre/MountConf for
3673 Severity : enhancement
3675 Description: dynamic OST addition
3676 Details : OSTs can now be added to a live filesystem
3678 Severity : enhancement
3680 Description: startup order invariance
3681 Details : MDTs and OSTs can be started in any order. Clients only
3682 require the MDT to complete startup.
3684 Severity : enhancement
3686 Description: parallel, asynchronous orphan cleanup
3687 Details : orphan cleanup is now performed in separate threads for each
3688 OST, allowing parallel non-blocking operation.
3690 Severity : enhancement
3692 Description: optimized stripe assignment
3693 Details : stripe assignments are now made based on ost space available,
3694 ost previous usage, and OSS previous usage, in order to try
3695 to optimize storage space and networking resources.
3697 Severity : enhancement
3699 Description: Permanently set tunables
3700 Details : All writable /proc/fs/lustre tunables can now be permanently
3701 set on a per-server basis, at mkfs time or on a live system.
3703 Severity : enhancement
3705 Description: Lustre message v2
3706 Details : Add lustre message format v2.
3708 Severity : enhancement
3710 Description: client OST exclusion list
3711 Details : Clients can be started with a list of OSTs that should be
3712 declared "inactive" for known non-responsive OSTs.
3716 Description: SPEC SFS validation failure on NFS v2 over lustre.
3717 Details : Changes the blocksize for regular files to be 2x RPC size,
3718 and not depend on stripe size.
3720 Severity : enhancement
3722 Description: Multiple MD RPCs in flight.
3723 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3724 To avoid overly-overloading MDS, introduce a limit on number of
3725 MDS RPCs in flight for a single client and add /proc controls
3726 to adjust this limit.
3728 Severity : enhancement
3730 Description: client read/write statistics
3731 Details : Add client read/write call usage stats for performance
3732 analysis of user processes.
3733 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3734 file access. extents_stats shows chunk size distribution.
3735 extents_stats_per_process show chunk size distribution per
3738 Severity : enhancement
3740 Description: mds statistics
3741 Details : Add detailed mds operations statistics in
3742 /proc/fs/lustre/mds/*/stats.
3746 Description: Failure of copying files with lustre special EAs.
3747 Details : Client side always return success for setxattr call for lustre
3748 special xattr (currently only "trusted.lov").
3753 Description: Refcount LNET uuids
3754 Details : The global LNET uuid list grew linearly with every startup;
3755 refcount repeated list entries instead of always adding to
3758 Severity : enhancement
3760 Description: Dynamic service threads
3761 Details : Within a small range, start extra service threads
3762 automatically when the request queue builds up.
3765 Frequency : mixed-endian client/server environments
3767 Description: mixed-endian crashes
3768 Details : The new msg_v2 system had some failures in mixed-endian
3771 Severity : enhancement
3773 Description: Easy OST removal
3774 Details : OSTs can be permanently deactivated with e.g. 'lctl
3775 conf_param lustre-OST0001.osc.active=0'
3777 Severity : enhancement
3779 Description: MGS proc entries
3780 Details : Added basic proc entries for the MGS showing what filesystems
3783 Severity : enhancement
3785 Description: provide MGS failover
3786 Details : Added config lock reacquisition after MGS server failover.
3788 Severity : enhancement
3790 Description: add Linux 2.4 support
3791 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3795 Description: a large application tries to do I/O to the same resource and dies
3796 in the middle of it.
3797 Details : Check the req->rq_arrival time after the call to
3798 ost_brw_lock_get(), but before we do anything about
3799 processing it & sending the BULK transfer request. This
3800 should help move old stale pending locks off the queue as
3801 quickly as obd_timeout.
3804 Frequency : when an incorrect nid is specified during startup
3806 Description: ptlrpc connect to non-existant node causes kernel crash
3807 Details : LNET can't be re-entered from an event callback, which
3808 happened when we expire a message after the export has been
3809 cleaned up. Instead, hand the zombie cleanup off to another
3812 Severity : enhancement
3814 Description: plain/inodebits lock performance improvement
3815 Details : Grouping plain/inodebits in granted list by their request modes
3816 and bits policy, thus improving the performance of search through
3820 Frequency : only if OST filesystem is corrupted
3822 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3823 Details : for a short time RPCs with bulk IO are in the replay list,
3824 but replay of bulk IOs is unimplemented. If the OST filesystem
3825 is corrupted due to disk cache incoherency and then replay is
3826 started it is possible to trip an assertion. Avoid putting
3827 committed RPCs into the replay list at all to avoid this issue.
3830 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3832 Description: System hang on startup
3833 Details : This bug allowed the liblustre (e.g. catamount) client to
3834 return to the app before handling all startup RPCs. This
3835 could leave the node unresponsive to lustre network traffic
3836 and manifested as a server ptllnd timeout.
3838 Severity : enhancement
3840 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3841 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3842 boolean to control whether to print peer debug info when a
3843 client's RPC times out.
3846 Frequency : only for kernels with patches from Lustre below 1.4.3
3848 Description: Remove old rdonly API
3849 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3852 Frequency : only for devices with external journals
3854 Description: Set external device read-only also
3855 Details : During a commanded failover stop, we set the disk device
3856 read-only while the server shuts down. We now also set any
3857 external journal device read-only at the same time.
3860 Frequency : when upgrading from 1.4 while trying to change parameters
3862 Description: The wrong (new) MDC name was used when setting parameters for
3863 upgraded MDT's. Also allows changing of OSC (and MDC)
3864 parameters if --writeconf is specified at tunefs upgrade time.
3867 Frequency : when setting specific ost indicies
3869 Description: QOS code breaks on skipped indicies
3870 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3871 created with --index need not be sequential.
3873 Severity : enhancement
3875 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3876 Details : The uninit_groups feature works in conjunction with the kernel
3877 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3878 up the pass1 processing of e2fsck. This is a read-only feature
3879 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3880 mount filesystems that have had this feature enabled.
3882 Severity : enhancement
3884 Description: Improve multi-block allocation algorithm to avoid fragmentation
3885 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3886 allocation locality and avoid filesystem fragmentation.
3888 ------------------------------------------------------------------------------
3890 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3892 * Support for kernels:
3893 2.4.21-47.0.1.EL (RHEL 3)
3894 2.6.5-7.283 (SLES 9)
3895 2.6.9-42.0.10.EL (RHEL 4)
3896 2.6.12.6 vanilla (kernel.org)
3897 2.6.16.27-0.9 (SLES 10)
3898 * Recommended e2fsprogs version: 1.39.cfs5
3900 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3904 Frequency : occasional, depends on client load and configuration
3905 Bugzilla : 12181, 12203
3906 Description: data loss for recently-modified files
3908 Details : In some cases it is possible that recently written or created
3909 files may not be written to disk in a timely manner (this should
3910 normally be within 30s unless client IO load is very high).
3911 The problem appears as zero-length files or files that are a
3912 multiple of 1MB in size after a client crash or client eviction
3913 that are missing data at the end of the file.
3915 This problem is more likely to be hit on clients where files are
3916 repeatedly created and unlinked in the same directory, clients
3917 have a large amount of RAM, have many CPUs, the filesystem has
3918 many OSTs, the clients are rebooted frequently, and/or the files
3919 are not accessed by other nodes after being written.
3921 The presence of the problem can be detected by looking at
3922 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3923 smaller than the second (nr_unused) then dirty files will not
3924 be flushed automatically to disk. "sync; sleep 10" should be
3925 run several times on the node before unmounting it to update
3926 Lustre (this is also safe to run on nodes without this problem).
3928 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3929 that can cause this same problem, so customers using that kernel
3930 also need to update the kernel in addition to Lustre. In order
3931 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3933 It is normal that files written just before a client crash (less
3934 than 30s) may not yet have been flushed to disk, even for local
3938 Frequency : frequent on thin XT3 nodes
3940 Description: UUID collision on thin XT3 Linux nodes
3941 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3942 randomly, since we relied on an insufficiently-seeded PRNG.
3947 Description: OSS hangs after "All ost request buffers busy"
3948 Details : A deadlock between quota and journal operations caused OSS
3949 hangs after printing "All ost request buffers busy."
3952 Frequency : always on liblustre builds
3954 Description: Cleanup compiler warnings on liblustre
3957 Frequency : always on liblustre builds on XT3
3959 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3963 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3964 > 32000 subdirectories
3965 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3966 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3967 subdirectory count crosses 32000. This will aid e2fsck to
3968 correctly handle more than 32000 subdirectories.
3971 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3973 Description: System hang on startup
3974 Details : This bug allowed the liblustre (e.g. catamount) client to
3975 return to the app before handling all startup RPCs. This
3976 could leave the node unresponsive to lustre network traffic
3977 and manifested as a server ptllnd timeout.
3979 Severity : enhancement
3981 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3982 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3983 boolean to control whether to print peer debug info when a
3984 client's RPC times out.
3989 Description: make O_SYNC working on 2.6 kernels
3990 Details : 2.6 kernels use different method for mark pages for write,
3991 so need add a code to lustre for O_SYNC work.
3996 Description: Failure to close file and release space on NFS
3997 Details : Put inode details into lock acquired in ll_intent_file_open.
3998 Use mdc_intent_lock in ll_intent_open to properly
3999 detect all kind of errors unhandled by mdc_enqueue.
4004 Description: proc file read during shutdown sometimes raced obd removal,
4006 Details : Add lock to prevent obd access after proc file removal.
4009 Frequency : Only for files larger than 4GB on 32-bit clients.
4011 Description: improperly doing page alignment of locks
4012 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4013 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4018 Description: RPCs being resent when they shouldn't be
4019 Details : Some RPCs that should not be resent are being resent. This
4020 can cause inconsistencies in the RPC state machine. Do not
4021 resend such requests.
4024 Frequency : rare, only with NFS export
4026 Description: Crash on NFS re-export node
4027 Details : under very unusual load conditions an assertion is hit in
4028 ll_intent_file_open()
4031 Frequency : only if OST filesystem is corrupted
4033 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4034 Details : for a short time RPCs with bulk IO are in the replay list,
4035 but replay of bulk IOs is unimplemented. If the OST filesystem
4036 is corrupted due to disk cache incoherency and then replay is
4037 started it is possible to trip an assertion. Avoid putting
4038 committed RPCs into the replay list at all to avoid this issue.
4043 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4044 Details : Large single O_DIRECT read and write calls can fail to allocate
4045 a sufficiently large buffer to process the request. In case of
4046 allocation failure the allocation is retried with a smaller
4047 buffer and broken into smaller requests.
4049 Severity : enhancement
4051 Description: Add -o localflock option to simulate old noflock behaviour.
4052 Details : This will achieve local-only flock/fcntl locks coherentness.
4057 Description: versioning check is incomplete
4058 Details : Checking the version difference of client vs. server, report
4059 error if the gap is too big.
4064 Description: add support PG_writeback bit
4065 Details : add support for PG_writeback bit for Lustre, for more carefull
4066 work with page cache in 2.6 kernel. This also fix some deadlocks
4067 and remove hack for work O_SYNC with 2.6 kernel.
4069 Severity : enhancement
4071 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4072 Details : The uninit_groups feature works in conjunction with the kernel
4073 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4074 up the pass1 processing of e2fsck. This is a read-only feature
4075 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4076 mount filesystems that have had this feature enabled.
4078 Severity : enhancement
4080 Description: Improve multi-block allocation algorithm to avoid fragmentation
4081 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4082 allocation locality and avoid filesystem fragmentation.
4084 ------------------------------------------------------------------------------
4086 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4088 * Support for kernels:
4089 2.6.9-42.0.3.EL (RHEL 4)
4090 2.6.5-7.276 (SLES 9)
4091 2.4.21-47.0.1.EL (RHEL 3)
4092 2.6.12.6 vanilla (kernel.org)
4093 2.6.16.21-0.8 (SLES10)
4094 * Recommended e2fsprogs version: 1.39.cfs2-0
4096 * The backwards-compatible /proc/sys/portals symlink has been removed
4097 in this release. Before upgrading, please ensure that you change
4098 any configuration scripts or /etc/sysctl.conf files that access
4099 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4100 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4101 in advance of the upgrade on any system running Lustre 1.4.6 or
4102 newer, since /proc/sys/lnet was added in that version.
4103 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4107 Frequency : only when quota is used
4109 Description: avoid scanning export list for quota master
4110 Details : Change the algorithms to avoid scanning export list in order
4111 to improve the efficiency.
4114 Frequency : MDS failover only, very rarely
4116 Description: "went back in time" messages on mds failover
4117 Details : The greatest transno may be lost when the current operation
4118 finishes with an error (transno==0) and the client's last_rcvd
4119 record is over-written. Save the greatest transno in the
4120 mds_last_transno for this case.
4123 Frequency : always for specific kernels and striping counts
4125 Description: client may get "Matching packet too big" without ACL support
4126 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4127 when trying to access files in certain configurations. The
4128 clients should in fact be denied when mounting because they do
4129 not understand ACLs.
4132 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4134 Description: many clients connecting with IO in progress causes connect timeouts
4135 Details : Avoid synchronous journal commits to avoid delays caused by many
4136 clients connecting/disconnecting when bulk IO is in progress.
4137 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4138 OST_IO_PORTAL to avoid delays behind potentially many pending
4142 Frequency : occasionally with multiple writers to a single file
4144 Description: shared writes to file may result in wrong size reported by stat()
4145 Details : Allow growing of kms when extent lock is cancelled
4148 Frequency : always with random mmap IO to multi-striped file
4150 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4151 Details : Only if the hole is at the end of OST object so that kms is too
4152 small. Fix is to increase kms accordingly in ll_nopage.
4155 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4157 Description: writes to a missing object would leak memory on the OST
4158 Details : If there is an inconsistency between the MDS and OST filesystems,
4159 such that the MDS references an object that doesn't exist, writes
4160 to that object will leak memory due to incorrect cleanup in the
4161 error handling path, eventually running out of memory on the OST.
