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 5.14.0-503.40.1.el9 (RHEL9.5)
7 5.14.0-427.42.1.el9 (RHEL9.4)
8 5.14.0-362.24.1.el9 (RHEL9.3)
9 4.18.0-553.51.1.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 5.14.0-570.16.1.el9 (RHEL9.6)
47 5.14.0-503.40.1.el9 (RHEL9.5)
48 5.14.0-427.42.1.el9 (RHEL9.4)
49 5.14.0-362.24.1.el9 (RHEL9.3)
50 4.18.0-553.51.1.el8 (RHEL8.10)
51 4.18.0-513.24.1.el8 (RHEL8.9)
52 4.18.0-477.27.1.el8 (RHEL8.8)
53 5.15.0-88 (Ubuntu 22.04)
54 6.8.0-35 (Ubuntu 24.04)
55 6.4.0-150600.23.50 (SLES15 SP6)
56 5.14.21-150500.55.73 (SLES15 SP5)
57 * Other clients known to build on these kernels at some point (others may also work):
58 3.10.0-862.14.4.el7 (RHEL7.5)
59 3.10.0-957.27.2.el7 (RHEL7.6)
60 4.14.0-49.13.1.el7a (RHEL7.5)
61 4.14.0-115.2.2.el7a (RHEL7.6)
62 3.10.0-1062.18.1.el7 (RHEL7.7)
63 3.10.0-1127.19.1.el7 (RHEL7.8)
64 3.10.0-1160.88.1.el7 (RHEL7.9)
65 4.18.0-80.11.2.el8 (RHEL8.0)
66 4.18.0-147.8.1.el8 (RHEL8.1)
67 4.18.0-193.28.1.el8 (RHEL8.2)
68 4.18.0-240.22.1.el8 (RHEL8.3)
69 4.18.0-305.25.1.el8 (RHEL8.4)
70 4.18.0-348.23.1.el8 (RHEL8.5)
71 4.18.0-372.32.1.el8 (RHEL8.6)
72 4.18.0-425.10.1.el8 (RHEL8.7)
73 5.14.0-70.30.1.el9 (RHEL9.0)
74 5.14.0-162.23.1.el9 (RHEL9.1)
75 5.14.0-284.30.1.el9 (RHEL9.2)
76 4.4.120-92.70 (SLES12 SP2)
77 4.4.180-94.100 (SLES12 SP3)
78 4.12.14-95.48 (SLES12 SP4)
79 4.12.14-122.133 (SLES12 SP5)
80 4.12.14-197.75 (SLES15 SP1)
81 5.3.18-24.96 (SLES15 SP2)
82 5.3.18-150300.59.93 (SLES15 SP3)
83 5.14.21-150400.24.100 (SLES15 SP4)
84 4.4.0-131 (Ubuntu 16.04)
85 4.15.0-48 (Ubuntu 18.04)
86 5.4.0-37 (Ubuntu 20.04)
87 5.8.0-53 (Ubuntu 20.04.2 HWE)
88 5.11.0-31 (Ubuntu 20.04.3 HWE)
89 5.11.0 (vanilla kernel.org)
90 5.10.0 (openEuler 22.03 LTS)
91 * Recommended e2fsprogs version: 1.47.2-wc1 or newer
92 * Recommended ZFS version: 2.3.2
93 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
94 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
95 stack overflow. For more information, please refer to bugzilla 17630.
96 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
97 the re-exporting nodes to be mounted with "32bitapi" mount option
99 --------------------------------------------------------------------------------
104 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
105 for currently supported client and server kernel versions.
106 * Server primary kernels built and tested during release cycle:
107 5.14.0-427.31.1.el9 (RHEL9.4)
108 5.14.0-362.24.1.el9 (RHEL9.3)
109 4.18.0-553.16.1.el8 (RHEL8.10)
110 4.18.0-513.24.1.el8 (RHEL8.9)
111 4.18.0-477.27.1.el8 (RHEL8.8)
112 * Other server kernels known to build and work at some point (others may also work):
113 3.10.0-862.14.4.el7 (RHEL7.5)
114 3.10.0-957.27.2.el7 (RHEL7.6)
115 3.10.0-1062.18.1.el7 (RHEL7.7)
116 3.10.0-1127.19.1.el7 (RHEL7.8)
117 3.10.0-1160.88.1.el7 (RHEL7.9)
118 4.18.0-80.11.2.el8 (RHEL8.0)
119 4.18.0-147.8.1.el8 (RHEL8.1)
120 4.18.0-193.28.1.el8 (RHEL8.2)
121 4.18.0-240.22.1.el8 (RHEL8.3)
122 4.18.0-305.25.1.el8 (RHEL8.4)
123 4.18.0-348.23.1.el8 (RHEL8.5)
124 4.18.0-372.32.1.el8 (RHEL8.6)
125 4.18.0-425.10.1.el8 (RHEL8.7)
126 5.14.0-162.23.1.el9 (RHEL9.1)
127 5.14.0-284.30.1.el9 (RHEL9.2)
128 4.4.120-92.70 (SLES12 SP2)
129 4.4.180-94.100 (SLES12 SP3)
130 5.14.21-150500.55.65 (SLES15 SP5)
131 4.4.0-131 (Ubuntu 16.04)
132 4.15.0-32 (Ubuntu 18.04)
133 5.4.0-48 (Ubuntu 20.04)
134 6.8.0-38 (Ubuntu 24.04)
135 6.10.0-15 (Ubuntu 24.04)
136 vanilla linux 5.4.0 (ZFS + ldiskfs)
137 vanilla linux 5.4.21 (ZFS + ldiskfs)
138 vanilla linux 5.4.136 (ZFS + ldiskfs)
139 vanilla linux 6.1.36 (ZFS + ldiskfs)
140 vanilla linux 6.6.13 (ZFS + ldiskfs)
141 5.10.0-60.94.0.118.oe2203 (openEuler 22.03 LTS)
142 5.10.0-136.32.0.108.oe2203sp1 (openEuler 22.03 LTS SP1)
143 5.10.0-153.19.0.95.oe2203sp2 (openEuler 22.03 LTS SP2)
144 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
145 * Client primary kernels built and tested during release cycle:
146 5.14.0-427.31.1.el9 (RHEL9.4)
147 5.14.0-362.24.1.el9 (RHEL9.3)
148 4.18.0-553.16.1.el8 (RHEL8.10)
149 4.18.0-513.24.1.el8 (RHEL8.9)
150 4.18.0-477.27.1.el8 (RHEL8.8)
151 5.4.0-37 (Ubuntu 20.04)
152 6.4.0-150600.23.17 (SLES15 SP6)
153 5.14.21-150500.55.73 (SLES15 SP5)
154 * Other clients known to build on these kernels at some point (others may also work):
155 3.10.0-862.14.4.el7 (RHEL7.5)
156 3.10.0-957.27.2.el7 (RHEL7.6)
157 4.14.0-49.13.1.el7a (RHEL7.5)
158 4.14.0-115.2.2.el7a (RHEL7.6)
159 3.10.0-1062.18.1.el7 (RHEL7.7)
160 3.10.0-1127.19.1.el7 (RHEL7.8)
161 3.10.0-1160.88.1.el7 (RHEL7.9)
162 4.18.0-80.11.2.el8 (RHEL8.0)
163 4.18.0-147.8.1.el8 (RHEL8.1)
164 4.18.0-193.28.1.el8 (RHEL8.2)
165 4.18.0-240.22.1.el8 (RHEL8.3)
166 4.18.0-305.25.1.el8 (RHEL8.4)
167 4.18.0-348.23.1.el8 (RHEL8.5)
168 4.18.0-372.32.1.el8 (RHEL8.6)
169 4.18.0-425.10.1.el8 (RHEL8.7)
170 5.14.0-70.30.1.el9 (RHEL9.0)
171 5.14.0-162.23.1.el9 (RHEL9.1)
172 5.14.0-284.30.1.el9 (RHEL9.2)
173 4.4.120-92.70 (SLES12 SP2)
174 4.4.180-94.100 (SLES12 SP3)
175 4.12.14-95.48 (SLES12 SP4)
176 4.12.14-122.133 (SLES12 SP5)
177 4.12.14-197.75 (SLES15 SP1)
178 5.3.18-24.96 (SLES15 SP2)
179 5.3.18-150300.59.93 (SLES15 SP3)
180 5.14.21-150400.24.100 (SLES15 SP4)
181 4.4.0-131 (Ubuntu 16.04)
182 4.15.0-48 (Ubuntu 18.04)
183 5.8.0-53 (Ubuntu 20.04.2 HWE)
184 5.11.0-31 (Ubuntu 20.04.3 HWE)
185 5.11.0 (vanilla kernel.org)
186 5.10.0 (openEuler 22.03 LTS)
187 * Recommended e2fsprogs version: 1.47.1-wc1 or newer
188 * Recommended ZFS version: 2.1.15
189 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
190 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
191 stack overflow. For more information, please refer to bugzilla 17630.
192 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
193 the re-exporting nodes to be mounted with "32bitapi" mount option
195 --------------------------------------------------------------------------------
200 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
201 for currently supported client and server kernel versions.
202 * Server primary kernels built and tested during release cycle:
203 4.18.0-348.2.1.el8 (RHEL8.5)
204 * Other server kernels known to build and work at some point (others may also work):
205 3.10.0-862.14.4.el7 (RHEL7.5)
206 3.10.0-957.27.2.el7 (RHEL7.6)
207 3.10.0-1062.18.1.el7 (RHEL7.7)
208 3.10.0-1127.19.1.el7 (RHEL7.8)
209 3.10.0-1160.49.1.el7 (RHEL7.9)
210 4.18.0-80.11.2.el8 (RHEL8.0)
211 4.18.0-147.8.1.el8 (RHEL8.1)
212 4.18.0-193.28.1.el8 (RHEL8.2)
213 4.18.0-240.22.1.el8 (RHEL8.3)
214 4.18.0-305.25.1.el8 (RHEL8.4)
215 4.4.120-92.70 (SLES12 SP2)
216 4.4.180-94.100 (SLES12 SP3)
217 4.4.0-131 (Ubuntu 16.04)
218 4.15.0-32 (Ubuntu 18.04)
219 5.4.0-48 (Ubuntu 20.04)
220 vanilla linux 5.4.0 (ZFS + ldiskfs)
221 vanilla linux 5.4.21 (ZFS + ldiskfs)
222 vanilla linux 5.4.136 (ZFS + ldiskfs)
223 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
224 * Client primary kernels built and tested during release cycle:
225 4.18.0-348.2.1.el8 (RHEL8.5)
226 5.4.0-37 (Ubuntu 20.04)
227 5.3.18-59.27 (SLES15 SP3)
228 * Other clients known to build on these kernels at some point (others may also work):
229 3.10.0-862.14.4.el7 (RHEL7.5)
230 3.10.0-957.27.2.el7 (RHEL7.6)
231 4.14.0-49.13.1.el7a (RHEL7.5)
232 4.14.0-115.2.2.el7a (RHEL7.6)
233 3.10.0-1062.18.1.el7 (RHEL7.7)
234 3.10.0-1127.19.1.el7 (RHEL7.8)
235 3.10.0-1160.49.1.el7 (RHEL7.9)
236 4.18.0-80.11.2.el8 (RHEL8.0)
237 4.18.0-147.8.1.el8 (RHEL8.1)
238 4.18.0-193.28.1.el8 (RHEL8.2)
239 4.18.0-240.22.1.el8 (RHEL8.3)
240 4.18.0-305.25.1.el8 (RHEL8.4)
241 4.4.120-92.70 (SLES12 SP2)
242 4.4.180-94.100 (SLES12 SP3)
243 4.12.14-95.48 (SLES12 SP4)
244 4.12.14-122.91 (SLES12 SP5)
245 4.12.14-197.75 (SLES15 SP1)
246 5.3.18-24.96 (SLES15 SP2)
247 4.4.0-131 (Ubuntu 16.04)
248 4.15.0-48 (Ubuntu 18.04)
249 5.8.0-53 (Ubuntu 20.04.2 HWE)
250 5.11.0-31 (Ubuntu 20.04.3 HWE)
251 5.11.0 (vanilla kernel.org)
252 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
253 * Recommended ZFS version: 2.0.7
254 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
255 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
256 stack overflow. For more information, please refer to bugzilla 17630.
257 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
258 the re-exporting nodes to be mounted with "32bitapi" mount option
260 --------------------------------------------------------------------------------
264 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
265 for currently supported client and server kernel versions.
266 * Server primary kernels built and tested during release cycle:
267 3.10.0-1062.18.1.el7 (RHEL7.7)
268 3.10.0-1127.19.1.el7 (RHEL7.8)
269 3.10.0-1160.6.1.el7 (RHEL7.9)
270 * Other server kernels known to build and work at some point (others may also work):
271 3.10.0-862.14.4.el7 (RHEL7.5)
272 3.10.0-957.27.2.el7 (RHEL7.6)
273 4.18.0-80.11.2.el8 (RHEL8.0)
274 4.18.0-147.8.1.el8 (RHEL8.1)
275 4.18.0-193.28.1.el8 (RHEL8.2)
276 4.18.0-240.1.1.el8 (RHEL8.3)
277 4.4.120-92.70 (SLES12 SP2)
278 4.4.180-94.100 (SLES12 SP3)
279 4.4.0-131 (Ubuntu 16.04)
280 4.15.0-32 (Ubuntu 18.04)
281 5.4.0-48 (Ubuntu 20.04)
282 vanilla linux 5.4.0 (ZFS + ldiskfs)
283 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
284 * Client primary kernels built and tested during release cycle:
285 3.10.0-1062.18.1.el7 (RHEL7.7)
286 3.10.0-1127.19.1.el7 (RHEL7.8)
287 3.10.0-1160.6.1.el7 (RHEL7.9)
288 4.12.14-95.48 (SLES12 SP4)
289 4.12.14-122.20 (SLES12 SP5)
290 4.12.14-197.75 (SLES15 SP1)
291 5.3.18-24.24 (SLES15 SP2)
292 4.15.0-48 (Ubuntu 18.04)
293 * Other clients known to build on these kernels at some point (others may also work):
294 3.10.0-862.14.4.el7 (RHEL7.5)
295 3.10.0-957.27.2.el7 (RHEL7.6)
296 4.14.0-49.13.1.el7a (RHEL7.5)
297 4.14.0-115.2.2.el7a (RHEL7.6)
298 4.18.0-80.11.2.el8 (RHEL8.0)
299 4.18.0-147.8.1.el8 (RHEL8.1)
300 4.18.0-193.28.1.el8 (RHEL8.2)
301 4.18.0-240.1.1.el8 (RHEL8.3)
302 4.4.120-92.70 (SLES12 SP2)
303 4.4.180-94.100 (SLES12 SP3)
304 4.4.0-131 (Ubuntu 16.04)
305 5.4.0-37 (Ubuntu 20.04)
306 5.4.0 (vanilla kernel.org)
307 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
308 * Recommended ZFS version: 2.0.0
309 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
310 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
311 stack overflow. For more information, please refer to bugzilla 17630.
312 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
313 the re-exporting nodes to be mounted with "32bitapi" mount option
315 --------------------------------------------------------------------------------
319 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
320 for currently supported client and server kernel versions.
321 * Server primary kernels built and tested during release cycle:
322 3.10.0-1062.1.1.el7 (RHEL7.7)
323 * Other server kernels known to build and work at some point (others may also work):
324 3.10.0-862.14.4.el7 (RHEL7.5)
325 3.10.0-957.27.2.el7 (RHEL7.6)
326 4.18.0-80.7.1.el8 (RHEL8.0)
327 4.4.120-92.70 (SLES12 SP2)
328 4.4.180-94.100 (SLES12 SP3)
329 4.4.0-131 (Ubuntu 16.04)
330 4.15.0-32 (Ubuntu 18.04)
331 vanilla linux 4.6.7 (ZFS only)
332 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
333 * Client primary unpatched kernels built and tested during release cycle:
334 3.10.0-1062.1.1.el7 (RHEL7.7)
335 4.12.14-95.29 (SLES12 SP4)
336 4.15.0-48 (Ubuntu 18.04)
337 * Other clients known to build on unpatched kernels at some point (others may also work):
338 3.10.0-862.14.4.el7 (RHEL7.5)
339 3.10.0-957.27.2.el7 (RHEL7.6)
340 4.14.0-49.13.1.el7a (RHEL7.5)
341 4.14.0-115.2.2.el7a (RHEL7.6)
342 4.18.0-80.7.1.el8 (RHEL8.0)
343 4.4.120-92.70 (SLES12 SP2)
344 4.4.180-94.100 (SLES12 SP3)
345 4.4.0-131 (Ubuntu 16.04)
346 vanilla linux 4.15.0 (ZFS only)
347 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
348 * Recommended ZFS / SPL version: 0.7.13
349 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
350 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
351 stack overflow. For more information, please refer to bugzilla 17630.
352 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
353 the re-exporting nodes to be mounted with "32bitapi" mount option
354 * Note for Ubuntu 14.04.5:
355 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
356 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
358 --------------------------------------------------------------------------------
362 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
363 for currently supported client and server kernel versions.
364 * Server known to build on the following kernels (others may also work):
365 2.6.32-696.18.7.el6 (RHEL6.9)
366 3.10.0-862.14.4.el7 (RHEL7.5)
367 3.10.0-957.el7 (RHEL7.6)
368 4.4.120-92.70 (SLES12 SP2)
369 4.4.162-94.69 (SLES12 SP3)
370 4.4.0-131 (Ubuntu 16.04)
371 4.15.0-32 (Ubuntu 18.04)
372 vanilla linux 4.6.7 (ZFS only)
373 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
374 * Client known to build on unpatched kernels (others may also work):
375 2.6.32-696.18.7.el6 (RHEL6.9)
376 2.6.32-754.6.3.el6 (RHEL6.10)
377 3.10.0-862.14.4.el7 (RHEL7.5)
378 3.10.0-957.el7 (RHEL7.6)
379 4.14.0-49.13.1.el7a (RHEL7.5)
380 4.14.0-115.2.2.el7a (RHEL7.6)
381 4.4.120-92.70 (SLES12 SP2)
382 4.4.162-94.69 (SLES12 SP3)
383 4.4.0-131 (Ubuntu 16.04)
384 4.15.0-32 (Ubuntu 18.04)
385 vanilla linux 4.15.0 (ZFS only)
386 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
387 * Recommended ZFS / SPL version: 0.7.9
388 * Tested with ZFS / SPL version: 0.7.9
389 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
390 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
391 stack overflow. For more information, please refer to bugzilla 17630.
392 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
393 the re-exporting nodes to be mounted with "32bitapi" mount option
394 * Note for Ubuntu 14.04.5:
395 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
396 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
398 --------------------------------------------------------------------------------
400 2018-04-03 Intel Corporation
402 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
403 for currently supported client and server kernel versions.
404 * Server known to build on patched kernels:
405 2.6.32-431.29.2.el6 (RHEL6.5)
406 2.6.32-504.30.3.el6 (RHEL6.6)
407 2.6.32-573.26.1.el6 (RHEL6.7)
408 2.6.32-642.15.1.el6 (RHEL6.8)
409 2.6.32-696.18.7.el6 (RHEL6.9)
410 3.10.0-693.21.1.el7 (RHEL7.4)
411 3.0.101-0.47.71 (SLES11 SP3)
412 3.0.101-107 (SLES11 SP4)
413 3.12.74-60.64.40 (SLES12 SP1)
414 4.4.114-92.64 (SLES12 SP2)
415 4.4.114-94.11 (SLES12 SP3)
416 3.13.0-101 (Ubuntu 14.04, ZFS only)
417 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
418 4.4.0-116 (Ubuntu 16.04)
419 vanilla linux 4.6.7 (ZFS only)
420 * Client known to build on unpatched kernels:
421 2.6.32-431.29.2.el6 (RHEL6.5)
422 2.6.32-504.30.3.el6 (RHEL6.6)
423 2.6.32-573.26.1.el6 (RHEL6.7)
424 2.6.32-642.15.1.el6 (RHEL6.8)
425 2.6.32-696.18.7.el6 (RHEL6.9)
426 3.10.0-693.21.1.el7 (RHEL7.4)
427 3.0.101-0.47.71 (SLES11 SP3)
428 3.0.101-107 (SLES11 SP4)
429 3.12.74-60.64.40 (SLES12 SP1)
430 4.4.114-92.64 (SLES12 SP2)
431 4.4.114-94.11 (SLES12 SP3)
432 3.13.0-101 (Ubuntu 14.04)
433 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
434 4.4.0-116 (Ubuntu 16.04)
436 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
437 * Recommended ZFS / SPL version: 0.7.6
438 * Tested with ZFS / SPL version: 0.7.6
439 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
440 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
441 stack overflow. For more information, please refer to bugzilla 17630.
442 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
443 the re-exporting nodes to be mounted with "32bitapi" mount option
444 * Note for Ubuntu 14.04.5:
445 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
446 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
448 --------------------------------------------------------------------------------
450 2017-07-05 Intel Corporation
452 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
453 for currently supported client and server kernel versions.
454 * Server known to build on patched kernels:
455 2.6.32-431.29.2.el6 (RHEL6.5)
456 2.6.32-504.30.3.el6 (RHEL6.6)
457 2.6.32-573.26.1.el6 (RHEL6.7)
458 2.6.32-642.15.1.el6 (RHEL6.8)
459 2.6.32-696.3.1.el6 (RHEL6.9)
460 3.10.0-514.21.1.el7 (RHEL7.3)
461 3.0.101-0.47.71 (SLES11 SP3)
462 3.0.101-100 (SLES11 SP4)
463 3.12.74-60.64.40 (SLES12 SP1)
464 4.4.59-92.17 (SLES12 SP2)
465 vanilla linux 4.6.7 (ZFS only)
466 * Client known to build on unpatched kernels:
467 2.6.32-431.29.2.el6 (RHEL6.5)
468 2.6.32-504.30.3.el6 (RHEL6.6)
469 2.6.32-573.26.1.el6 (RHEL6.7)
470 2.6.32-642.15.1.el6 (RHEL6.8)
471 2.6.32-696.3.1.el6 (RHEL6.9)
472 3.10.0-514.21.1.el7 (RHEL7.3)
473 3.0.101-0.47.71 (SLES11 SP3)
474 3.0.101-100 (SLES11 SP4)
475 3.12.74-60.64.40 (SLES12 SP1)
476 4.4.59-92.17 (SLES12 SP2)
478 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
479 * Recommended ZFS / SPL version: 0.6.5.9
480 * Tested with ZFS / SPL version: 0.6.5.9
481 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
482 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
483 stack overflow. For more information, please refer to bugzilla 17630.