4166 Description: Creating too long symlink causes lustre errors
4167 Details : Check symlink and name lengths before sending requests to MDS.
4170 Frequency : only if flock is enabled (not on by default)
4172 Description: posix locks not released on fd closure on 2.6.9+
4173 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4174 kernels, this caused such locks not to be released on fd close and
4175 then assertions on fs unmount about still used locks.
4178 Frequency : MDS failover only, very rarely
4180 Description: clients may get ASSERTION(granted_lock != NULL)
4181 Details : When request was taking a long time, and a client was resending
4182 a getattr by name lock request. The were multiple lock requests
4183 with the same client lock handle and
4184 mds_getattr_name->fixup_handle_for_resent_request found one of the
4185 lock handles but later failed with ASSERTION(granted_lock != NULL).
4190 Description: handle->h_buffer_credits > 0, assertion failure
4191 Details : h_buffer_credits is zero after truncate, causing assertion
4192 failure. This patch extends the transaction or creates a new
4196 Frequency : NFS re-export or patchless client
4197 Bugzilla : 11179, 10796
4198 Description: Crash on NFS re-export node (__d_move)
4199 Details : We do not want to hash the dentry if we don't have a lock.
4200 But if this dentry is later used in d_move, we'd hit uninitialised
4201 list head d_hash, so we just do this to init d_hash field but
4202 leave dentry unhashed.
4205 Frequency : NFS re-export or patchless client
4207 Description: NFS exports has problem with symbolic link
4208 Details : lustre client didn't properly install dentry when re-exported
4209 to NFS or running patchless client.
4212 Frequency : NFS re-export or patchless client
4214 Description: Various nfs/patchless fixes.
4215 Details : fixes reuse disconected alias for lookup process - this fixes
4216 warning "find_exported_dentry: npd != pd",
4217 fix permission error with open files at nfs.
4218 fix apply umask when do revalidate.
4221 Frequency : occasional
4223 Description: Crash on NFS re-export node
4224 Details : calling clear_page() on the wrong pointer triggered oops in
4225 generic_mapping_read().
4228 Frequency : rarely, using O_DIRECT IO
4230 Description: unaligned directio crashes client with LASSERT
4231 Details : check for unaligned buffers before trying any requests.
4234 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4236 Description: stale data returned from RAID cache
4237 Details : If only a small amount of IO is done to the RAID device before
4238 reading it again it is possible to get stale data from the RAID
4239 cache instead of reading it from disk.
4242 Frequency : always for sles10 kernel
4244 Description: sles10 support
4245 Details : ll_follow_link: compile fixes and using of nd_set_link
4246 under newer kernels.
4249 Frequency : depends on arch, kernel and compiler version, always on sles10
4252 Description: recursive or deep enough symlinks cause stack overflow
4253 Details : getting rid of large stack-allocated variable in
4257 Frequency : depends on hardware
4259 Description: lustre write performance loss in the SLES10 kernel
4260 Details : the performance loss is caused by using of write barriers in the
4261 ext3 code. The SLES10 kernel turns barrier support on by
4262 default. The fix is to undo that change for ldiskfs.
4264 ------------------------------------------------------------------------------
4266 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4268 * Support for kernels:
4269 2.6.9-42.0.3EL (RHEL 4)
4270 2.6.5-7.276 (SLES 9)
4271 2.4.21-47.0.1.EL (RHEL 3)
4272 2.6.12.6 vanilla (kernel.org)
4276 Frequency : quota enabled and large files being deleted
4278 Description: releasing more than 4GB of quota at once hangs OST
4279 Details : If a user deletes more than 4GB of files on a single OST it
4280 will cause the OST to spin in an infinite loop. Release
4281 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4286 Description: statfs data retrieved from /proc may be stale or zero
4287 Details : When reading per-device statfs data from /proc, in the
4288 {kbytes,files}_{total,free,avail} files, it may appear
4289 as zero or be out of date.
4292 Frequency : systems with MD RAID1 external journal devices
4294 Description: lconf's call to blkid is confused by RAID1 journal devices
4295 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4296 of returning all block devices that match the journal UUID.
4299 Frequency : always, for aggregate stripe size over 4GB
4301 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4302 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4303 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4306 Frequency : always if "lfs find" used on a local file/directory
4308 Description: "lfs find" segfaults if used on a local file/directory
4309 Details : The case where a directory component was not specified wasn't
4310 handled correctly. Handle this properly.
4313 Frequency : always on ppc64
4315 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4316 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4317 perform endianness conversion.
4320 Frequency : rarely (truncate to non-zero file size after write under load)
4321 Bugzilla : 10730, 10687
4322 Description: Files padded with zeros to next 4K multiple
4323 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4324 it is possible that files that are truncated to a non-zero size
4325 immediately after being written are filled with zero bytes beyond
4326 the truncated size. No file data is lost.
4328 Severity : enhancement
4330 Description: Allow recovery/failover for liblustre clients.
4331 Details : liblustre clients were unaware of failover configurations until
4334 Severity : enhancement
4336 Description: user file locks should fail when not mounting with flock option
4337 Details : Set up an error-returning stub in ll_file_operations.lock field
4338 to prevent incorrect behaviour when client is mounted without
4339 flock option. Also, set up properly f_op->flock field for
4343 Frequency : always on ia64
4345 Description: "lfs df" loops on printing out MDS statfs information
4346 Details : The obd_ioctl_data was not initialized and in some systems
4347 this caused a failure during the ioctl that did not return
4348 an error. Initialize the struct and return an error on failure.
4351 Frequency : SLES 9 only
4353 Description: Error of copying files with lustre special EAs as root
4354 Details : Client side always return success for setxattr call for lustre
4355 special xattr (currently only "trusted.lov").
4358 Frequency : rarely on clusters with both ia64+i386 clients
4360 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4361 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4362 clients are doing concurrent writes to the same file it is
4363 possible that the ia64 clients may LASSERT because the OST
4364 extent locks are not PAGE_SIZE aligned. Ensure that grown
4365 locks are always aligned on the request boundary.
4368 Frequency : specific use, occasional
4370 Description: Overwriting in use executable truncates on-disk binary image
4371 Details : If one node attempts to overwrite an executable in use by
4372 another node, we now correctly return ETXTBSY instead of
4373 truncating the file.
4375 Severity : enhancement
4377 Description: Async OSC create to avoid the blocking unnecessarily.
4378 Details : If an OST has no remaining object, system will block on the creating
4379 when need to create a new object on this OST. Now, ways use
4380 pre-created objects when available, instead of blocking on an
4381 empty osc while others are not empty. If we must block, we block
4382 for the shortest possible period of time.
4387 Description: chmod on Lustre root is propagated to other clients
4388 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4389 invalid by the follow_mount time.
4392 Frequency : liblustre clients only
4394 Description: Race in 'instant cancel' lock handling could lead to such locks
4395 never to be granted in case of SMP MDS
4396 Details : Do not destroy not yet granted but cbpending locks in
4400 Frequency : replay/resend of open
4402 Description: non null lock assetion failure in mds_intent_policy
4403 Details : Trying to replay/resend lockless open requests resulted in
4404 mds_open() returning 0 with no lock. Now it sets a flag if
4405 a lock is going to be returned.
4407 Severity : enhancement
4409 Description: Checksum enhancements
4410 Details : New checksum enhancements allow for resending RPCs that failed
4413 Severity : enhancement
4415 Description: Tunables on number of dirty pages in cacche
4416 Details : Allow to set limit on number of dirty pages cached.
4421 Description: client crash on unmount - lock still has references
4422 Details : In some error handling cases it was possible to leak a lock
4423 reference on a client while accessing a file. This was not
4424 harmful to the client during operation, but would cause the
4425 client to crash when the filesystem is unmounted.
4428 Frequency : specific case, rare
4430 Description: ETXTBSY on mds though file not in use by client
4431 Details : ETXTBSY is no longer incorrectly returned when attempting to
4432 chmod or chown a directory that the user previously tried to
4433 execute or a currently-executing binary.
4436 Frequency : extremely rare except on liblustre-based clients
4438 Description: Lustre space not freed when files are deleted
4439 Details : Clean up open-unlinked files after client eviction. Previously
4440 the unlink was skipped and the files remained as orphans.
4445 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4446 Details : In some cases it was possible to send a blocking callback to
4447 a client doing a glimpse, even though that client didn't get
4448 a lock granted. When the glimpse lock is cancelled on the OST
4449 the freed lock is left on the waiting list and corrupted the list.
4452 Frequency : all core dumps
4454 Description: Broke core dumps to lustre
4455 Details : Negative dentry may be unhashed if parent does not have UPDATE
4456 lock, but some callers, e.g. do_coredump, expect dentry to be
4457 hashed after successful create, hash it in ll_create_it.
4459 ------------------------------------------------------------------------------
4461 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4463 * Support for kernels:
4464 2.6.9-42.0.2.EL (RHEL 4)
4465 2.6.5-7.276 (SLES 9)
4466 2.4.21-40.EL (RHEL 3)
4467 2.6.12.6 vanilla (kernel.org)
4471 Frequency : always on RHEL 3
4473 Description: Number of open files grows over time
4474 Details : The number of open files grows over time, whether or not
4475 Lustre is started. This was due to a filp leak introduced
4476 by one of our kernel patches.
4478 ------------------------------------------------------------------------------
4480 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4482 * Support for kernels:
4483 2.6.9-42.EL (RHEL 4)
4484 2.6.5-7.276 (SLES 9)
4485 2.4.21-40.EL (RHEL 3)
4486 2.6.12.6 vanilla (kernel.org)
4491 Bugzilla : 5719, 9635, 9792, 9684
4492 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4493 Details : If a server is under heavy load and cannot reply to new
4494 connection requests before the client resends the (re)connect,
4495 the connection handling code can behave badly if two service
4496 threads are concurrently handing separate (re)connections from
4497 the same client. Add better locking to the connection handling
4498 code, and ensure that only a single connection will be processed
4499 for a given client UUID, even if the lock is dropped.
4501 Severity : enhancement
4503 Description: add TCP zero-copy support to kernel
4504 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4505 sends if the hardware supports scatter-gather and checksumming.
4506 This allows socklnd to do client-write and server-read more
4507 efficiently and reduce CPU utilization from skbuf copying.
4510 Frequency : only if NFS exporting from client
4512 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4513 Details : If setting attributes on a file created under NFS that had
4514 never been opened it would be possible to oops the client
4515 if the file had no objects.
4518 Frequency : always for liblustre
4520 Description: liblustre client does MDS+OSTs setattr RPC for each write
4521 Details : When doing a write from a liblustre client, the client
4522 incorrectly issued an RPC to the MDS and each OST the file was
4523 striped over in order to update the timestamps. When writing
4524 with small chunks and many clients this could overwhelm the MDS
4525 with RPCs. In all cases it would slow down the write because
4526 these RPCs are unnecessary.
4528 Severity : enhancement
4530 Description: allow number of MDS service threads to be changed at module load
4531 Details : It is now possible to change the number of MDS service threads
4532 running. Adding "options mds mds_num_threads={N}" to the MDS's
4533 /etc/modprobe.conf will set the number of threads for the next
4534 time Lustre is restarted (assuming the "mds" module is also
4535 reloaded at that time). The default number of threads will
4536 stay the same, 32 for most systems.
4541 Description: OST crash if filesystem is unformatted or corrupt
4542 Details : If an OST is started on a device that has never been formatted
4543 or if the filesystem is corrupt and cannot even mount then the
4544 error handling cleanup routines would dereference a NULL pointer.
4549 Description: NULL pointer deref in llap_from_page.
4550 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4551 incorrect) mapping if the page was truncated/reclaimed while it was
4552 searched for. Check for this condition and skip such pages when
4553 doing readahead. Introduce extra check to llap_from_page() to
4554 verify page->mapping->host is non-NULL (so page is not anonymous).
4557 Frequency : Sometimes when using sys_sendfile
4559 Description: "page not covered by a lock" warnings from ll_readpage
4560 Details : sendfile called ll_readpage without right page locks present.
4561 Now we introduced ll_file_sendfile that does necessary locking
4562 around call to generic_file_sendfile() much like we do in
4566 Frequency : with certain MDS communication failures at client mount time
4568 Description: NULL pointer deref after failed client mount
4569 Details : a client connection request may delayed by the network layer
4570 and not be sent until after the PTLRPC layer has timed out the
4571 request. If the client fails the mount immediately it will try
4572 to clean up before the network times out the request. Add a
4573 reference from the request import to the obd device and delay
4574 the cleanup until the network drops the request.
4577 Frequency : occasionally during client (re)connect
4579 Description: assertion failure during client (re)connect
4580 Details : processing a client connection request may be delayed by the
4581 client or server longer than the client connect timeout. This
4582 causes the client to resend the connection request. If the
4583 original connection request is replied in this interval, the
4584 client may trip an assertion failure in ptlrpc_connect_interpret()
4585 which thought it would be the only running connect process.
4588 Frequency : only with obd_echo servers and clients that are rebooted
4590 Description: kernel BUG accessing uninitialized data structure
4591 Details : When running an obd_echo server it did not start the ping_evictor
4592 thread, and when a client was evicted an uninitialized data
4593 structure was accessed. Start the ping_evictor in the RPC
4594 service startup instead of the OBD startup.
4596 Severity : enhancement
4597 Bugzilla : 10193 (patchless)
4598 Description: Remove dependency on various unexported kernel interfaces.
4599 Details : No longer need reparent_to_init, exit_mm, exit_files,
4600 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4603 Frequency : rare (only users of deprecated and unsupported LDAP config)
4605 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4606 Details : LDAP apparently contains 'lustreName' attributes instead of
4607 'name'. A simple remapping of the name is sufficient.
4610 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4612 Description: waiting_locks_callback trips kernel BUG if client is evicted
4613 Details : Running with the dump_on_timeout debug flag turned on makes
4614 it possible that the waiting_locks_callback() can try to dump
4615 the Lustre kernel debug logs from an interrupt handler. Defer
4616 this log dumping to the expired_lock_main() thread.