484 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
485 the re-exporting nodes to be mounted with "32bitapi" mount option
487 --------------------------------------------------------------------------------
489 12-07-2016 Intel Corporation
491 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
492 for currently supported client and server kernel versions.
493 * Server known to build on patched kernels:
494 2.6.32-431.29.2.el6 (RHEL6.5)
495 2.6.32-504.30.3.el6 (RHEL6.6)
496 2.6.32-573.26.1.el6 (RHEL6.7)
497 2.6.32-642.11.1.el6 (RHEL6.8)
498 3.10.0-514.el7 (RHEL7.3)
499 3.0.101-0.47.71 (SLES11 SP3)
500 3.0.101-80 (SLES11 SP4)
501 3.12.62-60.64.8 (SLES12 SP1)
502 vanilla linux 4.5.7 (ZFS only)
503 * Client known to build on unpatched kernels:
504 2.6.32-431.29.2.el6 (RHEL6.5)
505 2.6.32-504.30.3.el6 (RHEL6.6)
506 2.6.32-573.26.1.el6 (RHEL6.7)
507 2.6.32-642.11.1.el6 (RHEL6.8)
508 3.10.0-514.el7 (RHEL7.3)
509 3.0.101-0.47.71 (SLES11 SP3)
510 3.0.101-80 (SLES11 SP4)
511 3.12.62-60.64.8 (SLES12 SP1)
513 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
514 * Recommended ZFS / SPL version: 0.6.5.8
515 * Tested with ZFS / SPL version: 0.6.5.8
516 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
517 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
518 stack overflow. For more information, please refer to bugzilla 17630.
519 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
520 the re-exporting nodes to be mounted with "32bitapi" mount option
522 --------------------------------------------------------------------------------
524 02-29-2016 Intel Corporation
526 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
527 for currently supported client and server kernel versions.
528 * Server known to build on patched kernels:
529 2.6.32-431.29.2.el6 (RHEL6.5)
530 2.6.32-504.30.3.el6 (RHEL6.6)
531 2.6.32-573.12.1.el6 (RHEL6.7)
532 3.10.0-327.3.1.el7 (RHEL7.2)
533 3.0.101-0.47.71 (SLES11 SP3)
534 3.0.101-68 (SLES11 SP4)
535 vanilla linux 4.2.1 (ZFS only)
536 * Client known to build on unpatched kernels:
537 2.6.32-431.29.2.el6 (RHEL6.5)
538 2.6.32-504.30.3.el6 (RHEL6.6)
539 2.6.32-573.12.1.el6 (RHEL6.7)
540 3.10.0-327.3.1.el7 (RHEL7.2)
541 3.0.101-0.47.71 (SLES11 SP3)
542 3.0.101-68 (SLES11 SP4)
545 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
546 * Recommended ZFS / SPL version: 0.6.4.2
547 * Tested with ZFS / SPL version: 0.6.4.2
548 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
549 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
550 stack overflow. For more information, please refer to bugzilla 17630.
551 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
552 the re-exporting nodes to be mounted with "32bitapi" mount option
554 --------------------------------------------------------------------------------
556 03-10-2015 Intel Corporation
558 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
559 for currently supported client and server kernel versions.
560 * Server known to build on patched kernels:
561 2.6.32-431.29.2.el6 (RHEL6.5)
562 2.6.32-504.8.1.el6 (RHEL6.6)
563 3.0.101-0.46 (SLES11 SP3)
564 * Client known to build on unpatched kernels:
565 2.6.32-431.29.2.el6 (RHEL6.5)
566 2.6.32-504.8.1.el6 (RHEL6.6)
567 3.10.0-123.20.1.el7 (RHEL7)
568 3.0.101-0.46 (SLES11 SP3)
569 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
570 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
571 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
572 stack overflow. For more information, please refer to bugzilla 17630.
573 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
574 the re-exporting nodes to be mounted with "32bitapi" mount option
576 Severity : enhancement
578 Description: control OST-index in IDIF via ROCOMPAT flag.
579 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
580 last_rcvd file. For new formatted OST device, it will be auto set;
581 for the case of upgrading from old OST device, you can enable it
582 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
583 enabled, for new created OST-object, its IDIF-in-LMA will contain
584 the OST-index; for the existing OST-object, the OSD will convert
585 old format IDIF as new format IDIF with OST-index stored in the
586 LMA EA when accessing such OST-object or via OI scrub. Once such
587 flag is enabled, it cannot be reverted back, so the system cannot
588 be downgraded to the orignal incompatible version.
590 --------------------------------------------------------------------------------
592 07-30-2014 Intel Corporation
594 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
595 for currently supported client and server kernel versions.
596 * Server known to build on patched kernels:
597 2.6.32-431.20.3.el6 (RHEL6.5)
598 3.0.101-0.31 (SLES11 SP3)
599 * Client known to build on unpatched kernels:
600 2.6.32-431.20.3.el6 (RHEL6.5)
601 3.0.101-0.31 (SLES11 SP3)
604 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
605 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
606 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
607 stack overflow. For more information, please refer to bugzilla 17630.
608 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
609 the re-exporting nodes to be mounted with "32bitapi" mount option
611 --------------------------------------------------------------------------------
613 10-31-2013 Intel Corporation
615 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
616 * Server builds for kernels:
617 2.6.32-358.18.1.el6 (RHEL6.4)
618 * Client builds for unpatched kernels:
619 2.6.32-358.18.1.el6 (RHEL6.4)
620 3.0.92-0.8 (SLES11 SP3)
622 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
623 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
624 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
625 stack overflow. For more information, please refer to bugzilla 17630.
626 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
627 the re-exporting nodes to be mounted with "32bitapi" mount option
629 --------------------------------------------------------------------------------
631 05-30-2013 Intel Corporation
633 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
634 * Server builds for kernels:
635 2.6.32-358.6.2.el6 (RHEL6.4)
636 * Client builds for unpatched kernels:
637 2.6.32-358.6.2.el6 (RHEL6.4)
638 3.0.74-0.6.6 (SLES11 SP2)
640 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
641 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
642 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
643 stack overflow. For more information, please refer to bugzilla 17630.
644 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
645 the re-exporting nodes to be mounted with "32bitapi" mount option
647 --------------------------------------------------------------------------------
649 9-30-2012 Whamcloud, Inc.
651 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
652 * Server support for kernels:
653 2.6.32-279.5.1.el6 (RHEL6)
654 * Client support for unpatched kernels:
655 2.6.32-279.5.1.el6 (RHEL6)
657 2.6.18-238.19.1.el5 (RHEL5)
658 2.6.16.60-0.69.1 (SLES10)
659 * Recommended e2fsprogs version: 1.41.90.wc4
660 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
661 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
662 stack overflow. For more information, please refer to bugzilla 17630.
663 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
664 the re-exporting nodes to be mounted with "32bitapi" mount option
666 Severity : enhancement
668 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
670 Severity : enhancement
672 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
674 Severity : enhancement
676 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
678 Severity : enhancement
680 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
682 Severity : enhancement
684 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
686 --------------------------------------------------------------------------------
688 04-30-2012 Whamcloud, Inc.
690 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
691 * Server support for kernels:
692 2.6.32-220.4.2.el6 (RHEL6)
693 * Client support for unpatched kernels:
694 2.6.32-220.4.2.el6 (RHEL6)
696 2.6.18-238.19.1.el5 (RHEL5)
697 2.6.16.60-0.69.1 (SLES10)
698 * Recommended e2fsprogs version: 1.41.90.wc4
699 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
700 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
701 stack overflow. For more information, please refer to bugzilla 17630.
702 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
703 the re-exporting nodes to be mounted with "32bitapi" mount option
705 Severity : enhancement
707 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
709 Severity : enhancement
711 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
713 Severity : enhancement
715 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
717 Severity : enhancement
719 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
721 --------------------------------------------------------------------------------
723 09-30-2011 Whamcloud, Inc.
725 * Server support for kernels:
726 2.6.18-238.19.1.el5 (RHEL5)
727 2.6.32-131.6.1.el6 (RHEL6)
728 * Client support for unpatched kernels:
729 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
730 2.6.18-238.19.1.el5 (RHEL5)
731 2.6.32-131.6.1.el6 (RHEL6)
732 2.6.29.4-167.fc11 (FC11)
733 2.6.16.60-0.69.1 (SLES10)
735 * Recommended e2fsprogs version: 1.41.90.wc3
736 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
737 Lustre filesystem with 4K stack may cause a stack overflow. For more
738 information, please refer to bugzilla 17630.
739 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
740 mounted with "32bitapi" mount option
742 Severity : enhancement
743 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
745 Severity : enhancement
746 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
748 Severity : enhancement
749 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
752 Frequency : only with SLES10
754 Description: Use OFED "KMP" provided by Novell
755 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
756 Lustre is now built against this package. That means you need to
757 install the ofed-kmp package from Novell for the patchless client.
758 Note that the ofed-kmp that Novell ships may not exactly match the
759 kernel version but should still be compatible.
761 Severity : enhancement
763 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
764 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
766 --------------------------------------------------------------------------------
768 2010-07-15 Oracle, Inc.
770 * Server support for kernels:
771 2.6.18-164.11.1.el5 (RHEL 5)
772 2.6.18-164.11.1.0.1.el5 (OEL 5)
773 * Client support for unpatched kernels:
774 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
775 2.6.18-164.11.1.el5 (RHEL 5),
776 2.6.18-164.11.1.0.1.el5 (OEL 5)
777 2.6.16.60-0.42.8 (SLES 10),
779 2.6.29.4-167.fc11 (FC11)
780 * Recommended e2fsprogs version: 1.41.10-sun2
782 Severity : enhancement
784 Description: Distribute lfs_migrate script with lustre
785 Details : The lfs_migrate mentioned in the manual is now included in the
786 lustre tree in lustre/scripts and will distributed in lustre RPMs.
788 Severity : enhancement
790 Description: Failover nid list added to import proc file.
792 Severity : enhancement
794 Description: "som_preview" mount option required to use SOM
795 Details : The clients will need to use som_preview mount option
799 Description: Lustre does not do 1MB IOs to HW RAID
800 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
802 Severity : enhancement
804 Description: lfs quota output is not very convenient for awk/sed-parsing
805 Details : Some positions in lfs quota output table could be empty or
806 non-empty which made it hard to parse it with scripts, now
807 a dash is put instead of space where there is not supposed
812 Description: Update to OFED 1.5.1
813 Details : OFED 1.5.1 will be used in environments that don't use
814 vendor supplied OFED.