4618 Severity : enhancement
4620 Description: Support NFS exporting on 2.6 kernels.
4621 Details : Implement non-rawops metadata methods for NFS server to use without
4622 changing NFS server code.
4625 Frequency : very rare (synthetic metadata workload only)
4627 Description: two racing renames might cause an MDS thread to deadlock
4628 Details : Running the "racer" program may cause one MDS thread to rename
4629 a file from being the source of a rename to being the target of
4630 a rename at exactly the same time that another thread is doing
4631 so, and the second thread has already enqueued these locks after
4632 doing a lookup of the target and is trying to relock them in
4633 order. Ensure that we don't try to re-lock the same resource.
4636 Frequency : only very large systems with liblustre clients
4638 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4639 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4640 all OSTs in parallel. This allows the eviction of stale liblustre
4641 clients to proceed much faster than if they were done in series,
4642 and also offers similar improvements for other set_info RPCs.
4647 Description: excessive CPU usage during initial read phase on client
4648 Details : During the initial read phase on a client, it would agressively
4649 retry readahead on the file, consuming too much CPU and impacting
4650 performance (since 1.4.5.8). Improve the readahead algorithm
4651 to avoid this, and also improve some other common cases (read
4652 of small files in particular, where "small" is files smaller than
4653 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4658 Description: MDS crash when receiving packet with unknown intent.
4659 Details : Do not LBUG in unknown intent case, just return -EFAULT
4661 Severity : enhancement
4662 Bugzilla : 9293, 9385
4663 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4664 Details : Do not serialize getattr (non-intent version) and statfs.
4667 Frequency : occasional, when OST network is overloaded/intermittent
4669 Description: client evicted by OST after bulk IO timeout
4670 Details : If a client sends a bulk IO request (read or write) the OST
4671 may evict the client if it is unresposive to its data GET/PUT
4672 request. This is incorrect if the network is overloaded (takes
4673 too long to transfer the RPC data) or dropped the OST GET/PUT
4674 request. There is no need to evict the client at all, since
4675 the pinger and/or lock callbacks will handle this, and the
4676 client can restart the bulk request.
4679 Frequency : Always when mmapping file with no objects
4681 Description: client crashes when mmapping file with no objects
4682 Details : Check that we actually have objects in a file before doing any
4683 operations on objects in ll_vm_open, ll_vm_close and
4689 Description: Request leak when working with deleted CWD
4690 Details : Introduce advanced request refcount tracking for requests
4691 referenced from lustre intent.
4693 Severity : Enhancement
4695 Description: Cache open file handles on client.
4696 Details : MDS now will return special lock along with openhandle, if
4697 requested and client is allowed to hold openhandle, even if unused,
4698 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4699 closing files for every read/write openration.
4701 Severity : Enhancement
4703 Description: Cache open negative dentries on client when possible.
4704 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4705 negative dentries on lock revocation.
4710 Description: Remounting a client read-only wasn't possible with a zconf mount
4711 Details : It wasn't possible to remount a client read-only with llmount.
4713 Severity : enhancement
4714 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4715 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4716 MPICH is included to add Lustre-specific ADIO interfaces.
4717 This is based closely on the UFS ADIO layer and only differs
4718 in file creation, in order to allow the OST striping to be set.
4719 This is user-contributed code and not supported by CFS.
4724 Description: extended inode attributes (immutable, append-only) work improperly
4725 when 2.4 and 2.6 kernels are used on client/server or vice versa
4726 Details : Introduce kernel-independent values for these flags.
4728 Severity : enhancement
4731 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4732 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4733 granularity. Now, Fractional MB granularity is be supported,
4734 this is very useful for low memory system.
4736 Severity : enhancement
4738 Description: Getattr by fid
4739 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4740 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4741 have locks and avoids getattr rpc requests for it on every stat.
4744 Frequency : Always, for filesystems larger than 2TB
4746 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4747 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4748 filesystems larger than 2TB to be mounted with mballoc enabled.
4751 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4753 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4754 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4755 filesystems larger than 2TB is mounted then the kernel will
4756 silently corrupt the start of the filesystem. CONFIG_LBD is
4757 enabled for all CFS-supported kernels, but the possibility of
4758 this happening with a modified kernel config exists.
4760 Severity : enhancement
4762 Description: add client O_DIRECT support for 2.6 kernels
4763 Details : It is now possible to do O_DIRECT reads and writes to files
4764 in the Lustre client mountpoint on 2.6 kernel clients.
4766 Severity : enhancement
4768 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4769 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4770 parallel, not waiting for response from every OST before sending
4771 a rpc to the next OST.
4776 Description: setattr vs write race when updating file timestamps
4777 Details : Client processes that update a file timestamp into the past
4778 right after writing to the file (e.g. tar) it is possible that
4779 the updated file modification time can be reset to the current
4780 time due to a race between processing the setattr and write RPC.
4782 Severity : enhancement
4784 Description: Bring 'lfs find' closer in line with regular Linux find.
4785 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4786 -print0 options and obtains all the needed info through the lustre
4789 Severity : enhancement
4791 Description: support up to 1024 configured devices on one node
4792 Details : change obd_dev array from statically allocated to dynamically
4793 allocated structs as they are first used to reduce memory usage
4798 Description: Flush dirty partially truncated pages during truncate
4799 Details : Immediatelly flush partially truncated pages in filter_setattr,
4800 this way we completely avoid having any pages in page cache on OST
4801 and can retire ugly workarounds during writes to flush such pages.
4806 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4807 Details : protect inode from truncation within vfs_unlink() context
4808 just take a reference before calling vfs_unlink() and release it
4809 when parent's i_sem is free.
4814 Description: last_id value checked outside lock on OST caused LASSERT failure
4815 Details : If there were multiple MDS->OST object precreate requests in
4816 flight, it was possible that the OST's last object id was checked
4817 outside a lock and incorrectly tripped an assertion. Move checks
4818 inside locks, and discard old precreate requests.
4821 Frequency : always, if extents are used on OSTs
4823 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4824 Details : OSTs using the extents format would not zero the high 16 bits of
4825 the index physical block number. This is not a problem for any
4826 OST filesystems smaller than 16TB, and no kernels support ext3
4827 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4828 new/modified files) and can be fixed for existing filesystems
4829 with e2fsprogs-1.39-cfs1.
4834 Description: import connection selection may be incorrect if timer wraps
4835 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4836 import connections to be ignored if the 32-bit jiffies counter
4837 wraps. Use a 64-bit jiffies counter.
4840 Frequency : very large clusters immediately after boot
4842 Description: LNET request buffers exhausted under heavy short-term load
4843 Details : If a large number of client requests are generated on a service
4844 that has previously never seen so many requests it is possible
4845 that the request buffer growth cannot keep up with the spike in
4846 demand. Instead of dropping incoming requests, they are held in
4847 the LND until the RPC service can accept more requests.
4850 Frequency : Sometimes during replay
4852 Description: Assertion failure in ll_local_open after replay.
4853 Details : If replay happened on an open request reply before we were able
4854 to set replay handler, reply will become not swabbed tripping the
4855 assertion in ll_local_open. Now we set the handler right after
4856 recognising of open request
4859 Frequency : very rare
4861 Description: kernel reports "badness in vsnprintf"
4862 Details : Reading from the "recovery_status" /proc file in small chunks
4863 may cause a negative length in lprocfs_obd_rd_recovery_status()
4864 call to vsnprintf() (which is otherwise harmless). Exit early
4865 if there is no more space in the output buffer.
4867 Severity : enhancement
4869 Description: clear OBD RPC statistics by writing to them
4870 Details : It is now possible to clear the OBD RPC statistics by writing
4871 to the "stats" file.
4876 Description: Client mtime is not the same on different clients after utimes
4877 Details : In some cases, the client was using the utimes() syscall on
4878 a file cached on another node. The clients now validate the
4879 ctime from the MDS + OSTs to determine which one is right.
4884 Description: Inability to activate failout mode
4885 Details : lconf script incorrectly assumed that in pythong string's numeric
4886 value is used in comparisons.
4889 Frequency : always with multiple stripes per file
4891 Description: Inefficient object allocation for mutli-stripe files
4892 Details : When selecting which OSTs to stripe files over, for files with
4893 a stripe count that divides evenly into the number of OSTs,
4894 the MDS is always picking the same starting OST for each file.
4895 Return the OST selection heuristic to the original design.
4900 Description: mount failures may take full timeout to return an error
4901 Details : Under some heavy load conditions it is possible that a
4902 failed mount can wait for the full obd_timeout interval,
4903 possibly several minutes, before reporting an error.
4904 Instead return an error as soon as the status is known.
4906 Frequency : quota enabled and large files being deleted
4908 Description: releasing more than 4GB of quota at once hangs OST
4909 Details : If a user deletes more than 4GB of files on a single OST it
4910 will cause the OST to spin in an infinite loop. Release
4911 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4916 Description: statfs data retrieved from /proc may be stale or zero
4917 Details : When reading per-device statfs data from /proc, in the
4918 {kbytes,files}_{total,free,avail} files, it may appear
4919 as zero or be out of date.
4922 Frequency : systems with MD RAID1 external journal devices
4924 Description: lconf's call to blkid is confused by RAID1 journal devices
4925 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4926 of returning all block devices that match the journal UUID.
4929 Frequency : always, for aggregate stripe size over 4GB
4931 Description: assertion fails when trying to use 4GB stripe size
4932 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4933 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4936 Frequency : always on ppc64
4938 Description: the first write on an ext3 filesystem with mballoc got stuck
4939 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4940 perform endianness conversion.
4942 ------------------------------------------------------------------------------
4944 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4946 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4947 INTEROPERATE with older versions automatically. Please read the
4948 user documentation before upgrading any part of a live system.
4949 * WARNING: Lustre networking configuration changes are required with
4950 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4953 * Support for kernels:
4954 2.6.9-22.0.2.EL (RHEL 4)
4955 2.6.5-7.244 (SLES 9)
4956 2.6.12.6 vanilla (kernel.org)
4959 Severity : enhancement
4960 Bugzilla : 7981/8208
4961 Description: Introduced Lustre Networking (LNET)
4962 Details : LNET is new networking infrastructure for Lustre, it includes
4963 a reorganized network configuration mode (see the user
4964 documentation for full details) as well as support for routing
4965 between different network fabrics. Lustre Networking Devices
4966 (LNDs) for the supported network fabrics have also been
4967 created for this new infrastructure.
4969 Severity : enhancement
4970 Description: Introduced Access control lists
4971 Details : clients can set ACLs on files and directories in order to have
4972 more fine-grained permissions than the standard Unix UGO+RWX.
4973 The MDS must be started with the "-o acl" mount option.
4975 Severity : enhancement
4976 Description: Introduced filesystem quotas
4977 Details : Administrators may now establish per-user quotas on the
4980 Severity : enhancement
4982 Description: Configuration change for the XT3
4983 The PTLLND is now used to run Lustre over Portals on the XT3
4984 The configure option(s) --with-cray-portals are no longer used.
4985 Rather --with-portals=<path-to-portals-includes> is used to
4986 enable building on the XT3. In addition to enable XT3 specific
4987 features the option --enable-cray-xt3 must be used.
4992 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4993 Details : A race between allocating a new llcd and re-getting the llcd_lock
4994 allowed another thread to grab newly-allocated llcd.
4996 Severity : enhancement
4998 Description: 2.6 OST async journal commit and locking fix to improve performance
4999 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5000 2.6 kernels are now async as they already were in 2.4 kernels so
5001 that they can commit concurrently with the network bulk transfer.
5002 For block-allocated files the filter allocation semaphore is held
5003 to avoid filesystem fragmentation during allocation. BKL lock
5004 removed for 2.6 xattr operations where it is no longer needed.
5009 Description: lconf incorrectly determined whether two IP networks could talk
5010 Details : In some more complicated routing and multiple-network
5011 configurations, lconf will avoid trying to make a network
5012 connection to a disjoint part of the IP space. It was doing the
5013 math incorrectly for one set of cases.
5018 Description: Fix for potential infinite loop processing records in an llog.
5019 Details : If an llog record is corrupted/zeroed, it is possible to loop
5020 forever in llog_process(). Validate the llog record length
5021 and skip the remainder of the block on error.
5024 Frequency : occasional (liblustre only)
5026 Description: liblustre could not open files whose last component is a symlink
5027 Details : sysio_path_walk() would incorrectly pass the open intent to
5028 intermediate path components.
5031 Frequency : rare (liblustre only with non-standard tuning)
5032 Bugzilla : 7201 (7350)
5033 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5034 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5035 after it has been released, only do this for referenced locks
5037 Severity : enhancement
5039 Description: specifying an (invalid) directory default stripe_size of -1
5040 would reset the directory default striping
5041 Details : stripe_size -1 was used internally to signal directory stripe
5042 removal, now use "all default" to signal dir stripe removal
5043 as a directory striping of "all default" is not useful
5046 Frequency : common for large clusters running liblustre clients
5048 Description: doing an ls when liblustre clients are running is slow
5049 Details : sending a glimpse AST to a liblustre client waits for every AST
5050 to time out, as liblustre clients will not respond. Since they
5051 cannot cache data we refresh the OST lock LVB from disk instead.
5053 Severity : enhancement
5055 Description: doing an ls at the same time as file IO can be slow
5056 Details : enqueue and other "small" requests can be blocked behind many
5057 large IO requests. Create a new OST IO portal for non-IO
5058 requests so they can be processed faster.
5061 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5063 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5064 created file with mode 000
5065 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5066 allow an ftruncate() as a non-root user to a file with mode 000.
5067 Lustre now respects this flag to disable mode checks when
5068 truncating a file owned by the user
5071 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5074 Description: Revoking locks from clients that went dead or catatonic might take
5076 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5077 cancellation of such locks instant on servers without waiting for
5078 any reply from clients. Clients drops these locks when cancel
5079 notification from server is received without replying.