818 Description: Remove lnet. prefix from lctl params display
819 Details : Also renamed lnet.memused to lnet_memused to
824 Description: Use conf_param -d to remove permanant settings.
828 Description: Add list_param -R
829 Details : lctl list_param now supports recursive mode.
832 Bugzilla : 21829, 22312
833 Description: Fix llobdstat.
834 Details : 1. Now updating the write_bytes counter
835 2. Port script fixes from 1.8.
839 Description: decrease the usage of memory on clients.
840 Details : 1. On clients, recycle dentries and inodes unused.
841 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
846 Description: llobdstat fix and enhancement
847 Details : add a counter to set a limit to how many samples will be returned
848 fix a wildcard in the path to limit to obdfilter stats only
852 Description: lfs check only allowed for root
853 Details : Code cleanups and added a regresion test of non-root use of
858 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
859 Details : Fix hr->hr_index overload.
863 Description: Fix FIEMAP functionality.
867 Description: FC11 patchless client support added
871 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
872 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
876 Description: New acc-sm script and various test-framework enhancments.
880 Description: Previously landed patch was reverted.
884 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
888 Description: Add ldiskfs maxdirsize mount option.
892 Description: Fix mkfs.lustre support for 16TB LUNs.
896 Description: Print mmp_check_interval
897 Details : print mmp_check_interval and make it possible to abort
898 mount operation in case it takes too long.
902 Description: Add cascading_rw to lustre mpi tests.
906 Description: Store SOM data in EA.
907 Details : Also several SOM related cleanups.
911 Description: Add refcount to md_open_data
915 Description: Use refcount on lov_request_set to prevent memory corruption.
919 Description: Statahead: use parent dir's i_mutex for synchronization.
923 Description: loadgen improvements.
927 Description: Enable peer health detection in ksocklnd and ko2iblnd.
931 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
935 Description: Fix race on obd_devcie:md_stats
939 Description: Fix SLES 11 client builds.
940 Details : Resolved uninitalized pointer gcc warnings.
944 Description: Allow non-root access for "lfs check".
945 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
949 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
951 Severity : enhancement
953 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
954 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
958 Description: Sanity test_27p fails.
959 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
960 Sanity tests 27 were cleaned up and improved.
964 Description: Unify req format on client/servers
965 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
966 buffers and swabbers approach..
971 Description: drop join file stuff.
973 Severity : enhancement
975 Description: Add 2.6.27 and SLES11 (patchless client) support
979 Description: Fix in ptlrpc_expire_one_request() to print the signed time
982 Severity : enhancement
984 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
987 Severity : enhancement
989 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
991 Severity : enhancement
993 Description: correctly handle big reply message.
994 Details : send LNet event if reply is bigger than buffer and adjust this buffer
999 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
1001 Severity : enhancement
1003 Description: Add support for OEL5.
1007 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
1009 Severity : enhancement
1011 Description: "weak-modules" support
1012 Details : Implement "weak-modules" support which enables kernel modules
1013 to be used with any kernel that implements the same kABI. In
1014 order to achieve this modules are now installed in
1015 /lib/modules/$(uname -r)/updates/kernel on all distributions.
1019 Description: client could not reconnect to OST because of an active request.
1020 Details : The client now retries to reconnect to the same server, if a
1021 connect request failed with EBUSY or -EAGAIN. Test case 35b
1022 conf-sanity.sh has been created.
1026 Description: Don't allow make backward step on assignin osc next id.
1027 Details : race between allocation next id and ll_sync thread can be cause
1028 of set wrong osc next id and can be kill valid ost objects.
1032 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
1036 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
1040 Description: Error handling in osc_statfs_interpret() has been improved.
1041 Details : Check in osc_statfs_interpret() for EBADR.
1045 Description: Do not send statfs() requests to OSTs disabled by administrator.
1046 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
1048 Severity : enhancement
1050 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
1052 Severity : enhancement
1054 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
1056 Severity : enhancement
1058 Description: Enhancement for lfs(1) command to use numeric uid/gid.
1062 Description: Improved cleanup procedure for conf-sanity test_46a.
1066 Description: Do not start bulk transfer, if client tried to reconnect.
1067 Details : Do not start bulk transfer, if client tried to reconnect,
1068 but failed due to active rpcs. Check for exp_abort_active_req
1069 has been implemented.
1073 Description: Prevent inconsistences between linux and lustre mount structures.
1074 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
1075 Make the sleep interruptible.
1077 Severity : enhancement
1079 Description: provide server to client comms path
1080 Details : server to client communications path via new LDLM_SET_INFO rpc
1084 Description: truncate starts GFP_FS allocation under transaction causing deadlock
1085 Details : ldiskfs_truncate calls grab_cache_page which may start page
1086 allocation under an open transaction. This may lead to
1087 calling prune_icache with consequent lustre reentrance.
1091 Description: send correctly lsm on open replay
1092 Details : MDS is trust to LSM size on replay open, but client can set wrong size
1097 Description: enable client interface failover
1098 Details : When a child reconnects from another NID, properly update export
1099 nid hash position and ldlm reverse import.
1101 Severity : enhancement
1103 Description: Communicate OST degraded/readonly state via statfs to MDS
1104 Details : Flags in the statfs returned from OSTs indicate whether the
1105 OST is in a degraded RAID state, or if the filesystem has
1106 turned read-only after a filesystem error is detected.
1111 Description: don't panic if EPROTO was hit when reading symlink
1112 Details : correctly handling request reference in error cases.
1114 Severity : enhancement
1116 Description: Add LustreNetLink, a kernel-userspace communcation path.
1118 Severity : enhancement
1120 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
1125 Description: access to llog context before init.
1126 Details : move handling CATALOGS file at osc layer and forbid access to llog
1127 context before init.
1130 Frequency : rare, only when client is writing to OST on local node
1132 Description: Avoid deadlock for local client writes
1133 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1134 memory freeing context. This allows OST threads to set the
1135 PF_MEMALLOC flag on task structures in order to allocate memory
1136 from reserved pools and complete IO.
1137 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1138 so that the OST threads generate memory pressure and allow
1139 inactive pages to be reclaimed.
1141 Severity : enhancement
1143 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1148 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1149 Details : move ldlm namespace creation in setup phase to avoid grab
1150 _lprocfs_lock with cli_sem held.
1154 Description: Temporarily disable grant shrink.
1155 Details : Disable the feature for debugging.
1158 Frequency : only when formatting test filesystems
1160 Description: Unable to run several mkfs.lustre on loop devices at the same time
1161 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1162 formatting. The solution is to proper handle the error.
1164 Severity : enhancement
1166 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1168 Severity : enhancement
1170 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1175 Description: resolve race between obd_disconnect and class_disconnect_exports
1176 Details : if obd_disconnect will be called to already disconnected export he
1177 forget release one reference and osc module can't unloaded.
1179 Severity : enhancement
1181 Description: Allow tuning service thread via /proc
1182 Details : For each service a new
1183 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1184 created that can be used to set min/max thread counts, and get the
1185 current number of running threads.
1190 Descriptoin: don't return error if have partially created objects for file.
1191 Details : lov_update_create_set uses set->set_success as index for created
1192 objects, so if some requests will be failed, they will have hole
1193 at end of array and qos_shrink_lsm for allocate correct lsm.
1195 Severity : enhancement
1197 Description: Update OFED support to 1.4.1
1202 Description: Reduce small size read RPC
1203 Details : Set read-ahead limit for every file and only do read-ahead when
1204 available read-ahead pages are bigger than 1M to avoid small size
1207 Severity : enhancement
1209 Description: move AT tunable parameters for more consistent usage
1210 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1212 Severity : enhancement
1214 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1215 Details : allow skip disconnected ost for send statfs request and hide error
1221 Description: fix racy locking of mballoc block bitmaps causing BUG
1222 Details : The locking of the mballoc buddy bitmap and the in-memory
1223 block bitmap was using two different spin locks in some
1224 cases. This made it possible to incorrectly access the
1225 mballoc bitmap while another process was modifying it,
1226 causing a sanity assertion to fail. While no on-disk corruption
1227 was reported, there was some risk of this happening.
1230 Frequency : rare, on llog test 6
1232 Descriptoin: don't allow connect to already connected import
1233 Details : allowing connect to already connected import is hide connecting problem.
1238 Descriptoin: allow kill process which wait statahead result
1239 Details : for some reasons 'ls' can stick in waiting result from statahead,
1240 in this case need way for kill this process.
1242 Severity : enhancement
1244 Description: Add state history info file, enhance import info file
1245 Details : Track import connection state changes in a new osc/mdc proc file;
1246 add overview-type data to the osc/mdc import proc file.
1248 Severity : enhancement
1250 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1252 Severity : enhancement
1254 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1259 Descriptoin: OST grant too much space to client even there are not enough space.
1260 Details : Client will shrink its grant cache to OST if there are no write
1261 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1262 this grant cache if there are already not enough avaible space
1263 (left_space < total_clients * 32M).
1266 Frequency : start MDS on uncleanly shutdowned MDS device
1268 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1269 Details : stay in waiting mds<>ost recovery finished produce random bugs
1270 due race between two ll_sync thread for one lov target. send
1271 ACTIVATE event only if connect realy finished and import have
1275 Frequency : rare, connect and disconnect target at same time
1277 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1278 Details : don't call obd_disconnect under lov_lock. this long time
1279 operation and can block ptlrpcd which answer to connect request.
1284 Descriptoin: don't lose wakeup for imp_recovery_waitq
1285 Details : recover_import_no_retry or invalidate_import and import_close can
1286 both sleep on imp_recovery_waitq, but we was send only one wakeup
1290 Frequency : always with long access acl
1292 Descriptoin: mds can't pack reply with long acl.
1293 Details : mds don't control size of acl but they limited by reint/getattr
1296 Severity : enhancement
1298 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1300 Severity : enhancement
1302 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1305 Frequency : start MDS on uncleanly shutdowned MDS device
1307 Descriptoin: aborting recovery hang on MDS
1308 Details : don't throttle destroy RPCs for the MDT.
1311 Frequency : on remount
1313 Description: external journal device not working after the remount
1314 Details : clear dev_rdonly flag for external journal devices in
1320 Description: shutdown vs evict race
1321 Details : client_disconnect_export vs connect request race.
1322 if client will evicted at this time - we start invalidate
1323 thread without referece to import and import can be freed
1327 Frequency : rare, need acl's on inode.
1329 Description: client can't handle ost additional correctly
1330 Details : if ost was added after client connected to mds client can have
1331 hit lnet_try_match_md ... to big messages to wide striped files.
1332 in this case need teach client to handle config events about add
1333 lov target and update client max ea size at that event.
1335 Severity : enhancement
1337 Description: Changelogs
1338 Details : Changelogs are a lightweight mechanism to track filesystem
1339 metadata and namespace changes. The changelog is recorded
1340 permanently on the MDTs, and is periodically "consumed" / purged
1341 when records are no longer needed.
1343 Severity : enhancement
1345 Description: compact fld format with extents
1346 Details : Store range of seq rather than every seq in FLD. Seq
1347 controller update FLD rather than clients. In Case of CMD, mdt0
1348 has FLD, all other metadata server act as non persistent proxy
1349 for FLD queries and cache fld entries in fld cache.
1354 Description: don't skip ost target if they assigned to file
1355 Details : Drop slow OSCs if we can, but not for requested start idx.
1356 This means "if OSC is slow and it is not the requested
1357 start OST, then it can be skipped, otherwise skip it only
1358 if it is inactive/recovering/out-of-space.
1362 Description: more cleanup in mds_lov
1363 Details : not send LOV EA under replay, we can't know about they size at this
1364 time. Don't allow client connect to mds before any ost connected,
1365 for avoid problems with LOV EA size and returning EIO to client.
1367 Severity : enhancement
1369 Description: Interoperability at server side (Disk interoperability)
1371 Severity : enhancement
1373 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1375 Severity : enhancement
1377 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1379 Severity : enhancement
1381 Description: New client IO stack (CLIO).
1383 Severity : enhancement
1385 Description: Commit on sharing. Eliminate inter-client dependencies between
1386 uncommitted transactions by doing transaction commits.
1387 Thereby clients may recovery independently.
1390 Frequency : Create a symlink file with a very long name
1392 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1393 Details : If there is no extra space in the request for early cancels,
1394 ldlm_req_handles_avail() returns 0 instead of a negative value.
1396 Severity : enhancement
1398 Description: Add /proc entry for import status
1399 Details : The mdc, osc, and mgc import directories now have
1400 an import directory that contains useful import data for debugging
1401 connection problems.
1403 Severity : enhancement
1405 Description: Re-disable certain /proc logging
1406 Details : Enable and disable client's offset_stats, extents_stats and
1407 extents_stats_per_process stats logging on the fly.
1410 Frequency : Only on FC kernels 2.6.22+
1412 Description: oops in statahead
1413 Details : Do not drop reference count for the dentry from VFS when lookup,
1414 VFS will do that by itself.
1416 Severity : enhancement
1418 Description: Generic /proc file permissions
1419 Details : Set /Proc file permissions in a more generic way to enable non-
1420 root users operate on some /proc files.
1424 Description: Hitting mdc_commit_close() ASSERTION
1425 Details : Properly handle request reference release in
1426 ll_release_openhandle().
1430 Description: quota recovery deadlock during mds failover
1431 Details : This patch includes att18982, att18236, att18237 in bz14840.
1433 1. fix osts hang when mds does failover with quotaon
1434 2. prevent watchdog storm when osts threads wait for the
1439 Frequency : only patchless client
1440 Description: add workaround for race between add/remove dentry from hash
1442 Severity : enhancement
1444 Description: Allow OST glimpses to return PW locks
1448 Description: LBUG when llog conf file is full
1449 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1453 Description: Prevent import from entering FULL state when server in recovery
1457 Description: service mount cannot take device name with ":"
1458 Details : Only when device name contains ":/" will mount treat it as
1464 Description: replace ptlrpcd with the statahead thread to interpret the async
1465 statahead RPC callback
1469 Frequency : on recovery
1470 Description: I/O failures after umount during fail back
1471 Details : if client reconnected to restarted server we need join to recovery
1472 instead of find server handler is changed and process self
1473 eviction with cancel all locks.
1475 Severity : enhancement
1477 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1479 Severity : enhancement
1481 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1483 Severity : enhancement
1485 Description: Upcall on Lustre log has been dumped
1486 Details : Allow for a user mode script to be called once a Lustre log has
1487 been dumped. It passes the filename of the dumped log to the
1488 script, the location of the script can be specified via
1489 /proc/sys/lnet/debug_log_upcall.
1494 Description: avoid idr_remove called for id which is not allocated.
1495 Details : Move assigment s_dev for clustered nfs to end of initialization,
1496 to avoid problem with error handling.
1501 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1502 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1503 into CONN_USED_HASH and this prodice warning when put connection
1504 again in unused hash.
1507 Severity : enhancement
1509 Description: File striping can now be set to use an arbitrary pool of OSTs.
1511 Severity : enhancement
1513 Description: Export bytes_read/bytes_write count on OSC/OST.
1517 Description: Early reply size mismatch, MGC loses connection
1518 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1519 the connect flags are properly negotiated.
1524 Description: quotas are not honored with O_DIRECT
1525 Details : all writes with the flag O_DIRECT will use grants which leads to
1526 this problem. Now using OBD_BRW_SYNC to guard this.
1530 Description: add quota statistics
1531 Details : 1. sort out quota proc entries and proc code.
1532 2. add quota statistics
1534 Severity : enhancement
1536 Description: enable quota support for HEAD.
1540 Description: Properly propagate oinfo flags from lov to osc for statfs
1541 Details : restore missing copy oi_flags to lov requests.
1543 Severity : enhancement
1545 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1546 : l_getgroups(8), lst(8), routerstat(8)
1547 Details : included man pages for llobdstat(8), llstat(8),
1548 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1550 Severity : enhancement
1552 Description: configure's --enable-quota should check the
1553 : kernel .config for CONFIG_QUOTA
1554 Details : configure is terminated if --enable-quota is passed but
1555 : no quota support is in kernel
1559 Description: Remove portals compatibility
1560 Details : Remove portals compatibility, not interoperable with releases
1565 Description: Resolve device initialization race
1566 Details : Prevent proc handler from accessing devices added to the
1567 obd_devs array but yet be intialized.
1569 Severity : enhancement
1571 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1573 Severity : enhancement
1575 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1580 Description: Using wrong pointer in osc_brw_prep_request
1581 Details : Access to array[-1] can produce panic if kernel compiled with
1582 CONFIG_PAGE_ALLOC enabled
1586 Description: Client runs out of low memory
1587 Details : Consider only lowmem when counting initial number of llap pages
1591 Description: *optional* service tags registration
1592 Details : if the "service tags" package is installed on a Lustre node
1593 When the filesystem is mounted, a local-node service tag will
1594 be created. See http://inventory.sun.com/ for more information
1595 about the Service Tags asset management system.
1599 Description: Kernel BUG tries to release flock
1600 Details : Lustre does not destroy flock lock before last reference goes
1601 away. So always drop flock locks when client is evicted and
1602 perform unlock regardless of successfulness of speaking to MDS.
1606 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1610 Description: sanity test 65a fails if stripecount of -1 is set
1611 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1616 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1617 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1620 Severity : enhancement
1622 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1626 Description: do not process already freed flock
1627 Details : flock can possibly be freed by another thread before it reaches
1628 to ldlm_flock_completion_ast.
1632 Description: LBUG during stress test
1633 Details : Need properly lock accesses the flock deadlock detection list.
1637 Description: oops in page fault handler
1638 Details : kernel page fault handler can return two special 'pages' in error
1639 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1643 Description: timeout with invalidate import.
1644 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1645 should be handled by ptlrpcd. This produce long age waiting and
1646 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1648 Severity : enhancement
1650 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1654 Description: enable MGS and MDT services start separately
1655 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1656 a co-located MGS without starting the MGS, which is a complement
1657 to 'nosvc' mount option.
1661 Frequency : after recovery
1662 Description: precreate to many object's after del orphan.
1663 Details : del orphan st in oscc last_id == next_id and this triger growing
1664 count of precreated objects. Set flag LOW to skip increase count
1665 of precreated objects.
1669 Frequency : rare, on clear nid stats
1670 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1671 Details : when clean nid stats sometimes try destroy live entry,
1672 and this produce panic in free.
1676 Description: Stack overflow during MDS log replay
1677 ease stack pressure by using a thread dealing llog_process.
1681 Description: wait until IO finished before start new when do lock cancel.
1682 Details : VM protocol want old IO finished before start new, in this case
1683 need wait until PG_writeback is cleared until check dirty flag and
1684 call writepages in lock cancel callback.
1686 Severity : enhancement
1688 Description: using special macro for print time and cleanup in includes.
1692 Description: mds_mfd_close() ASSERTION(rc == 0)
1693 Details : In mds_mfd_close(), we need protect inode's writecount change
1694 within its orphan write semaphore to prevent possible races.
1698 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1702 Frequency : rare, on shutdown ost
1703 Description: don't hit live lock with umount ost.
1704 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1705 use shrink_dcache_sb instead.
1709 Description: don't panic with use echo client
1710 Details : echo client pass NULL as client nid pointer and this produce null
1711 pointer dereference.
1715 Description: fix build on ppc32
1716 Details : compile code with -m64 flag produce wrong object file for ppc32.
1720 Description: add message levels for liblustreapi
1724 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1725 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1726 will result in an -EAGAIN and in the exit call path we call
1727 alloc_rr() which will with increasing aggressiveness attempt to
1728 aquire precreated objects on the minimum number of required OSCs.
1732 Description: Use old size assignment to avoid deadlock
1733 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1734 the scheduling while holding a spinlock. We do not need locking
1735 for size in ll_update_inode() because size is only updated from
1736 the MDS for directories or files without objects, so there is no
1737 other place to do the update, and concurrent access to such inodes
1738 are protected by the inode lock.
1742 Description: resolve "_IOWR redefined" build error on SLES10
1746 Description: dump the memory debugging after all modules are unloaded to
1747 suppress false negative in conf_sanity test 39
1749 Severity : enhancement
1751 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1756 Description: fix SLES kernel versioning
1757 Details : the kernel version for our SLES 10 kernel did not include a "-"
1758 before the "smp" at the end. while this was not a problem in
1759 general, it did mean that software trying to use the kernel
1760 version to try to detect a vendor specific kernel would fail.
1761 this was most evident by the OFED build scripts.
1765 Description: Don't update lov_desc members until making sure they are valid
1766 Details : When updating lov_desc members via proc fs, need fix their
1767 validities before doing the real update.
1771 Description: don't put request into delay list while invalidate in flight.
1772 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1773 import in flight. this produce timeout for invalidate and sometimes
1774 can cause stale data.
1778 Frequency : on ppc only
1779 Description: not convert ost objects for directory because it's not exist.
1780 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1784 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1786 Severity : enhancement
1788 Description: Update to support 2.6.22.14 vanilla kernel.
1792 Frequency : rare, on recovery
1793 Description: read procfs can produce deadlock in some situation
1794 Details : Holding lprocfs lock which send rpc can produce block for destroy
1795 obd objects and this also block reconnect with -EALREADY. This
1796 isn't fix all lprocfs bugs - but make it rare.
1798 Severity : enhancement
1800 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1803 Frequency : frequent on X2 node
1805 Description: mdc_set_open_replay_data LBUG
1806 Details : Set replay data for requests that are eligible for replay.
1810 Description: lustre_mgs: operation 101 on unconnected MGS
1811 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1812 MGC, and late on MGC cannot successfully connect to MGS and a lot
1813 of the error messages complaining that MGS is not connected.