5082 Frequency : liblustre-only, when liblustre client dies or becomes busy
5084 Description: Doing ls on Linux clients can take a long time with active
5086 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5087 granting such locks to it in the first place if possible. Locks
5088 are taken by proxy on the OST during the read or write and
5089 dropped immediately afterward. Add connect flags handling, do
5090 not grant locks to liblustre clients for glimpse ASTs.
5092 Severity : enhancement
5094 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5095 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5096 sequential reads and behaves badly with applications doing random
5097 reads. Improve it by reading ahead at least read region, and
5098 avoiding excessive large RPC for small reads.
5100 Severity : enhancement
5102 Description: Creating more than 1000 files for a single job may cause a load
5103 imbalance on the OSTs if there are also a large number of OSTs.
5104 Details : qos_prep_create() uses an OST index reseed value that is an
5105 even multiple of the number of available OSTs so that if the
5106 reseed happens in the middle of the object allocation it will
5107 still utilize the OSTs as uniformly as possible.
5112 Description: OST or MDS may oops in ping_evictor_main()
5113 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5114 but doesn't restart at beginning of obd_exports_timed list
5117 Severity : enhancement
5119 Description: improve by-nid export eviction on the MDS and OST
5120 Details : allow multiple exports with the same NID to be evicted at one
5121 time without re-searching the exports list.
5124 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5126 Description: MDS may oops in groups_free()
5127 Details : in rare race conditions a newly allocated group_info struct is
5128 freed again, and this can be NULL. The 2.4 compatibility code
5129 for groups_free() checked for a NULL pointer, but 2.6 did not.
5132 Frequency : common for liblustre clients doing little filesystem IO
5133 Bugzilla : 9352, 7313
5134 Description: server may evict liblustre clients accessing contended locks
5135 Details : if a client is granted a lock or receives a completion AST
5136 with a blocking AST already set it would not reply to the AST
5137 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5140 Frequency : lfs setstripe, only systems with more than 160 OSTs
5142 Description: unable to set striping with a starting offset beyond OST 160
5143 Details : llapi_create_file() incorrectly limited the starting stripe
5144 index to the maximum single-file stripe count.
5147 Frequency : LDAP users only
5149 Description: lconf did not handle in-kernel recovery with LDAP properly
5150 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5152 Severity : enhancement
5154 Description: bind OST threads to NUMA nodes to improve performance
5155 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5156 node and do their allocations there to localize memory access
5158 Severity : enhancement
5160 Description: llmount can determine client NID directly from Myrinet (GM)
5161 Details : the client NID code from gmnalnid was moved directly into
5162 llmount, removing the need to use this or specifying the
5163 client NID explicitly when mounting GM clients with zeroconf
5166 Frequency : if client is started with down MDS
5168 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5169 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5170 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5171 ptlrpc_import_delayed_req() because we didn't check if the
5172 request was interrupted, and we also didn't break out of the
5173 event loop if there was no timeout
5178 Description: data loss during non-page-aligned writes to a single file from
5179 both multiple nodes and multiple threads on one node at same time
5180 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5181 inconsistency led to false short-reads, that were cached and later
5182 used by ->prepare_write() to fill in partially written page,
5183 leading to data loss.
5186 Frequency : always, if lconf --abort_recovery used
5188 Description: lconf --abort_recovery fails with 'Operation not supported'
5189 Details : lconf was attempting to abort recovery on the MDT device and not
5192 Severity : enhancement
5194 Description: remove cleanup logs
5195 Details : replace lconf-generated cleanup logs with lustre internal
5196 cleanup routines. Eliminates the need for client-cleanup and
5199 Severity : enhancement
5201 Description: add support for EAs (user and system) on lustre filesystems
5202 Details : it is now possible to store extended attributes in the Lustre
5203 client filesystem, and with the user_xattr mount option it
5204 is possible to allow users to store EAs on their files also
5206 Severity : enhancement
5208 Description: Add possibility (config option) to show minimal available OST free
5210 Details : When compiled with --enable-mindf configure option, statfs(2)
5211 (and so, df) will return least minimal free space available from
5212 all OSTs as amount of free space on FS, instead of summary of
5213 free spaces of all OSTs.
5215 Severity : enhancement
5217 Description: do not expand extent locks acquired on OST-side
5218 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5219 by server: they are not cached anyway.
5222 Frequency : when mmap is used/binaries executed from Lustre
5224 Description: Unmmap pages before throwing them away from read cache.
5225 Details : llap_shrink cache now attempts to unmap pages before discarding
5226 them (if unmapping failed - do not discard). SLES9 kernel has
5227 extra checks that trigger if this unmapping is not done first.
5232 Description: lconf didn't resolve symlinks before checking to see whether a
5233 given mountpoint was already in use
5236 Frequency : when migrating failover services
5237 Bugzilla : 6395, 9514
5238 Description: When migrating a subset of services from a node (e.g. failback
5239 from a failover service node) the remaining services would
5240 time out and evict clients.
5241 Details : lconf --force (implied by --failover) sets the global obd_timeout
5242 to 5 seconds in order to quickly disconnect, but this caused
5243 other RPCs to time out too quickly. Do not change the global
5244 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5246 Severity : enhancement
5247 Frequency : if MDS is started with down OST
5248 Bugzilla : 9439,5706
5249 Description: Allow startup/shutdown of an MDS without depending on the
5250 availability of the OSTs.
5251 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5252 Add appropriate locking and lov-osc refcounts for safe
5253 cleaning. Add osc abort_inflight calls in case the
5254 synchronize never started.
5257 Frequency : occasional (Cray XT3 only)
5259 Description: root not authorized to access files in CRAY_PORTALS environment
5260 Details : The client process capabilities were not honoured on the MDS in
5261 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5262 been accessed by an authorized user then root was able to access
5263 the file on the local client also. The root user capabilities
5264 are now allowed on the MDS, as this environment has secure UID.
5267 Frequency : occasional
5269 Description: ldiskfs "too long searching" message happens too often
5270 Details : A debugging message (otherwise harmless) prints too often on
5271 the OST console. This has been reduced to only happen when
5272 there are fragmentation problems on the filesystem.
5277 Description: Division by zero in statfs when all OSCs are inactive
5278 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5279 lov_statfs divides by value returned by lov_get_stripecnt().
5283 Bugzilla : 9489, 3273
5284 Description: First write from each client to each OST was only 4kB in size,
5285 to initialize client writeback cache, which caused sub-optimal
5286 RPCs and poor layout on disk for the first written file.
5287 Details : Clients now request an initial cache grant at (re)connect time
5288 and so that they can start streaming writes to the cache right
5289 away and always do full-sized RPCs if there is enough data.
5290 If the OST is rebooted the client also re-establishes its grant
5291 so that client cached writes will be honoured under the grant.
5296 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5297 Details : Now I/O RPCs go to different portal number and (presumably) fast
5298 lock requests (and glimses) and other RPCs get their own service
5299 threads pool that should be able to service those RPCs
5302 Severity : enhancement
5304 Description: Ability to exchange lustre version between client and servers and
5305 issue warnings at client side if client is too old. Also for
5306 liblustre clients there is ability to refuse connection of too old
5308 Details : New 'version' field is added to connect data structure that is
5309 filled with version info. That info is later checked by server and
5313 Frequency : rare, liblustre only.
5314 Bugzilla : 9296, 9581
5315 Description: Two simultaneous writes from liblustre at offset within same page
5316 might proceed at the same time overwriting eachother with stale
5318 Details : I/O lock withing llu_file_prwv was released too early, before data
5319 actually was hitting the wire. Extended lock-holding time until
5320 server acknowledges receiving data.
5323 Frequency : extremely rare. Never observed in practice.
5325 Description: avoid generating lustre_handle cookie of 0.
5326 Details : class_handle_hash() generates handle cookies by incrementing
5327 global counter, and can hit 0 occasionaly (this is unlikely, but
5328 not impossible, because initial value of cookie counter is
5329 selected randonly). Value of 0 is used as a sentinel meaning
5330 "unassigned handle" --- avoid it. Also coalesce two critical
5331 sections in this function into one.
5333 Severity : enhancement
5335 Description: allow liblustre clients to delegate truncate locking to OST
5336 Details : To avoid overhead of locking, liblustre client instructs OST to
5337 take extent lock in ost_punch() on client's behalf. New connection
5338 flag is added to handle backward compatibility.
5340 Severity : enhancement
5341 Bugzilla : 4928, 7341, 9758
5342 Description: allow number of OST service threads to be specified
5343 Details : a module parameter allows the number of OST service threads
5344 to be specified via "options ost ost_num_threads={N}" in the
5345 OSS's /etc/modules.conf or /etc/modprobe.conf.
5349 Bugzilla : 6146, 9635, 9895
5350 Description: servers crash with bad pointer in target_handle_connect()
5351 Details : In rare cases when a client is reconnecting it was possible that
5352 the connection request was the last reference for that export.
5353 We would temporarily drop the export reference and get a new
5354 one, but this may have been the last reference and the export
5355 was just destroyed. Get new reference before dropping old one.
5357 Severity : enhancement
5358 Frequency : if client is started with failover MDS
5360 Description: Allow multiple MDS hostnames in the mount command
5361 Details : Try to read the configuration from all specified MDS
5362 hostnames during a client mount in case the "primary"
5365 Severity : enhancement
5367 Description: Stop sending data to evicted clients as soon as possible.
5368 Details : Check if the client we are about to send or are sending data to
5369 was evicted already. (Check is done every second of waiting,
5370 for which l_wait_event interface was extended to allow checking
5371 of exit condition at specified intervals).
5374 Frequency : rare, normally only when NFS exporting is done from client
5376 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5378 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5381 Frequency : always, if srand() is called before liblustre initialization
5383 Description: Liblustre uses system PRNG disturbing its usage by user application
5384 Details : Introduce internal to lustre fast and high-quality PRNG for
5385 lustre usage and make liblustre and some other places in generic
5386 lustre code to use it.
5388 Severity : enhancement
5389 Bugzilla : 9477, 9557, 9870
5390 Description: Verify that the MDS configuration logs are updated when xml is
5391 Details : Check if the .xml configuration logs are newer than the config
5392 logs stored on the MDS and report an error if this is the case.
5393 Request --write-conf, or allow starting with --old_conf.
5395 Severity : enhancement
5397 Description: Handle symlinks in the path when checking if Lustre is mounted.
5398 Details : Resolve intermediate symlinks when checking if a client has
5399 mounted a filesystem to avoid duplicate client mounts.
5404 Description: lconf can hit an error exception but still return success.
5405 Details : The lconf command catches the Command error exception at the top
5406 level script context and will exit with the associated exit
5407 status, but doesn't ensure that this exit status is non-zero.
5412 Description: failure of ptlrpc thread startup can cause oops
5413 Details : Starting a ptlrpc service thread can fail if there are a large
5414 number of threads or the server memory is very fragmented.
5415 Handle this without oopsing.
5418 Frequency : always, only if liblustre and non-default acceptor port was used
5420 Description: liblustre cannot connect to servers with non-default acceptor port
5421 Details : tcpnal_set_default_params() was not called and was therefore
5422 ignoring the environment varaible TCPNAL_PORT, as well as other
5423 TCPNAL_ environment variables
5428 Description: two objects could be created on the same OST for a single file
5429 Details : If an OST is down, in some cases it was possible to create two
5430 objects on a single OST for a single file. No problems other
5431 than potential performance impact and spurious error messages.
5435 Bugzilla : 5681, 9562
5436 Description: Client may oops in ll_unhash_aliases
5437 Details : Client dcache may become inconsistent in race condition.
5438 In some cases "getcwd" can fail if the current directory is
5444 Description: Inode refcounting problems in NFS export code
5445 Details : link_raw functions used to call d_instantiate without obtaining
5446 extra inode reference first.
5450 Bugzilla : 9942, 9903
5451 Description: Referencing freed requests leading to crash, memleaks with NFS.
5452 Details : We used to require that call to ll_revalidate_it was always
5453 followed by ll_lookup_it. Also with revalidate_special() it is
5454 possible to call ll_revalidate_it() twice for the same dentry
5455 even if first occurence returned success. This fix changes semantic
5456 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5457 reference on a request referred from the intent.
5458 ll_intent_release() then releases such a request.
5461 Frequency : rare, normally benchmark loads only
5463 Description: unlinked inodes were kept in memory on the client
5464 Details : If a client is repeatedly creating and unlinking files it
5465 can accumulate a lot of stale inodes in the inode slab cache.
5466 If there is no other client load running this can cause the
5467 client node to run out of memory. Instead flush old inodes
5468 from client cache that have the same inode number as a new inode.
5471 Frequency : SLES9 2.6.5 kernel and long filenames only
5472 Bugzilla : 9969, 10379
5473 Description: utime reports stale NFS file handle
5474 Details : SLES9 uses out-of-dentry names in some cases, which confused
5475 the lustre dentry revalidation. Change it to always use the
5479 Frequency : rare, unless heavy write-truncate concurrency is continuous
5480 Bugzilla : 4180, 6984, 7171, 9963, 9331
5481 Description: OST becomes very slow and/or deadlocked during object unlink
5482 Details : filter_destroy() was holding onto the parent directory lock
5483 while truncating+unlinking objects. For very large objects this
5484 may block other threads for a long time and slow overall OST
5485 responsiveness. It may also be possible to get a lock ordering
5486 deadlock in this case, or run out of journal credits because of
5487 the combined truncate+unlink. Solution is to do object truncate
5488 first in one transaction without parent lock, and then do the
5489 final unlink in a new transaction with the parent lock. This
5490 reduces the lock hold time dramatically.
5493 Frequency : rare, 2.4 kernels only
5495 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5496 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5497 before unmounting their filesystem in order to avoid deadlock.
5498 The kernel_locked() function in 2.4 kernels only checks whether
5499 the kernel lock is held, not whether it is this process that is
5500 holding it as 2.6 kernels do.