1817 Frequency : on network error
1818 Description: panic with double free request if network error
1819 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1820 but it's true only for synchronus enqueue, for async enqueue
1821 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1824 Severity : enhancement
1826 Description: client-side metadata stat-ahead during readdir(directory readahead)
1827 Details : perform client-side metadata stat-ahead when the client detects
1828 readdir and sequential stat of dir entries therein
1831 Frequency : on start mds
1833 Description: Implement get_info(last_id) in obdfilter.
1836 Frequency : occasional
1838 Description: Correctly check stale fid, not start epoch if ost not support SOM
1839 Details : open with flag O_CREATE need set old fid in op_fid3 because
1840 op_fid2 was overwritten with new generated fid, but mds can answer
1841 with one of these two fids and both is not stale. Setattr
1842 incorrectly started an epoch and assume will be called
1843 done_writing, but without SOM done_writing ever being called.
1846 Frequency : rare, depends on device drivers and load
1848 Description: MDS or OSS nodes crash due to stack overflow
1849 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1850 In some cases, in conjunction with device drivers that use a lot
1851 of stack the MDS (or possibly OSS) service threads could overflow
1852 the stack. One change which was identified to consume additional
1853 stack has been reworked to avoid the extra stack usage.
1856 Frequency : occasional
1858 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1859 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1860 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1861 oscc_has_objects and see OSC has no objests and call
1862 oscc_internal_create to resend create request.
1864 Severity : enhancement
1866 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1868 Severity : enhancement
1870 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1873 Frequency : very rare
1875 Description: Fix replay if there is an un-replied request and open
1876 Details : In some cases, older replay request will revert the
1877 mcd->mcd_last_xid on MDS which is used to record the client's
1878 latest sent request.
1880 Severity : enhancement
1882 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1884 Severity : enhancement
1886 Description: Add rhel5 support to HEAD.
1888 Serverity : enhancement
1890 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1895 Description: Don't allow skipping OSTs if index has been specified.
1896 Details : Don't allow skipping OSTs if index has been specified, make
1897 locking in internal create lots better.
1901 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1902 Details : access to bitfield in structure is always rounded to long
1903 and this produce problem with not atomic change any bit.
1907 Description: Lustre make rpms failed.
1908 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1909 builds Lustre rpms from tarball.
1912 Frequency : rare on shutdown ost
1914 Description: If llog cancel was not send before clean_exports phase, this can
1915 produce deadlock in llog code.
1916 Details : If llog thread has last reference to obd and call class_import_put
1917 this produce deadlock because llog_cleanup_commit_master wait when
1918 last llog_commit_thread exited, but this never success because was
1919 called from llog_commit_thread.
1923 Description: allow userland application know is lost one of stripes.
1924 Details : fill lvb_blocks with error code on ost and return it to
1925 application if error flag found.
1929 Description: NULL lov_tgts causing MDS oops
1930 Details : more safe checks for NULL lov_tgts for avoid oops.
1932 Severity : enhancement
1934 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1938 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1939 Details : If an OST is down the MDS will hang indefinitely in
1940 obd_statfs_rqset() waiting for the statfs data. While for
1941 MDS QOS usage of statfs, it should not stuck in waiting.
1943 Severity : enhancement
1945 Description: remote_acl support
1946 Details : Support ACL-based permission check for remote user.
1947 Support setfacl/getfacl for remote user with the utils
1948 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1949 the system "{s,g}etfacl" utils.
1951 Severity : enhancement
1953 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1955 Severity : enhancement
1957 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1961 Description: make mgs_setparam() handle fsname containing dash
1962 Details : fsname containing a dash does not work with lctl conf_param
1964 Severity : enhancement
1966 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1968 Severity : enhancement
1970 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1974 Frequency : rare, at shutdown
1975 Description: access already free / zero obd_namespace.
1976 Details : if client_disconnect_export was called without force flag set,
1977 and exist connect request in flight, this can produce access to
1978 NULL pointer (or already free pointer) when connect_interpret
1979 store ocd flags in obd_namespace.
1983 Frequency : only at startup
1984 Description: not alloc memory with spinlock held.
1985 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1986 if any spinlock held.
1988 Severity : enhancement
1990 Description: make lustre randomly fail allocating memory
1991 Details : Make lustre randomly failed allocating memory for testing purpose.
1993 Severity : enhancement
1995 Description: lost problems with lov objid file
1996 Details : Fixes some scability and access to not inited memory problems
1997 in work with lov objdid file.
2002 Description: lfs find does not continue on file error
2003 Details : Continue other files processing when a file/dir is absent.
2007 Description: Inconsistent usage of lustre_pack_reply()
2008 Details : Standardize the usage of lustre_pack_reply() such that it
2009 always generate a CERROR on failure.
2012 Frequency : occasional
2014 Description: MDS hang or stay in waiting lock
2015 Details : If client receive lock with CBPENDING flag ldlm need send lock
2016 cancel as separate rpc, to avoid situation when cancel request
2017 can't processed due all i/o threads stay in wait lock.
2021 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
2025 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
2029 Description: Update to SLES9 kernel-2.6.5-7.287.3.
2033 Description: Update to RHEL4 latest kernel.
2035 Severity : enhancement
2037 Description: Build SLES10 patchless client fails
2038 Details : The configure was broken by run ./configure with
2039 --with-linux-obj=.... argument for patchless client. When the
2040 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
2041 can't search header adequately. Use absolute path such as
2042 -I($LINUX)/include instead.
2046 Description: interrupt oig_wait produce painc on resend.
2047 Details : brw_redo_request can be used for resend requests from ptlrpcd and
2048 private set, and this produce situation when rq_ptlrpcd_data not
2049 copyed to new allocated request and triggered LBUG on assert
2050 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
2051 ptlrpcd set if request is changed and can be safety changed to use
2056 Description: LASSERT_{REQ,REP}SWAB macros are buggy
2057 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
2058 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
2059 these macros and replace them with their difinitions instead.
2063 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
2064 Details : Update which_patch & target file for SLES10 latest kernel.
2067 Description: Remove client patches from SLES 10 kernel.
2068 Details : This causes SLES 10 clients to behave as patchless clients
2069 even on a Lustre-patched (server) kernel.
2071 Severity : enhancement
2073 Description: self-adjustable client's lru lists
2074 Details : use adaptive algorithm for managing client cached locks lru
2075 lists according to current server load, other client's work
2076 pattern, memory activities, etc. Both, server and client
2077 side namespaces provide number of proc tunables for controlling
2080 Severity : enhancement
2082 Description: light-weight GSS support
2083 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
2084 and reduce performance overhead.
2086 Severity : enhancement
2088 Description: Linux keyring support
2089 Details : Support using service of Linux keyring for Lustre GSS internal
2090 context refresh/cache mechanism.
2094 Description: Fix errors in lfs documentation
2095 Details : Fixes man pages
2099 Description: don't use GFP_* in generic Lustre code.
2100 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2104 Description: obdclass is limited by single OBD_ALLOC(idarray)
2105 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2109 Description: warnings with build patchless client with vanila 2.6.19 and up
2110 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2111 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2115 Frequency : Only for SLES
2116 Description: sanity_quota fail test_1
2117 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2118 file, which makes TSTID[2] inunique.
2122 Frequency : Only for SLES9
2123 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2124 Details : Update target/ChangeLog/which_patch .
2128 Frequency : Only for test_10 in sanity-quota.sh
2129 Description: a bug in quota test code
2130 Details : checking a test flag in a wrong place
2134 Frequency : Only for test_13 in sanity-quota.sh
2135 Description: a bug in quota test script
2136 Details : shouldn't check overfull allocated quota
2140 Description: jbd statistics
2141 Details : Port older jbd statistics patch for sles10
2143 Severity : enhancement
2145 Description: Add printing inode info into message about error in writepage.
2150 Description: reply_lock_interpret crash due to race with it and lock cancel.
2151 Details : Do not replay locks that are being cancelled. Do not reference
2152 locks by their address during replay, just by their handle.
2156 Frequency : When flocks are used.
2157 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2158 Details : Flock locks might destroy just granted lock if it could be merged
2159 with another existing flock, this is done in completion handler,
2160 so teach ldlm_cli_enquque_fini that this is a valid case for
2166 Description: Oops in read and write path when failing to allocate lock.
2167 Details : Check if lock allocation failed and return error back.
2171 Description: lstripe command fails for valid OST index
2172 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2173 instead of lov->desc.ld_active_tgt_count.
2177 Description: sanity.sh failed test 103
2178 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2183 Description: df doesn't work properly if diskfs blocksize != 4K
2184 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2188 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2189 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2191 Severity : enhancement
2193 Description: Adaptive timeouts
2194 Details : RPC timeouts adapt to changing server load and network
2195 conditions to reduce resend attempts and improve recovery time.
2199 Description: llapi_file_create() does not allow some changes
2200 Details : add llapi_file_open() that allows specifying the mode and
2201 open flags, and also returns an open file handle.
2205 Description: merge and cleanup kernel patches.
2206 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2210 Description: Add journal checksum support.(Kernel part)
2211 Details : The journal checksum feature adds two new flags i.e
2212 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2213 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2214 indicates that the commit block contains the checksum for
2215 the blocks described by the descriptor blocks. Now commit
2216 record can be sent to disk without waiting for descriptor
2217 blocks to be written to disk. This behavior is controlled
2218 using JBD2_FEATURE_ASYNC_COMMIT flag.
2222 Description: OSS needs mutliple precreate threads
2223 Details : Add ability to start more than one create thread per OSS.
2227 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2228 Details : execute lfs setstripe on client
2232 Description: mds_obd_create error creating tmp object
2233 Details : When the user sets quota on root, llog will be affected and can't
2234 create files and write files.
2237 Frequency : Always on ia64 patchless client, and possibly others.
2239 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2240 Details : This allows the patchless client to be loaded on architectures
2241 without this export.
2245 Description: RedHat Update kernel for RHEL5
2246 Details : Add the kernel config file for RHEL5.
2250 Description: RedHat Update kernel for RHEL5
2251 Details : Modify the kernel config file more closer RHEL5.
2255 Description: Build failure against Centos5 (RHEL5)
2256 Details : Define PAGE_SIZE when it isn't present.
2260 Description: test_7 on sanity-quota.sh can't be used on separate machines
2261 Details : add facet to handle it
2265 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2266 Details : with NFS, the anon dentry's parent was set to itself in
2267 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2268 corresponding dentry other than use rec->ur_name.
2270 Severity : enhancement
2272 Description: lfs setstripe enhancement
2273 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2277 Description: enable data checksumming by default
2278 Details : enable checksum by default, allow --disable-checksum
2279 configure option and "-o nochecksum" mount option. Checksums
2280 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2284 Description: lustre support for RHEL5
2285 Details : Add support for RHEL5.
2289 Description: Client eviction due to failover config
2290 Details : after a connection loss, the lustre client should attempt to
2291 reconnect to the last active server first before trying the
2292 other potential connections.
2296 Description: when mds and osts use different quota unit(32bit and 64bit),
2297 quota will be released repeatly.
2298 Details : void sending multiple quota reqs to mds, which will keep the
2299 status between the reqs.
2303 Description: osts not allocated evenly to files
2304 Details : change the condition to increase offset_idx
2307 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2309 Description: Data corruption for OSTs that are formatted larger than 2TB
2311 Details : When generating the bio request for lustre file writes the
2312 sector number would overflow a temporary variable before being
2313 used for the IO. The data reads correctly from Lustre (which
2314 will overflow in a similar manner) but other file data or
2315 filesystem metadata may be corrupted in some cases.
2319 Description: Tune the kernel for good SCSI performance.
2320 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2321 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2326 Description: rewrite ext2-derived code in obdclass/uuid.c
2327 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2328 from scratch preserving functionality.
2332 Description: "lfs find -obd UUID" prints directories
2333 Details : "lfs find -obd UUID" will return all directory names instead
2334 of just file names. It is incorrect because the directories
2335 do not reside on the OSTs.
2339 Description: use i_size_read and i_size_write in 2.6 port
2340 Details : replace inode->i_size access with i_size_read/write()
2343 Frequency : when removing large files
2345 Description: scheduling issue during removal of large Lustre files
2346 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2347 It causes scheduling issues when removing large files (17TB in the
2351 Frequency : only with liblustre clients on XT3
2353 Description: evictions taking too long
2354 Details : allow llrd to evict clients directly on OSTs
2357 Frequency : only on ppc
2359 Description: /proc/fs/lustre/devices broken on ppc
2360 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2361 We should make sure the type of 'index' is loff_t and then cast
2362 explicitly as needed below. Do not assign an explicitly cast
2366 Frequency : only for rhel5
2368 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2369 Details : Modify the target file & which_kernel.
2371 Severity : enhancement
2373 Description: omit set fsid for export NFS
2374 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2375 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2379 Frequency : Always, for kernels after 2.6.16
2380 Description: Fix warning idr_remove called for id=.. which is not allocated.
2381 Details : Last kernels save old s_dev before kill super and not allow
2382 to restore from callback - restore it before call kill_anon_super.
2386 Description: buffer overruns could theoretically occur
2387 Details : llapi_semantic_traverse() modifies the "path" argument by
2388 appending values to the end of the origin string, and a
2389 overrun may occur. Adding buffer overrun check in liblustreapi.
2393 Description: Fix error on 'ls .' at the top of the Lustre mount.
2394 Details : Don't revalidate dentry if it is a root dentry.
2398 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2399 Details : Modify vm-tunables-rhel4.patch.
2403 Description: Kernel config for 2.6.18-vanilla.
2404 Details : Modify targets/2.6-vanilla.target.in.
2405 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2406 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2407 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2408 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2412 Description: improve handling recoverable errors
2413 Details : if request processig with error which can be recoverable on server
2414 request should be resend, otherwise page released from cache and
2419 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2420 Details : Modify target file & which_patch.
2421 A flaw was found in the IA32 system call emulation provided
2422 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2423 value could be stored in the %RAX register, which could trigger an
2424 out-of-bounds system call table access. An untrusted local user
2425 could exploit this flaw to run code in the kernel
2426 (ie a root privilege escalation). (CVE-2007-4573).
2430 Description: O_DIRECT bypasses client statistics.
2431 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2432 read_ahead_stats were updated. copied the stats section from
2433 osc_send_oap_rpc() into async_internal().
2437 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2441 Description: change order of libsysio includes
2442 Details : '#include sysio.h' should always come before '#include xtio.h'
2446 Description: handle "serious error: objid * already exists" more gracefully
2447 Details : If LAST_ID value on disk is smaller than the objects existing in
2448 the O/0/d* directories, it indicates disk corruption and causes an
2449 LBUG(). If the object is 0-length, then we should use the existing
2450 object. This will help to avoid a full fsck in most cases.
2452 Severity : enhancement
2454 Description: adapt the lustre_config script to support the upgrade case
2455 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2456 server targets to 1.6 in parallel.
2460 Description: To avoid grant space > avaible space when the disk is almost
2461 full. Without this patch you might see the error "grant XXXX >
2462 available" or some LBUG about grant, when the disk is almost
2464 Details : In filter_check_grant, for non_grant cache write, we should
2465 check the left space by if (*left > ungranted + bytes), instead
2466 of (*left > ungranted), because only we are sure the left space
2467 is enough for another "bytes", then the ungrant space should be
2468 increase. In client, we should update cl_avail_grant only there
2469 is OBD_MD_FLGRANT in the reply.
2473 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2474 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2475 described in CVE-2007-4573. This problem could allow untrusted
2476 local users to gain root access.
2479 Frequency : when using O_DIRECT and quotas
2481 Description: Incorrect file ownership on O_DIRECT output files
2482 Details : block usage reported by 'lfs quota' does not take into account
2483 files that have been written with O_DIRECT.
2488 Description: touch file failed when fs is not full
2489 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2490 otherwise we can get ENOSP while fs is not full.
2494 Description: parallel lock callbacks
2495 Details : Instead of sending blocking and completion callbacks as separated
2496 requests, adding them to a set and sending in parallel.
2499 Frequency : only for Cray XT3
2500 Bugzilla : 12829/13455
2501 Description: Changing primary group doesn't change the group lustre assigns to
2503 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2504 overridden with the primary group provided by the group upcall,
2505 whereas the supplied fsgid can be trusted if it is in the list of
2506 supplementary groups returned by the group upcall.
2508 Severity : enhancement
2510 Description: Allow masking D_WARNING, D_ERROR messages from console
2511 Details : Console messages can now be disabled via lnet.printk.
2515 Description: User code with malformed file open parameter crashes client node
2516 Details : Before packing join_file req, all the related reference should be
2517 checked carefully in case some malformed flags cause fake
2518 join_file req on client.
2522 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2523 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2524 mode is equal to the granted mode. However, it should also check
2525 that LDLM_FL_CP_REQD is not set.
2529 Description: Heavy nfs access might result in deadlocks
2530 Details : After ELC code landed, it is now improper to enqueue any mds
2531 locks under och_sem, because enqueue might want to decide to
2532 cancel open locks for same inode we are holding och_sem for.
2536 Description: Client eviction while running blogbench
2537 Details : A lot of unlink operations with concurrent I/O can lead to a
2538 deadlock causing evictions. To address the problem, the number of
2539 oustanding OST_DESTROY requests is now throttled to
2540 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2541 callbacks are processed in priority.
2545 Description: enable ACLs on MDS by default
2546 Details : ACLs must be enabled on MDS by default.
2549 Frequency : PPC/PPC64 only
2551 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2552 Details : fix duplicated definitions between asm-ppc64/types.h and
2553 lustre_types.h on PPC.
2556 Frequency : PPC/PPC64 only
2558 Description: asm-ppc/segment.h does not exist
2559 Details : fix compile issue on PPC.
2563 Description: better handle error messages in extents code
2566 Frequency : RHEL4 only
2568 Description: mkfs is very slow on IA64/RHEL4
2569 Details : A performance regression has been discovered in the MPT Fusion
2570 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2571 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2572 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2574 Severity : enhancement
2576 Description: SNMP support enhancement
2577 Details : Adding total number of sampled request for an MDS node in snmp
2580 Severity : enhancement
2582 Description: Optimize ldlm waiting list processing for PR extent locks
2583 Details : When processing waiting list for read extent lock and meeting read
2584 lock that is same or wider to it that is not contended, skip
2585 processing rest of the list and immediatelly return current
2586 status of conflictness, since we are guaranteed there are no
2587 conflicting locks in the rest of the list.
2591 Description: Time out and refuse to reconnect
2592 Details : When the failover node is the primary node, it is possible
2593 to have two identical connections in imp_conn_list. We must
2594 compare not conn's pointers but NIDs, otherwise we can defeat
2595 connection throttling.
2599 Description: port llog fixes from b1_6 into HEAD
2600 Details : Port llog reference couting and some llog cleanups from b1_6
2601 (bug 10800) into HEAD, for protect from panic and access to already
2602 free llog structures.
2606 Description: Detect stride IO mode in read-ahead
2607 Details : When a client does stride read, read-ahead should detect that and
2608 read-ahead pages according to the detected stride pattern.
2612 Description: data checksumming impacts single node performance
2613 Details : add support for several checksum algorithm. Currently, only CRC32
2614 and Adler-32 are supported. The checksum type can be changed on
2615 the fly via /proc/fs/lustre/osc/*/checksum_type.
2619 Description: use adler32 for page checksums
2620 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2625 Description: build for x2 fails
2626 Details : fix compile issue on Cray systems.
2630 Description: Properly match for duplicate locks
2631 Details : Due to different lock order from skiplists code, we need to
2632 traverse entire list for now
2635 Frequency : only on PPC/SLES10
2637 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2638 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2639 be defined. Add a hack in mkfs_lustre.c to work around this compile
2644 Description: LASSERT on MDS when client holding flock lock dies
2645 Details : ldlm pool logic depends on number of granted locks equal to
2646 number of released locks which is not true for flock locks, so
2647 just exclude such locks from consideration.
2651 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2652 Details : Use fsfilt_sync() for both the whole filesystem sync and
2653 individual file sync to eliminate dangerous inode locking
2654 with I_LOCK that can lead to a deadlock.
2658 Description: performance in 1.6.3
2659 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2660 sure that 1MB requests are not fragmented by the block layer.
2662 Severity : enhancement
2664 Description: organize the server-side client stats on per-nid basis
2665 Details : Change the structure of stats under obdfilter and mds to
2673 The "uuid"s file would list the uuids of _active_ exports.
2674 And the clear entry is to clear all stats and stale nids.
2676 Severity : enhancement
2678 Description: eliminate client locks in face of contention
2679 Details : file contention detection and lockless i/o implementation
2680 for contended files.
2684 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2687 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2689 Description: MDS may lose file striping (and hence file data) in some cases
2690 Details : If there are additional extended attributes stored on the MDS,
2691 in particular ACLs, SELinux, or user attributes (if user_xattr
2692 is specified for the client mount options) then there is a risk
2693 of attribute loss. Additionally, the Lustre file striping
2694 needs to be larger than default (e.g. striped over all OSTs),
2695 and an additional attribute must be stored initially in the
2696 inode and then increase in size enough to be moved to the
2697 external attribute block (e.g. ACL growing in size) for file
2702 Description: skiplist implementation simplification
2703 Details : skiplists are used to group compatible locks on granted list
2704 that was implemented as tracking first and last lock of each
2705 lock group the patch changes that to using doubly linked lists
2709 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2710 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2711 because LASSERT on both the data supplied by a client, and the
2712 data on disk is dangerous and incorrect.
2714 Severity : enhancement
2716 Description: Slow truncate/writes to huge files at high offsets.
2717 Details : Directly associate cached pages to lock that protect those pages,
2718 this allows us to quickly find what pages to write and remove
2719 once lock callback is received.
2723 Description: more ldlm soft lockups
2724 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2725 starve other threads from the resource lock for a long time in
2726 case of long waiting queue, so change the debug level from
2727 D_OTHER to the less frequently used D_INFO.
2729 Severity : enhancement
2731 Description: add -gid, -group, -uid, -user options to lfs find
2735 Description: Hung threads in invalidate_inode_pages2_range
2736 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2737 once one is completed. As a result, some RPCs are stuck in the
2738 queue and are never sent.