5505 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5506 Details : The client ptlrpc code may be trying to reconnect to a down
5507 server before a previous connection attempt has timed out.
5508 Increase the reconnect interval to be longer than the connection
5509 timeout interval to avoid sending duplicate connections to
5513 Frequency : echo_client brw_test command
5515 Description: fix echo_client to work with OST preallocated code
5516 Details : OST preallocation code (5137) didn't take echo_client IO path
5517 into account: echo_client calls filter methods outside of any
5518 OST thread and, hence, there is no per-thread preallocated
5519 pages and buffers to use. Solution: hijack pga pages for IO. As
5520 a byproduct, this avoids unnecessary data copying.
5524 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5525 Description: Client can oops in mdc_commit_close() after open replay
5526 Details : It was possible for the MDS to return an open request with no
5527 transaction number in mds_finish_transno() if the client was
5528 evicted, but without actually returning an error. Clients
5529 would later try to replay that open and may trip an assertion
5530 Simplify the client close codepath, and always return an error
5531 from the MDS in case the open is not successful.
5534 Frequency : rare, 2.6 OSTs only
5536 Description: OST may deadlock under high load on fragmented files
5537 Details : If there was a heavy load and highly-fragmented OST filesystems
5538 it was possible to have all the OST threads deadlock waiting on
5539 allocation of biovecs, because the biovecs were not released
5540 until the entire RPC IO was completed. Instead, release biovecs
5541 as soon as they are complete to ensure forward IO progress.
5543 Severity : enhancement
5545 Description: Support for specifying external journal device at mount
5546 Details : If an OST or MDS device is formatted with an external journal
5547 device, this device major/minor is stored in the ext3 superblock
5548 and may not be valid for failover. Allow detecting and
5549 specifying the external journal at mount time.
5554 Description: Mounting an MDS with pending unlinked files may cause oops
5555 Details : target_finish_recovery() calls mds_postrecov() which returned
5556 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5557 considers this an error and immediately begins cleaning up the
5558 lov, just after starting the mds_lov process
5560 Severity : enhancement
5562 Description: Implement 'lfs df' to report actual free space on per-OST basis
5563 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5564 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5565 usage of objects; '-h' to report in human readable format.
5567 ------------------------------------------------------------------------------
5569 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5576 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5577 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5578 at startup during orphan unlinking. Instead, skip buddy update
5579 before setup, as it will be regenerated after recovery is complete.
5584 Description: If an OST is inactive, its locks might reference stale inodes.
5585 Details : lov_change_cbdata() must iterate over all namespaces, even if
5586 they are inactive to clear inode references from the lock.
5588 Severity : enhancement
5589 Frequency : occasional, if non-standard max_dirty_mb used
5591 Description: Client will block write RPCs if not enough grant
5592 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5593 then the client will block writes while waiting for another RPC
5594 to complete instead of consuming its dirty limit. With change
5595 we get improved performance when max_dirty_mb is small.
5597 Severity : enhancement
5598 Bugzilla : 3389, 6253
5599 Description: Add support for supplementary groups on the MDS.
5600 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5601 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5602 lookups for user supplementary groups into a cache.
5606 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5607 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5608 for liblustre O_EXCL is enforced by MDS.
5612 Description: readdir never returns NULL in liblustre.
5613 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5614 dentry in struct dirent.
5618 Frequency : liblustre only, depends on application IO pattern
5619 Description: liblustre clients evicted if not contacting servers
5620 Details : Don't put liblustre clients into the ping_evictor list, so
5621 they will not be evicted by the pinger ever.
5623 Severity : enhancement
5625 Description: Add ability to evict clients by NID from MDS.
5626 Details : By echoing "nid:$NID" string into
5627 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5628 $NID would be instantly evicted from this MDS and from all active
5629 OSTs connected to it.
5633 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5634 Details : lookup_it_finish() used to query file size from OSTs that was not
5639 Description: service threads change working directory to that of init
5640 Details : Starting lustre service threads may pin the working directory
5641 of the parent thread, making that filesystem busy. Threads
5642 now change to the working directory of init to avoid this.
5646 Frequency : during shutdown only
5647 Description: shutdown with a failed MDS or OST can cause unmount to hang
5648 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5653 Frequency : 2.6 only
5654 Description: chmod/chown may include an extra supplementary group
5655 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5656 supplementary group and if none is specified this is used.
5659 Bugzilla : 5479 (6816)
5661 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5662 Details : If lookup is in progress on a file that is unlinked we might try
5663 to revalidate the inode and fail in revalidate after lookup is
5664 complete and ll_file_open() enqueues the open again but
5665 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5668 Frequency : always, if lconf --abort_recovery used
5670 Description: lconf --abort_recovery fails with 'Operation not supported'
5671 Details : lconf was attempting to abort recovery on the MDT device and not
5674 ------------------------------------------------------------------------------
5676 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5681 Frequency : rare (only unsupported configurations with a node running as an
5683 Bugzilla : 6514, 5137
5684 Description: Mounting a Lustre file system on a node running as an OST could
5686 Details : OSTs now preallocates memory needed to write out data at
5687 startup, instead of when needed, to avoid having to
5688 allocate memory in possibly low memory situations.
5689 Specifically, if the file system is mounted on on OST,
5690 memory pressure could force it to try to write out data,
5691 which it needed to allocate memory to do. Due to the low
5692 memory, it would be unable to do so and the node would
5693 become unresponsive.
5695 Severity : enhancement
5697 Description: Addition of lconf --service command line option
5698 Details : lconf now accepts a '--service <arg>' option, which is
5699 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5701 Severity : enhancement
5703 Description: Failover mode is now the default for OSTs.
5704 Details : By default, OSTs will now run in failover mode. To return to
5705 the old behaviour, add '--failout' to the lmc line for OSTs.
5707 Severity : enhancement
5709 Description: Health checks are now provided for MDS and OSTs
5710 Details : Additional detailed health check information on MSD and OSTs
5711 is now provided through the procfs health_check value.
5714 Frequency : occasional, depends on IO load
5716 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5717 after numerous create/delete cycles
5718 Details : The ext3 inode allocation policy would not allocate new inodes
5719 very well on the OSTs because there are no new directories
5720 being created. Instead we look for groups with free space if
5721 the parent directories are nearly full.
5726 Description: Network or server problems during mount may cause partially
5727 mounted clients instead of returning an error.
5728 Details : The config llog parsing code may overwrite the error return
5729 code during mount error handling, returning success instead
5735 Description: MDS can fail to allocate large reply buffers
5736 Details : After long uptimes the MDS can fail to allocate large reply
5737 buffers (e.g. zconf client mount config records) due to memory
5738 fragmentation or consumption by the buffer cache. Preallocate
5739 some large reply buffers so that these replies can be sent even
5740 under memory pressure.
5744 Frequency : rare (liblustre)
5745 Description: fsx running with liblustre complained that using truncate() to
5746 extend the file doesn't work. This patch corrects that issue.
5747 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5748 ATTR_SIZE and lsm use in llu_setattr_raw.
5752 Frequency : rare, only 2.6 kernels
5753 Description: Unusual file access patterns on the MDS may result in inode
5754 data being lost in very rare circumstances.
5755 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5756 code in the 2.6 kernel caused the inode and/or EA data not to
5757 be read from disk, causing single-file corruption.
5761 Frequency : rare, only 2.6 filesystems using extents
5762 Description: Heavy concurrent write and delete load may cause data corruption.
5763 Details : It was possible under high-load situations to have an extent
5764 metadata block in the block device cache from a just-unlinked
5765 file overwrite a newly-allocated data block. We now unmap any
5766 metadata buffers that alias just-allocated data blocks.
5770 Frequency : filesystems with default stripe_count larger than 77
5771 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5772 Details : lconf specifies an inode size of 4096 bytes when the default
5773 stripe_count is larger than 77. This conflicts with the default
5774 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5775 this case to avoid pinning too much memory for large EAs.
5777 ------------------------------------------------------------------------------
5779 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5784 Frequency : rare (extremely heavy IO load with hundreds of clients)
5786 Description: Client is evicted, gets IO error writing to file
5787 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5788 caused clients to be evicted for AST timeouts. The fixes in
5789 bug 5192 mean we no longer need to have such short AST timeouts
5790 so ldlm_timeout has been increased.
5793 Frequency : occasional during --force or --failover shutdown under load
5794 Bugzilla : 5949, 4834
5795 Description: Server oops/LBUG if stopped with --force or --failover under load
5796 Details : a collection of import/export refcount and cleanup ordering
5797 issues fixed for safer force cleanup
5800 Frequency : only filesystems larger than 120 OSTs
5801 Bugzilla : 5990, 6223
5802 Description: lfs getstripe would oops on a very large filesystem
5803 Details : lov_getconfig used kfree on vmalloc'd memory
5806 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5807 Bugzilla : 6242, 6243
5808 Description: reading from files that had been truncated to a non-zero size
5809 but never opened returned no data
5810 Details : ll_file_read() reads zeros from no-object files to EOF
5815 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5816 Details : There's little harm in aborting MDS/OSS recovery and letting it
5817 try again, so I removed the LASSERT and return an error instead.
5819 Severity : enhancement
5821 Description: New debugging infrastructure for tracking down data corruption
5822 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5823 (b) cover more of the client-side code path, and (c) try to narrow
5824 down where problems occurred
5828 Bugzilla : 3819, 4364, 4397, 6313
5829 Description: Racing close and eviction MDS could cause assertion in mds_close
5830 Details : It was possible to get multiple mfd references during close and
5831 client eviction, leading to one thread referencing a freed mfd.
5833 Severity: : enhancement
5834 Bugzilla : 3262, 6359
5835 Description: Attempts to reconnect to servers are now more aggressive.
5836 Details : This builds on the enhanced upcall-less recovery that was added
5837 in 1.4.2. When trying to reconnect to servers, clients will
5838 now try each server in the failover group every 10 seconds. By
5839 default, clients would previously try one server every 25 seconds.
5844 Description: After recovery, certain operations trigger a failed
5845 assertion on a client.
5846 Details : Failing over an mds, using lconf -d --failover, while a
5847 client was doing a readdir() call would cause the client to
5848 LBUG after recovery completed and the readdir() was resent.
5850 Severity : enhancement
5852 Description: Default groups are now added by lconf
5853 Details : You can now run lconf --group <servicename> without having to
5854 manually add groups with lmc.
5857 Frequency : occasional
5859 Description: Nodes with an elan id of 0 trigger a failed assertion
5862 Frequency : always when accessing e.g. tty/console device nodes
5864 Description: tty and some other devices nodes cannot be used on lustre
5865 Details : file's private_data field is used by device data and lustre
5866 values in there got lost. New field was added to struct file to
5867 store fs-specific private data.
5870 Frequency : when exporting Lustre via NFS
5872 Description: NFSD failed occasionally when looking up a path component
5873 Details : NFSD is looking up ".." which was broken in ext3 directories
5874 that had grown large enough to become hashed.
5877 Frequency : Clusters with multiple interfaces not on the same subnet
5879 Description: Nodes will repeatedly try to reconnect to an interface which it
5880 cannot reach and report an error to the log.
5881 Details : Extra peer list entries will be created by lconf with some peers
5882 unreachable. lconf now validates the peer before adding it.
5885 Frequency : Only if a default stripe is set on the filesystem root.
5887 Description: Setting a default stripe on the filesystem root prevented the
5888 filesystem from being remounted.
5889 Details : The client was sending extra request flags in the root getattr
5890 request and did not allocate a reply buffer for the dir EA.
5893 Frequency : occasional, higher if lots of files are accessed by one client
5894 Bugzilla : 6159, 6097
5895 Description: Client trips assertion regarding lsm mismatch/magic
5896 Details : While revalidating inodes the VFS looks up inodes with ifind()
5897 and in rare cases can find an inode that is being freed.
5898 The ll_test_inode() code will free the lsm during ifind()
5899 when it finds an existing inode and then the VFS later attaches
5900 this free lsm to a new inode.
5904 Bugzilla : 6422, 7030
5905 Description: MDS deadlock between mkdir and client eviction
5906 Details : Creating a new file via mkdir or mknod (starting a transaction
5907 and getting the ns lock) can deadlock with client eviction
5908 (gets ns lock and trying to finish a synchronous transaction).
5911 Frequency : occasional
5912 Description: While starting a server, the fsfilt_ext3 module could not be
5914 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5915 kernels. Previously, lconf would still use the ext3 name
5916 when trying to load modules. Now, it will correctly use
5917 ext3 on 2.4 and ldiskfs on 2.6.
5919 Severity : enhancement
5920 Description: The default stripe count has been changed to 1
5921 Details : The interpretation of the default stripe count (0, to lfs
5922 or lmc) has been changed to mean striping across a single
5923 OST, rather than all available. For general usage we have
5924 found a stripe count of 1 or 2 works best.
5926 Severity : enhancement
5927 Description: Add support for compiling against Cray portals.
5928 Details : Conditional compiling for some areas that are different
5932 Frequency : occasional
5933 Bugzilla : 6409, 6834
5934 Description: Creating files with an explicit stripe count may lead to
5935 a failed assertion on the MDS
5936 Details : If some OSTs are full or unavailable, creating files may
5937 trigger a failed assertion on the MDS. Now, Lustre will
5938 try to use other servers or return an error to the
5942 Frequency : occasional
5944 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5945 caused lock timeout during readahead (since 1.4.2).
5946 Details : Processes doing readahead might match a lock that hasn't been
5947 granted yet if there are overlapping and conflicting lock
5948 requests. The readahead process waits on ungranted lock
5949 (original lock is CBPENDING), while OST waits for that process
5950 to cancel CBPENDING read lock and eventually evicts client.
5952 Severity : enhancement
5954 Description: Initial enabling of flock support for clients
5955 Details : Implements fcntl advisory locking and file status functions.
5956 This feature is provided as an optional mount flag (default
5957 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5958 locking are implemented yet, and those that are are not guaranteed
5959 to be completely correct in production environments.
5960 mount -t lustre -o [flock|noflock] ...