2742 Description: filter threads hungs on waiting journal commit
2743 Details : Cleanup filter group llog code, then only filter group llog will
2744 be only created in the MDS/OST syncing process.
2748 Description: Procfs and llog threads access destoryed import sometimes.
2749 Details : Sync the import destoryed process with procfs and llog threads by
2750 the import refcount and semaphore.
2752 Severity : enhancement
2754 Description: openlock cache of b1_6 port to HEAD
2759 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2760 Details : If insertion of an extent fails, then discard the inode
2761 preallocation and free data blocks else it can lead to duplicate
2766 Description: don't always update ctime in ext3_xattr_set_handle()
2767 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2768 In some cases the ctime should not be updated, for example for
2769 2.0->1.8 compatibility it is necessary to delete an xattr and it
2770 should not update the ctime.
2774 Bugzilla : 15713/16362
2775 Description: Assertion in iopen_connect_dentry in 1.6.3
2776 Details : looking up an inode via iopen with the wrong generation number can
2777 populate the dcache with a disconneced dentry while the inode
2778 number is in the process of being reallocated. This causes an
2779 assertion failure in iopen since the inode's dentry list contains
2780 both a connected and disconnected dentry.
2784 Description: assertion failure in ldlm_handle2lock()
2785 Details : fix a race between class_handle_unhash() and class_handle2object()
2786 introduced in lustre 1.6.5 by bug 13622.
2791 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2792 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2797 Description: oops in sd_iostats_seq_show()
2798 Details : unloading/reloading the scsi low level driver triggers a kernel
2799 bug when trying to access the sd iostat file.
2804 Description: Kernel panics during QLogic driver reload
2805 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2806 patch, causing memory corruption.
2811 Description: journal_dev option does not work in b1_6
2812 Details : pass mount option during pre-mount.
2814 Severity : enhancement
2816 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2817 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2818 extent information of a file. It can be used to map logical blocks
2819 in a file to physical blocks in the block device.
2823 Description: LDLM soft lockups - improvement
2824 Details : It is be possible to send the lock handle along with each read
2825 or write request because the client is already doing a lock match
2826 itself so there isn't any reason the OST should have to re-do that
2830 Frequency : only on Cray X2
2832 Description: X2 build failures
2833 Details : fix build failures on Cray X2.
2837 Description: xid & resent requests
2838 Details : Initialize RPC XID from clock at startup (randomly if clock is
2841 Severity : enhancement
2843 Description: Add lustre_start utility to start or stop multiple Lustre servers
2848 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2849 Details : In case of memory pressure, list_del() can be called twice on
2850 req->rq_history_list, causing a kernel oops.
2854 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2855 Details : fix stack overflow in the distributed lock manager by defering
2856 export eviction after a failed AST to the elt thread instead of
2857 handling it in the dlm interpret routine.
2861 Description: Convert some comments to new format.
2862 Details : Update documenting comments to match doxygen conventions.
2866 Description: Grammar fixes.
2867 Details : A couple of trivial spelling fixes.
2871 Description: OSD_COUNTERS-mandatory
2872 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2873 to simplify the code.
2877 Description: simplify cmm_device freeing logic.
2878 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2882 Description: Add lockdep support to dt_object_operations locking interface.
2883 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2884 parameter indicating lock ordering. Update mdd code to use new
2889 Description: Introduce failloc constants for lockless IO tests.
2890 Details : Add two new failloc constants to test lockless IO. Only one of
2891 them in implemented---another is checked in yet to be landed
2896 Description: Add lockdep support for inode mutex.
2897 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2898 in the situations where Lustre has to lock more than one inode
2903 Description: Add optional invariants checking support.
2904 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2905 configure switch. This macro is to be used for consistency and
2906 sanity checks that are too expensive to be left in `production'
2911 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2912 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2917 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2918 Details : Use spin_lock_nested() in (the only) situation where more than
2919 one ldlm_lock is locked simultaneously. Also, fix possible
2920 dead-lock in ldlm_lock_change_resource() by enforcing particular
2925 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2926 Details : Instead of specifying each ldlm_lock call-back through separate
2927 parameter, wrap them into struct ldlm_callback_suite.
2931 Description: Kill join_lru obd method and its callers.
2932 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2933 in memory---a requirement implemented through ->o_join_lru() obd
2934 method in HEAD. Remove this method and its users.
2938 Description: Add asynchronous ldlm ENQUEUE completion handler.
2939 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2940 case that stock ldlm_completion_ast() cannot handle as it waits
2941 until lock is granted. Introduce new ldlm_completion_ast_async()
2942 for this. Also comment ldlm_completion_ast().
2946 Description: ldlm_error <-> errno conversion.
2947 Details : Add functions to map (rather arbitrary) between LDLM error codes
2948 and standard errno values. CLIO needs this to prevent LDLM specific
2949 constants from escaping ldlm and osc.
2953 Description: Kill unused ldlm_handle2lock_ns() function.
2954 Details : Kill unused ldlm_handle2lock_ns() function.
2958 Description: Add lu_ref support to ldlm_lock
2959 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2960 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2961 and ldlm_resource. LDLM interface has to be changed, because code that
2962 releases a reference on a lock, has to "know" what reference this is.
2963 In the most frequent case
2965 lock = ldlm_handle2lock(handle);
2967 LDLM_LOCK_PUT(lock);
2969 no changes are required. When any other reference (received _not_ from
2970 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2971 instead of LDLM_LOCK_PUT().
2973 Arguably, changes are pervasive, and interface requires some discipline
2974 for proper use. On the other hand, it was very instrumental in finding
2975 a few leaked lock references.
2979 Description: Add ldlm_lock_addref_try().
2980 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2981 attempts to addref a lock that might be being canceled
2986 Description: Add ldlm_weigh_callback().
2987 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2988 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2989 hard-coded `number of pages' logic.
2993 Description: Add lockdep annotations to llog code.
2994 Details : Use appropriately tagged _nested() locking calls in the places
2995 where llog takes more than one ->lgh_lock lock.
2999 Description: Add loi_kms_set().
3000 Details : Wrap kms updates into a helper function.
3004 Description: Constify instances of struct lsm_operations.
3005 Details : Constify instances of struct lsm_operations.
3009 Description: lu_conf support.
3010 Details : On a server, a file system object is uniquely identified
3011 by a fid, which is sufficient to locate and load all object
3012 state (inode). On a client, on the other hand, more data are
3013 necessary instantiate an object. Change lu_object_find() and
3014 friends to take additional `lu_conf' argument describing object.
3015 Typically this includes layout information.
3019 Description: lu_context fixes.
3020 Details : Introduce new lu_context functions that are needed on the client
3021 side, where some system threads (ptlrpcd) are shared by multiple
3022 modules, and so cannot be stopped during module shutdown.
3026 Description: Add start and stop methods to lu_device_type_operations.
3027 Details : Introduce two new methods in lu_device_type_operations, that are
3028 invoked when first instance of a given type is created and last one
3029 is destroyed respectively. This is need by CLIO.
3033 Description: Add lu_ref support to struct lu_device.
3034 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
3035 track leaked references.
3039 Description: Introduce lu_kmem_descr.
3040 Details : lu_kmem_descr and its companion interface allow to create
3041 and destroy a number of kmem caches at once.
3045 Description: Fix lu_object finalization race.
3046 Details : Fix a race between lu_object_find() finding an object and its
3047 concurrent finalization. This race is (most likely) not possible
3048 on the server, but might happen on the client.
3052 Description: Introduce lu_ref interface.
3053 Details : lu_ref is a debugging module allowing to track references to
3054 a given object. It is quite cpu expensive, and has to be
3055 explicitly enabled with --enable-lu_ref. See usage description
3060 Description: Factor lu_site procfs stats into a separate function.
3061 Details : Separate lu_site stats printing code into a separate function
3062 that can be reused on a client.
3066 Description: Constify instances of struct {lu,dt,md}_device_operations.
3067 Details : Constify instances of struct {lu,dt,md}_device_operations.
3071 Description: Introduce struct md_site and move meta-data specific parts of
3072 struct lu_site here.
3073 Details : Move md-specific fields out of struct lu_site into special struct
3074 md_site, so that lu_site can be used on a client.
3078 Description: Kill mdd_lov_destroy().
3079 Details : Remove unused mdd code.
3083 Description: Add st_block checking to multistat.c.
3084 Details : Add st_block checking to multistat.c.
3088 Description: Add lu_ref support to struct obd_device.
3089 Details : Add lu_ref tracking to obd_device.
3093 Description: Kill obd_set_fail_loc().
3094 Details : Remove unused code.
3098 Description: Add special type for ptlrpc_request interpret functions.
3099 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
3100 there. Actual usage will be in CLIO.
3104 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
3105 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
3106 doesn't work with lockdep.
3110 Description: Add rwv.c test program.
3111 Details : New testing program exercising readv(2) and writev(2) (Qian).
3115 Description: Add sendfile.c test program.
3116 Details : New testing program exercising sendfile(2) (Jay).
3120 Description: Ratelimit a message that can be very frequent.
3121 Details : Ratelimit a memory allocation failure message that can
3126 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3127 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3131 Description: Liblustre build fixes.
3132 Details : Liblustre build fixes.
3136 Description: libcfs: add cfs_{need,cond}_resched() interface.
3137 Details : libcfs: add cfs_{need,cond}_resched() definition and
3138 implementations for Linux, NT, and liblustre.
3140 Severity : enhancement
3142 Description: More exported tunables for mballoc
3143 Details : Add support for tunable preallocation window and new tunables for
3144 large/small requests
3148 Description: Detect corruption of block bitmap and checking for preallocations
3149 Details : Checks validity of on-disk block bitmap. Also it does better
3150 checking of number of applied preallocations. When corruption is
3151 found, it turns filesystem readonly to prevent further corruptions.
3155 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3156 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3157 window to 0. If it does hit cache miss, and read-pattern is still
3158 stride-io mode, does not reset the stride window, but also does not
3159 increase the stride window length in this case.
3163 Frequency : only for big-endian servers
3164 Description: Check if system is big-endian while mounting fs with extents feature
3165 Details : Mounting a filesystem with extents feature will fail on big-endian
3166 systems since ext3-based ldiskfs is not supported on big-endian
3167 systems. This can be over-riden with "bigendian_extents" mount option.
3169 Severity : enhancement
3171 Description: The root squash functionality
3172 Details : A security feature, which is to prevent users from being able
3173 to mount lustre on their desktop, run as root, and delete
3174 all of the files in the filesystem. The goal is accomplished by
3175 remapping user id (UID) and group id (GID) of the root user to
3176 a UID and GID specified by the system administartor via Lustre
3177 configuration management server (MGS). The functionality also
3178 allows to specify sets of clients for which the remapping does
3183 Description: Excessive recovery window
3184 Details : With AT enabled, the recovery window can be excessively long (6000+
3185 seconds). To address this problem, we no longer use
3186 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3187 timeout no longer depends on the service time, it is set to
3188 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3189 time via pb_service_time.
3193 Description: Watchdog triggered on MDS failover
3194 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3195 the OSTs know that the MDS "UUID" can be reused for the same export
3196 from a different NID