5963 Frequency : occasional
5965 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5966 assertion in the ext3 JBD (journaling) layer.
5967 Details : The b_committed_data struct is protected by the big kernel lock
5968 in 2.4 kernels, serializing journal_commit_transaction() and
5969 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5970 access to this struct. In 2.6 kernels there is finer grained
5971 locking to improve SMP performance of the JBD layer.
5975 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5980 Description: merge and cleanup kernel patches.
5983 -----------------------------------------------------------------------------
5985 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5987 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5988 versions of Lustre. Please update all servers and clients to
5989 version 1.4.2 or later at the same time. You must also run
5990 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5991 to update the configuration logs.
5993 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5994 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5995 - swab llog records of type '0' so we get proper header size/idx (5861)
5996 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5997 - fix rename of one directory over another leaking an inode (5953)
5998 - avoid SetPageDirty on 2.6 (5981)
5999 - don't re-add just-being-destroyed locks to the waiting list (5653)
6000 - when creating new directories, inherit the parent's custom
6001 striping settings if present parent (3048)
6002 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6003 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6004 - don't hold client locks on temporary worklist from l_lru (5666)
6005 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6006 - automatically evict dead clients (5921)
6007 - Update file size properly in create+truncate+fstat case (6196)
6008 - Do not unhash mountpoint dentries, do not allow removal of
6010 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6011 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6012 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6013 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6014 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6015 - Correctly handle reads of files with no objects (6243)
6016 - lctl recover will also mark a device active if deactivate used (5933)
6018 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6019 - allow --write-conf on an MDS with different nettype than client (5619)
6020 - don't write config llogs to MDS for mounts not from that MDS (5617)
6021 - lconf should create multiple TCP connections from a client (5201)
6022 - init scripts are now turned off by default; run chkconfig --on
6023 lustre and chkconfig --on lustrefs to use them
6024 - upcalls are no longer needed for clients to recover to failover
6026 - add --abort-recovery option to lconf to abort recovery on device
6028 - add support for an arbitrary number of OSTs (3026)
6029 - Quota support protocol changes.
6030 - forward compatibility changes to wire structs (6007)
6031 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6032 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6033 - improved "lustre status" script
6034 - initialize blocksize for non-regular files (6062)
6035 - added --disable-server and --disable-client configure options (5782)
6036 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6037 - Vanilla 2.4.29 support
6038 - increase maximum number of obd devices to 520 (6242)
6039 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6040 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6041 SLES 9 2.6.5 kernels
6042 - limit stripes per file to 160 (the maximum EA size) (6093)
6044 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6047 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6048 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6049 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6050 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6051 - small ext3 extents cleanups and fixes (5733)
6052 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6053 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6054 - avoid needless client->OST connect, fix handle mismatch (5317)
6055 - fix DLM error path that led to out-of-sync client, long delays (5779)
6056 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6057 - fix several locking issues related to i_size (5492,5624,5654,5672)
6058 - don't move pending lock onto export if it is already evicted (5683)
6059 - fix kernel oops when creating .foo in unlinked directory (5548)
6060 - fix deadlock in obdfilter statistics vs. object create (5811)
6061 - use time_{before,after} to avoid timer jiffies wrap (5882)
6062 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6063 - Do not leak request if server was not able to process it (5154)
6064 - If mds_open unable to find parent dir, make that negative lookup(5154)
6065 - don't create new directories with extent-mapping (5909, 5936)
6067 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6068 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6069 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6070 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6071 - support for RHEL4 kernel on i686 (5773)
6072 - provide error messages when incompatible logs are encountered (5898)
6074 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6075 * version 1.4.0.10 (1.4.1 release candidate 1)
6077 - don't keep a lock reference when lock is not granted (4238)
6078 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6079 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6080 - drop import inflight refcount on signal_completed_replay error (5255)
6081 - unlock page after async write error during send (3677)
6082 - handle missing objects in filter_preprw_read properly (5265)
6083 - no transno return for symlink open, don't save no-trasno open (3440)
6084 - don't try to complete elan receive that already failed (4012)
6085 - free RPC server reply state on error (5406)
6086 - clean up thread from ptlrpc_start_thread() on error (5160)
6087 - readahead could read extra page into cache that wasn't ejected (5388)
6088 - prevent races in class_attach/setup/cleanup/detach (5260)
6089 - don't dereference de->d_inode after l_dput of de (5458)
6090 - use "int" for stripe value returned from lock_to_stripe (5544)
6091 - mballoc allocation and error-checking fixes in 2.6 (5504)
6092 - block device patches to fix I/O request sizes in 2.6 (5482)
6093 - look up hostnames for IB nals (5602)
6094 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6095 - don't start multiple acceptors for the same port (5277)
6096 - fix incorrect LASSERT in mds_getattr_name (5635)
6097 - export a proc file for general "ping" checking (5628)
6098 - fix "lfs check" to not block when the MDS is down (5628)
6100 - service request history (4965)
6101 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6102 - create an "evict_client" /proc entry on OSTs, like the MDS has
6103 - fix mount usage message, return errors per mount(8) (5168)
6104 - change grep [] to grep "[]" in tests so they work in more UMLs
6105 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6106 - remove ancient LOV_MAGIC_V0 EA support (5047)
6107 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6108 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6109 - allow readahead allocations to fail when low on memory (5383)
6110 - mmap locking landed again, after considerable improvement (2828)
6111 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6113 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6116 - send OST transaction number in read/write reply to free req (4966)
6117 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6118 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6119 - the watchdog thread now runs as interruptible (5246)
6120 - flock/lockf fixes (but it's still disabled, pending 5135)
6121 - don't use EXT3 constants in llite code (5094)
6122 - memory shortage at startup could cause assertion (5176)
6124 - reorganization of lov code
6125 - single portals codebase
6127 - add extents/mballoc support (5025)
6128 - direct I/O reads in the obdfilter (4048)
6129 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6131 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6134 - send OST transaction number in read/write reply to free req (4966)
6135 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6136 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6137 - the watchdog thread now runs as interruptible (5246)
6138 - handle missing objects in filter_preprw_read properly (5265)
6139 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6140 - drop import inflight refcount on signal_completed_replay error (5255)
6141 - unlock page after async write error during send (3677)
6142 - return original error code on reconstructed replies (3761)
6143 - no transno return for symlink open, don't save no-trasno open (3440)
6145 - add pid to ldlm debugging output (4922)
6146 - bump the watchdog timeouts -- we can't handle 30sec yet
6147 - extra debugging for orphan dentry/inode bug (5259)
6149 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6152 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6153 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6154 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6155 - fix NULL dereference in /proc/sys/portals/routes (4827)
6156 - allow failed mdc_close() operations to be interrupted (4561)
6157 - stop precreate on OST before MDS would time out on it (4778)
6158 - don't send partial-page writes before EOF from client (4410)
6159 - discard client grant for sub-page writes on large-page clients (4520)
6160 - don't free dentries not owned by NFS code, check generation (4806)
6161 - fix lsm leak if mds_create_objects() fails (4801)
6162 - limit debug_daemon file size, always print CERROR messages (4789)
6163 - use transno after validating reply (3892)
6164 - process timed out requests if import state changes (3754)
6165 - update mtime on OST during writes, return in glimpse (4829)
6166 - add mkfsoptions to LDAP (4679)
6167 - use ->max_readahead method instead of zapping global ra (5039)
6168 - don't interrupt __l_wait_event() during strace
6170 - add software watchdogs to catch hung threads quickly (4941)
6171 - make lustrefs init script start after nfs is mounted
6172 - fix CWARN/ERROR duplication (4930)
6173 - return async write errors to application if possible (2248)
6174 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6175 - print NAL number in %x format (4645)
6176 - update barely-supported suse-2.4.21-171 series (4842)
6177 - support for sles 9 %post scripts
6178 - support for building 2.6 kernel-source packages
6179 - support for sles km_* packages
6181 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6184 - ignore -ENOENT errors in osc_destroy (3639)
6185 - notify osc create thread that OSC is being cleaned up (4600)
6186 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6187 - reconstruct ost_handle() like mds_handle() (4657)
6188 - create a new thread to do import eviction to avoid deadlock (3969)
6189 - let lconf resolve symlinked-to devices (4629)
6190 - don't unlink "objects" from directory with default EA (4554)
6191 - hold socknal file ref over connect in case target is down (4394)
6192 - allow more than 32000 subdirectories in a single directory (3244)
6193 - fix blocks count for O_DIRECT writes (3751)
6194 - OST returns ENOSPC from object create when no space left (4539)
6195 - don't send truncate RPC if file size isn't changing (4410)
6196 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6197 - bind to privileged port in socknal and tcpnal (3689)
6199 - rate limit CERROR/CWARN console message to avoid overload (4519)
6200 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6201 - basic mmap support (3918)
6202 - kernel patch series update from b1_4 (4711)
6204 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6207 - avoid crash during MDS cleanup with OST shut down (2775)
6208 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6209 - don't use bad inodes on the MDS (3744)
6210 - dynamic object preallocation to improve recovery speed (4236)
6211 - don't hold spinlock over lock dumping or change debug flags (4401)
6212 - don't zero obd_dev when it is force cleaned (3651)
6213 - print grants to console if they go negative (4431)
6214 - "lctl deactivate" will stop automatic recovery attempts (3406)
6215 - look for existing locks in ldlm_handle_enqueue() (3764)
6216 - don't resolve lock handle twice in recovery avoiding race (4401)
6217 - revalidate should check working dir is a directory (4134)
6219 - don't always mark "slow" obdfilter messages as errors (4418)
6221 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6224 - don't close LustreDB during write_conf until it is done (3860)
6225 - fix typo in lconf for_each_profile (3821)
6226 - allow dumping logs from multiple threads at one time (3820)
6227 - don't allow multiple threads in OSC recovery (3812)
6228 - fix debug_size parameters (3864)
6229 - fix mds_postrecov to initialize import for llog ctxt (3121)
6230 - replace config semaphore with spinlock (3306)
6231 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6232 - don't allow enqueue to complete on a destroyed export (3822)
6233 - down write_lock before checking llog header bitmap (3825)
6234 - recover from lock replay timeout (3764)
6235 - up llog sem before sending rpc (3652)
6236 - reduce ns lock hold times when setting kms (3267)
6237 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6238 - fix NULL deref and obd_dev leak on setup error (3312)
6239 - replace some LBUG about llog ops with error handling (3841)
6240 - don't match INVALID dentries from d_lookup and spin (3784)
6241 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6242 - fix invalid assertion in ptlrpc_set_wait (3880)
6244 - add libwrap support for the TCP acceptor (3996)
6245 - add /proc/sys/portals/routes for non-root route listing (3994)
6246 - allow setting MDS UUID in .xml (2580)
6247 - print the stack of a process that LBUGs (4228)
6249 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6252 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6253 - make sure to unset replay flag from failed open requests (3440)
6254 - if default stripe count is 0, use OST count for inode size (3636)
6255 - update parent mtime/ctime on client for create/unlink (2611)
6256 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6257 - free recovery state on server during a forced cleanup (3571)
6258 - unregister_reply for resent reqs (3063)
6259 - loop back devices mounting and status check on 2.6 (3563)
6260 - fix resource-creation race that can provoke i_size == 0 (3513)
6261 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6262 - more debugging for page-accounting assertion (3746)
6263 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6264 - avoid deadlock after precreation failure (3758)
6265 - fix race and lock order deadlock in orphan handling (3450, 3750)
6266 - add validity checks when grabbing inodes from l_ast_data (3599)
6268 - add /proc/.../recovery_status to obdfilter (3428)
6269 - lightweight CDEBUG infrastructure, debug daemon (3668)
6270 - change default OSC RPC parameters to be better on small clusters
6271 - turn off OST read cache for files smaller than 32MB
6272 - install man pages and include them in rpms (3100)
6273 - add new init script for (un)mounting lustre filesystems (2593)
6274 - run chkconfig in %post for init scripts (3701)
6275 - drop scimac NAL (unmaintained)
6277 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6280 - clean kiobufs before and after use (3485)
6281 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6282 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6283 - change init script to fail more clearly if not run as root (1528)
6284 - allow clients to reconnect during replay (1742)
6285 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6286 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6287 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6288 - fix cond_resched (3554)
6289 - only evict unfinished clients after recovery (3515)
6290 - allow bulk resend, prevent data loss (3570)
6291 - dynamic ptlrpc request buffer allocation (2102)
6292 - don't allow unlinking open directory if it isn't empty (2904)
6293 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6294 - remove extraneous obd dereference causing LASSERT failure (3334)
6295 - don't use get_cycles() when creating temp. files on the mds (3156)
6296 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6297 - handle EEXIST for set-stripe, set proper directory name (3336)
6299 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6300 - fix ksocknal_fmb_callback() error messages (2918)
6302 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6305 - don't copy lvb into (possibly NULL) reply on error (2983)
6306 - don't deref dentry after dput, don't free lvb on error (2922)
6307 - use the kms to determine writeback rpc length (2947)
6308 - increment oti_logcookies when osc is inactive (2948)
6309 - update client's i_blocks count via lvb messages (2543)
6310 - handle intent open/close of special files properly (1557)
6311 - mount MDS with errors=remount-ro, like obdfilter (2009)
6312 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6313 - don't use cancelling-locks' kms values (2947)
6314 - use highest lock extent for kms, not last one (2925)
6315 - don't dereference ERR_PTR() dentry in error handling path (3107)
6316 - fix thread race in portals_debug_dumplog() (3122)
6317 - create lprocfs device entries at setup instead of at attach (1519)
6318 - common AST error handler, don't evict client on completion race (3145)
6319 - zero nameidata in detach_mnt in 2.6 (3118)
6320 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6321 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6322 - initialize RPC timeout timer earlier for 2.6 (3219)
6323 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6324 - print nal/nid for unknown nid (3258)
6325 - additional checks for oscc recovery before doing precreate (3284)
6326 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6327 - don't crash in mdc_close for bad permissions on open (3285)
6328 - zero i_rdev for non-device files (3147)
6329 - clear page->private before handing to FS, better assertion (3119)
6330 - tune the read pipeline (3236)
6331 - fix incorrect decref of invalidated dentry (2350)
6332 - provide read-ahead stats and refine rpc in flight stats (3328)
6333 - don't hold journal transaction open across create RPC (3313)
6334 - update atime on MDS at close time (3265)
6335 - close LDAP connection when recovering to avoid server load (3315)
6336 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6337 - don't leak open file on MDS after open resend (3325)
6338 - serialize filter_precreate and filter_destroy_precreated (3329)
6339 - loop device shouldn't call sync_dev() for nul device (3092)
6340 - clear page cache after eviction (2766)
6341 - resynchronize MDS->OST in background (2824)
6342 - refuse to mount the same filesystem twice on same mountpoint (3394)
6343 - allow llmount to create routes for mounting behind routers (3320)
6344 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6345 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6346 - fix rare problem with rename on htree directories (3417)
6348 - allow default OST striping configuration per directory (1414)
6349 - fix compilation for qswnal for 2.6 kernels (3125)
6350 - increase maximum number of MDS request buffers for large systems
6351 - change liblustreapi to be useful for external progs like lfsck (3098)
6352 - increase local configuration timeout for slow disks (3353)
6353 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6355 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6358 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6359 - don't overwrite extent policy data in reply if lock was blocked (2901)
6360 - drop filter export grants atomically with removal from device (2663)
6361 - del obd_self_export from work_list in class_disconnect_exports (2908)
6362 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6363 - swab reply message in mdc_close, other PPC fixes (2464)
6364 - fix destroying of named logs (2325)
6365 - overwrite old logs when running lconf --write_conf (2264)
6366 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6367 - fix race in target_handle_connect (2898)
6368 - mds_reint_create() should take same inode create lock (2926)
6369 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6370 - don't close files for self_export to avoid uninitialized obd (2936)
6371 - allow MDS with the same name as client node (2939)
6372 - hold dentry reference for closed log files for unlink (2325)
6373 - reserve space for all logs during transactions (2059)
6374 - don't evict page beyond end of stripe extent (2925)
6375 - don't oops on a deleted current working directory (2399)
6376 - handle hard links to targets without a parent properly (2517)
6377 - don't dereference NULL lock when racing during eviction (2867)
6378 - don't grow lock extents when lots of conflicting locks (2919)
6380 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6383 - account for cache space usage on clients to avoid data loss (974)
6384 - lfsck support in lustre kernel code (2349)
6385 - reduce journal credits needed for BRW writes (2370)
6386 - orphan handling to avoid losing space on client/server crashes
6387 - ptlrpcd can be blocked, stopping ALL progress (2477)
6388 - use lock value blocks to assist in proper KMS, faster stat (1021)
6389 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6390 - recovery for initial connections (2355)
6391 - fixes for mds_cleanup_orphans (1934)
6392 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6393 - block all file creations until orphan recovery completes (1901)
6394 - client remove rq_connection from request struct (2423)
6395 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6396 - recovery timer race (2670)
6397 - mdc_close recovey bug (2532)
6398 - ptlrpc cleanup bug (2710)
6399 - mds timeout on local locks (2588)
6400 - namespace lock held during RPCs (2431)
6401 - handle interrupted sync write properly (2503)
6402 - don't try to handle a message that hasn't been replied to (2699)
6403 - client assert failure during cleanup after abort recovery (2701)
6404 - leak mdc device after failed mount (2712)
6405 - ptlrpc_check_set allows timedout requests to complete (2714)
6406 - wait for inflight reqs when ptlrpcd finishes (2710)
6407 - make sure unregistered services are removed from the srv_list
6408 - reset bulk XID's when resending them (caught by 1138 test)
6409 - unregister_bulk after timeout
6410 - fix lconf error (2694)
6411 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6412 - readahead locks pages, leaves pending causing memory pressure (2673)
6413 - increase OST request buffers to 4096 on large machines (2729)
6414 - fix up permission of existing directories in simple_mkdir (2661)
6415 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6416 - don't assign transno to errored transactions (2742)
6417 - don't delete objects on OST if given a bogus objid from MDS (2751)
6418 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6419 - if rq_no_resend, then timeout request after recovery (2432)
6420 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6421 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6422 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6423 - MDS recovery shouldn't delete objects that it has given out (2730)
6424 - if enqueue arrives after completion, don't clobber LVB (2819)
6425 - don't unlock pages twice when trigger_group_io returns error (2814)
6426 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6427 - don't write pages to disk if there was an error (1450)
6428 - don't ping imports that have recovery disabled (2676)
6429 - take buffered bytes into account when balancing socknal conn (2817)
6430 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6431 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6432 - remove little-endian swabbing of llog records (1987)
6433 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6434 - retry reposting request buffers if they fail (1191)
6435 - grow extent at grant time to avoid granting a revoked lock (2809)
6436 - lock revoke doesn't evict page if covered by a second lock (2765)
6437 - disable VM readahead to avoid reading outside lock extents (2805)
6439 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6440 - updated kernel patches for hp-2.4.20 kernel (2681)
6442 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6445 - fix truncated write corruption (2366)
6446 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6448 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6449 - ensure len(uuid) < 37 in lmc (1171)
6450 - fix ia64 OOPS in llog_test (2255)
6451 - zero end of page at obdfilter for partial page writes (2648)
6452 - don't leave stale dentries around after renames (bug 2428)
6453 - fix timeouts when evicting a client with a single lock held (2642)
6454 - set deadline for the initial HELLO message to drain (2634)
6455 - print out dotted-quad IP addresses in the socknal (2302)
6457 - additional debugging for MDS client eviction problem (2443)
6458 - fix mkfsoptions support for osts (2603, 2604)
6460 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6463 - add series for the vanilla 2.6.0 kernel
6464 - add series for the vanilla 2.4.24 kernel
6465 - add series for a cray x86/64 UL kernel drop
6466 - fix xattr patches for the vanilla 2.4.19 series
6468 - generate true UUIDs in lmc (1171)
6469 - have portals stack dumping break in UML (2466)
6470 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6471 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6472 - wake callback waiting threads on client eviction (2460)
6473 - Add --ptldebug and --subsystem to lmc (1719)
6474 - update assertion to allow safe interrupt allocation
6475 - set rq_no_resend for cancel requests (2432)
6476 - recalculate ptlrpcd timeout after resend (2494)
6477 - call vfs_rmdir when removing pending directories (2368)
6478 - fix renaming a file to itself (2429)
6479 - lmc creates a default one-stripe lov (2454)
6480 - expand procfs space to handle large clusters (2326)
6481 - increase UML stack to avoid overflow
6482 - update lconf's list of debug and subsystem masks
6483 - fix lfs find --obd (2510)
6484 - /proc tunable for disabling filter read caching (2591)
6485 - stop rpm packages from altering slapd.conf (2301)
6486 - disable nagle in the socknal under 0conf (2578)
6487 - choose mds inode size based on stripe count (2572)
6488 - fix kernel-source rpm problems (2516)
6490 - add --disable-doc to avoid pdf generation (2421)
6491 - update documentation, tests, type-os, comments
6492 - avoid format warnings on ia64
6493 - remove the TOE NAL
6494 - tiny code cleanups by removing unused fields
6496 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6499 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6500 - remove the most blatant lies from BUILDING file (2371)
6501 - change default debug level to reasonable production setting
6502 - reduce client side cache size to reduce cache flush time
6503 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6504 - make TCP zerocopy and pinger support enabled by default (2476)
6505 - sync writes completed after process exits caused crashes (2319)
6506 - maintain correct mount count on the MDS (2356)
6507 - backout 1557, because 2316 wasn't really fixed
6508 - better file I/O statistics gathering in /proc
6509 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6510 - another kernel patch to fix zero-copy TCP function export
6511 - don't take duplicate lock when processing re-sent getattr (2420)
6512 - lctl uses obd_self_export instead of creating new conn (2353)
6513 - MDS/OST recovery case which requires object creation asserted (2425)
6514 - move lfs from /usr/sbin to /usr/bin in packages
6515 - fix race between mds_client_add and mds_client_free (2417)
6516 - use kmalloc instead of slabs in portals (2430)
6517 - don't create duplicate records when a failover MDS is present (2442)
6518 - remove unnecessary mount age check (2332)
6519 - don't remove directory inodes from locks prematurely (2451)
6520 - don't break if MDS service name is the same as hostname (2103)
6521 - fix races in client write RPC generation when cache full (2482)
6523 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6526 - remove now-unused request->rq_obd (278)
6527 - if an allocation fails, print out how much memory we've used (1933)
6528 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6529 - add the "configurable stack size" patch to most series files (1256)
6530 - ability to write large log records, for 100+ OST configs (2306)
6531 - fix NULL deref when filter_prep fails (2314)
6532 - fix operator precedence error in filter_sync
6533 - dynamic allocation of socknal TX descriptors (2315)
6534 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6535 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6536 - ensure that CWARN messages in Portals always get to the syslog
6537 - __init/__exit are not for prototype decls (ldlm_init/exit)
6538 - x86-64 compile warning fixes
6539 - fix gateway LMC keyword conflict (2318)
6540 - fix MDS lock inversions in getattr/reint paths (1844)
6541 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6542 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6543 - DLM race condition prevented some lock evictions (2328)
6544 - ENOMEM detection and retry on socknal sends (2230)
6545 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6546 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6547 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6548 - better error messages when a client is rejected during recovery (1505)
6549 - avoid cancelling locks which were never granted, after failure (2330)
6550 - fix i_sem/journal inversion in mds_client_add (2333)
6551 - fix truncate/getattr lock cycle deadlock (2334)
6552 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6553 - fix two rare exit paths which could leak an l_lock() ref (2321)
6554 - fencepost error in MDS/OST orphan recovery (2226)
6555 - make log record alignment 8 bytes (1988)
6556 - lstripe now fails when requested offset > ost_count (2237)
6557 - ensure that all kernel series have a complete list.h (1607)
6558 - fix crashes in special-file operations (2316)
6559 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6561 - allow configurable automake binary, for testing new versions
6562 - small update to the lfs documentation
6564 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6566 * fix negative export reference count in fsfilt_sync (2312)
6568 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6569 * release candidate 0.9.1
6571 - orphans are moved into the PENDING directory for possible recovery
6572 - replayed opens now open by fid for orphan/rename safety (1042)
6573 - last close of an orphan inode generates a transno (683)
6574 - chdir() and mount() now pin the directory entry (1020)
6575 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6576 - discard very old requests without processing them (1502)
6577 - remove some common, well-understood CERRORs (1505)
6578 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6579 - clear "grant" flags in OST replies until OST grant code lands (1644)
6580 - fix read performance by not clobbering i_blksize on client (1598)
6581 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6582 - make LRU size a /proc tunable, clears locks when reduced (707)
6583 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6584 - server locks take references on exports now (1558)
6585 - build fixes for 2.4.20-rh trees (1663)
6586 - return an error from lov_create if all OSCs are inactive (1751)
6587 - fix import levels when a reconnect happens without a timeout (1597)
6588 - exit early from mds_open if we get a lookup error (1749)
6589 - partial page read at EOF wouldn't wait for disk before sending (1642)
6590 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6591 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6592 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6593 - write/truncate lock inversion (1639)
6594 - Don't auto-load obdclass, portals modules during cleanup (1495)
6595 - fix timestamps from jumping to "now" (1763)
6596 - extra journal assertions (1648)
6597 - add an extra multiunlink test (1771)
6598 - fix read_record/write_record API (1776)
6599 - fix leak of offset_extent, possible incorrect i_size later (1772)
6600 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6601 - Debugging for the kqswnal_get_idle_tx problems (1820)
6602 - Allow recovery to be attempted multiple times (1536)
6603 - Write out MDS last_rcvd file after it is first created (1600)
6604 - Fix tx_descriptor leak in failed transmit situations (1827)
6605 - ext3 journaling fixes for assertion failure after IO error (1871)
6606 - class_export_put() on freed export after completion AST error (1896)
6607 - Fix revalidate looping in VFS (1322)
6608 - Don't access a freed export during MDS_REINT timeout (1521)
6609 - Add open-unlink recovery support on the MDS (1673,1764)
6610 - Return an error if no MDS data was read from last_rcvd (1946)
6611 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6612 - Don't setup a disk device that doesn't match exported UUID (317)
6613 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6614 - avoid committing NULL handle in force close
6615 - local.sh is now a one-stripe LOV configuration
6616 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6617 - fix ext3 htree duplicate directory entry corruption (1516)
6618 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6619 - update to new LOV EA format (2097)
6620 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6622 2003-06-15 Phil Schwan <phil@clusterfs.com>
6625 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6626 - per-import/export recovery handling (958, 931, 959)
6627 - multiple last-rcvd slots, for serving multiple FSes (949)
6628 - connections are again shared between multiple imp/exports (963, 964)
6629 - "umount -f" would hang if any requests needed to be sent (393, 978)
6630 - avoid pinning large req buffer by copying for queued messages (989)
6631 - add "uuid" to "lctl device" command to help upcalls (991)
6632 - "open" RPCs with transnos would confuse recovery counters (1037)
6633 - do proper endian conversion of all wire messages (288, 340, 891)
6634 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6635 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6636 - fix ptlrpc_connection leak in target_handle_connect (1174)
6637 - fix import refcounting bug in OST and MDS cleanup (1134)
6638 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6639 - fix very unlikely obd_types race condition (501)
6640 - remove osc_open hack for echo_client (1187)
6641 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6642 - a failure in read_inode2 leads to deadlock (1139)
6643 - cancel ack-locks as soon as transaction is committed (1072)
6644 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6645 - make sure to commitrw after any preprw to avoid deadlock (1162)
6646 - failing to execute a file in a lustre FS would lock inode (1203)
6647 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6648 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6649 - obd_uuid2tgt would walk off the end of the list (1255)
6650 - on IA64 the osc would give portals incorrect bulk size (1258)
6651 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6652 - fix lock inversion caused by new llite matching code (1282)
6653 - limit the number of dirty pages on a client to 10MB (1286)
6654 - timed out locks were not being corrected cancelled (1289)
6655 - fix O_DIRECT above 4GB on IA-32 (1292)
6656 * major user-visible changes
6657 - fail out/fail over policy now controlled by the upcall (993)
6659 - add OBD_PING to check server availability and failure (954)
6660 - lustre messages are now sent in sending host order (288, 340, 891)
6661 - add eadatalen to MDS getattr reply (340)
6662 - OST read replies may contain second buffer, with per-page status (593)
6664 2003-03-11 Phil Schwan <phil@clusterfs.com>
6667 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6668 - failed open()s could cause deadlock; fixed (867, 869)
6669 - stop cancelling OST locks when files are closed (481)
6670 - overlapping XID spaces caused network corruption (851, 853)
6671 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6672 - setattr_raw added, to do single-RPC, server-side setattrs
6673 - lmc/lconf syntax change for OST UUIDs
6674 - fix crashy race condition between ptlrpc_free_req and osc_close
6675 - don't use request in mdc_enqueue if we hit a timeout (889)
6676 - don't set the inode i_size for regular files from the MDS (896)
6677 - handle out of order completion AST (842)
6678 - don't LBUG if a lock request times out after receiving AST (913)
6679 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6680 - if a bad lock AST arrives, send an error instead of dropping entirely
6681 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6682 - fix leak in bulk IO when only partially completed (899, 900, 926)
6683 - fix O_DIRECT for ia64 (55)
6684 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6685 - C-z after timeout could hang a process forever; fixed (977)
6687 - client-side I/O cache (678, 924, 929, 941, 970)
6689 - READPAGE and SETATTRs which don't take server-side locks get
6692 2003-02-11 Phil Schwan <phil@clusterfs.com>
6695 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6696 - Fix fsfilt_extN_readpage() to read a full page of directory
6697 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6698 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6699 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6700 - Most intent operations were removed, in favour of a new RPC mode
6701 that does a single RPC to the server and bypasses most of the VFS
6702 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6703 - Aggressively cancel local locks on DLM servers
6704 - mds_reint_unlink sends EA to the client if it's the last nlink.
6705 client uses that EA to unlink OST objects.
6706 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6707 - recursive symlinks were fixed (439)
6708 - fixed NULL deref in DEBUG_REQ
6709 - filter_update_lastobjid no longer calls sync, which annoyed extN
6710 - fixed multi-client small-writes to a single file problem (445)
6711 - fixed mtime updates during file writes (607)
6712 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6713 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6714 - fixed Linux OST size reporting problem (444, 656)
6715 - OST now updates object mtime with writes or setattr (607, 619)
6716 - client verifies file size before zeroing page past EOF (445)
6717 - OST now writes last allocated objid to disk with allocation (108)
6718 - LOV on echo now works (409)
6720 - mds_reint_unlink sends a new buffer, with the EA included. this
6721 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6722 set if a regular file was being unlinked, and it was the last link
6723 - use PtlGet from the target for bulk writes (315)
6724 - OST now updates object mtime with writes or setattr (607, 619)
6725 - LDLM now has a grant-time callback to revalidate locked items, if
6727 - Many MDS operations were reorganized to combat race conditions
6729 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6730 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6732 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6735 - Fully reactivate OST imports after reconnection (512, others)
6736 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6737 - More graceful error handling for truncating on dead OST (515)
6738 - Don't error out unless we're actually accessing dead stripes (474)
6739 - Fix garbage sizes when stripes are missing (410)
6740 - LRU counters were broken, causing constant lock purge (433, 432)
6741 - garbage on read from stripes with failed OSTs (441)
6742 - mark OSCs as active before reconnecting during recovery (438)
6743 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6744 - lfind did not preserve OST order in output (443)
6745 - symlinks cause hung clients, incorrect data (439)
6746 - stop dereferencing request after dropping refcount (457)
6747 - don't LASSERT(spin_is_locked) on non-SMP (455)
6748 - fixes for many rename() bugs
6749 - fstat didn't correctly synchronize attributes (399)
6750 - server must handle lock cancellation during blocking AST prep (487)
6751 - bulk descriptors were free()d too soon (511)
6752 - fix paths in lconf, which would load incorrect modules (451, 507)
6753 - fix confusing lconf 'host not found' error message (386)
6754 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6755 - fix race condition in mdc_blocking_ast() for inode access (526)
6756 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6757 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6758 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6759 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6760 - fix lstripe in several ways (kernel side) (527)
6761 - fix request leak in ldlm_cli_enqueue (262)
6762 - incorrect OSC was marked inactive after OST failure
6763 - call mds_fs_cleanup before unmounting filesystem (524)
6764 - fix races between taking ns_lock and ldlm_lock_change_resource
6765 - fix races updating LOV export open file list
6766 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6767 - fix recovery NULL deref in ldlm_cli_cancel_unused
6768 - fix some DLM races by using new hash table for lock handles (419)
6769 - permit the client to specify desired inodes, at replay
6770 - duplicate requests when we queue them for replay reintegration
6771 - fix last_rcvd offset calculation
6772 - sync after each recovered transaction, so we always make progress
6773 - never, not always, ERESTART requests without transnos
6774 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6776 - skip replay if the MDS says that the client is already connected
6777 - don't check for a recovery-enabled export to match lctl's UUID
6778 - don't INC_USE_COUNT for phantom exports
6779 - don't crash when cleaning up phantom exports (567)
6780 - don't double-finish or set replay data for errored mdc_open requests
6781 - abort requests when they time out, so we don't get old replies
6782 - send/receive replies for AST messages again
6783 - if the client says that it doesn't have the lock, cancel it on the
6785 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6786 mark it as destroyed, it will all work out when decref is called
6787 - fix module use counts (22, 581)
6789 - ASTs now expect a reply (server cancels lock on error reply)
6791 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6794 - fix many simultaneous client startup (392)
6795 - fix dentry->d_it clobbering
6796 - credentials weren't being shipped for readdir/getattr operations
6797 - remove invalid assertions triggered during some concurrent MD
6799 - proper Lustre versions added (336, 389)
6800 - fix memory leak for create error case (398)
6801 - fix LOV locking bug that would get cli/srv out of sync
6802 - fix echo client over LOV (409)
6803 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6804 - fix double-O_EXCL intent crash (424)
6805 - avoid sending multiple lock CANCELs (352)
6807 - MDS can do multi-client recovery (modulo bugs in new code)
6809 - many updates, edits, cleanups
6811 2002-11-18 Phil Schwan <phil@clusterfs.com>
6814 - fix null d_it dereference (346)
6815 - fix full OST/dbench hang (333)
6816 - fix permission problem with file removal (286)
6817 - fix removal of OSCs from LOV when they fail
6818 - fix NULL deref during bulk timeout (214)
6819 - fix problems related to multiple filesystems on one MDS (241)
6820 - fixed serious subtle metadata locking bugs
6821 - free locks on clients when inodes are removed due to memory
6823 - fix inode pointer in lock data (285)
6824 - partial support for multiple MDS on a single host (241)
6825 - data locks weren't cancelled at clear_inode time (290, 311)
6826 - intent locks could lead to unbounded lock growth (205)
6827 - added a maximum lock count, an LRU list, and a flusher
6828 - fix multiple rename (365)
6829 - properly abstracted the echo client
6830 - OSC locked 1 byte too many; fixed
6831 - rewrote brw callback code:
6832 - fixed recovery bugs related to LOVs (306)
6833 - fixed too-many-pages-in-one-write crash (191)
6834 - fixed (again) crash in sync_io_timeout (214)
6835 - probably fixed callback-related race (385)
6837 - Add capability to MDS protocol
6838 - LDLM cancellations and callbacks on different portals
6840 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6843 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6844 - defer open object destruction to close time (601981, 138)
6845 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6846 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6847 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6848 - don't call obd_setattr in ll_file_release for destroyed objects
6849 * protocol change to lustre_msg: move |version| and add |flags|
6850 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6851 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6852 * added replay of create, unlink, link and rename operations during
6853 MDS failover; recovery should be much more robust now
6854 * remove failed OSCs from LOVs (only lov_create uses this so far)
6855 * the lustre-HOWTO was brought (more) up to date (582544)
6857 2002-10-23 Phil Schwan <phil@clusterfs.com>
6860 - in-use dentries weren't being reused properly (617851)
6861 - prevent multiple LDLM setup (599178)
6862 - fix LOV size calculations for truncate (617853)
6863 - fix client handling of MDS intent errors (POSIX)
6864 - fix permission bug in lovstripe.c test (624321)
6865 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6866 - truncate past end of file could corrupt data
6867 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6868 - a race in recovery could return ETIMEDOUT to apps (623947)
6869 - building outside the source directory was fixed
6870 * the lustre-HOWTO was brought (more) up to date (582544)
6871 * major progress was made on recovery functionality
6873 2002-10-10 Phil Schwan <phil@clusterfs.com>
6876 - recovery deadlock fix
6877 - rm -rf causes LBUG fix (617817)
6878 - file open by multiple tasks fix (618962)
6879 - directory permissions bugs (602707 and 620007)
6880 - journal_stop fixed with locking (611313)
6881 - O_APPEND failures resolved (618273, perhaps 614459)
6882 - lconf PATH fix (619770)
6883 - IA64 build fix (621450)
6884 - RPC buffer sizes scale with amount of memory
6886 2002-10-01 Phil Schwan <phil@clusterfs.com>
6889 - locks would be cancelled without throwing away data pages,
6890 resulting in inconsistent data (605627)
6891 - inode attributes were not always being refreshed (605627, 612449)
6892 - lconf now continues to cleanup after lctl reports an error
6893 - MDS now enforces user permissions (602707)
6894 - lprocfs cleanup fixed, but not yet enabled (614157)
6895 - fixed infinite server hang, should a client not respond to an AST
6896 - avoid going into recovery if user calls readlink() with a buffer
6897 that's too small (613941)
6898 - AST RPCs no longer require replies (614867) -- this may be changed
6899 - don't crash server if client sends an IOV that's too big (611336)
6900 - fixed lock conversion deadlock (611892)
6901 - fixed the following of symlinks (614622)
6902 * recovery: the server can remove locks from a client that dies, other
6903 clients can make progress
6904 * more extN patch fixes
6905 * compile-time configurable ptlrpc buffer allocations
6907 - collaborative read cache document
6908 - Lustre Lite Performance CDR document-in-progress
6910 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6913 - fix typo in patch-2.4.18
6915 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6918 - clear ptlrpc request each time in handle_incoming_request()
6919 - unlink of files now destroys the object on the OST
6921 2002-09-19 Peter Braam <braam@clusterfs.com>
6923 * add hard link support
6924 * change obdfile creation method
6925 * kernel patch changed
6927 2002-09-19 Peter Braam <braam@clusterfs.com>
6930 - stack overflow bug in extN fixed
6932 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6934 * documentation updates
6935 - add man pages for config tools
6936 - update tests/README to describe testing with new config tools
6937 - finish metadata API descriptions
6938 * bug fixes and cleanups
6939 - statfs workaround for 16TB limit
6940 - LOV stripe allocation improved, can stripe on subset of OSTs
6941 - LOV file size/IO offset was wrong for files > 4GB in size
6942 - object EA data was being dropped, caused files to be unreadable
6943 - memory overflow with non-LOV OST caused memory corruption
6944 - fixed regression tests to work with new config tools, obdfilter
6945 - fixed bug when directory size became larger than 1 block
6946 - fixed bug (for single client case) when PWD was deleted
6947 - invalidate local directory pages when doing intent-based ops
6948 - avoid LDLM oops when lock callback contained bad data
6950 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6952 * documentation updates
6953 * bug fixes and cleanups
6954 - configuration tools
6957 - 64-bit compile warnings
6958 - 64-bit internal statfs data
6960 * test_brw on persistent OST devices
6962 * lprocfs (disabled)
6964 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6966 * documentation updates
6967 * bug fixes and cleanups
6968 * configuration tools
6970 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6973 * many small fixes to 0_5_4
6974 * io/network handling
6975 * thinkos in MDS operations
6977 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6980 * crucial basic fixes to 0.5.3
6981 * IOR, Iozone work over Elan
6984 2002-08-07 Phil Schwan <phil@clusterfs.com>
6985 * version 0_5_3, our first alpha
6986 * we use the new Portals iovs
6987 * documentation updates
6988 * bug fixes and cleanups
6989 * small changes in the DLM wire protocol
6991 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6992 * version 0_5_1 with some initial stability,
6993 * locking on MD and file I/O.
6994 * documentation updates
6995 * several bug fixes since 0.5.0
6996 * small changes in wire protocol
6998 2002-07-18 Phil Schwan <phil@clusterfs.com>
7000 * delivered as Lustre Light Alpha
7001 * fixed a crash after handling invalid MDS requests
7002 * fixed directory pages for architectures with non-4k pages sizes
7004 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7005 * release version v0_4_4
7006 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7008 * Fixes a number of interruption problems with OST operations.
7009 * Update documentation for portals header changes
7010 * Move all wire protocol structs/defines to lustre_idl.h
7011 * Fixes symlink length bug.
7012 * Add tcpdump to repository.
7014 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7015 * release version v0_4_3
7016 * Fixes statfs for inodes on extN.
7017 * Fixes bug in runtests which would delete /etc/hosts.
7018 * Use 64-bit object IDs wherever possible (not into VFS though)
7019 Remove ost_get_info, which is unused by lustre, and out of date.
7021 2002-07-03 Peter Braam <braam@clusterfs.com>
7022 * release version v0_4_2 Fixes a lookup error (type not passed)
7023 * move forward to head of Portals
7024 * move forward to latest Lustre kernel
7026 2002-06-25 Peter Braam <braam@clusterfs.com>
7027 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob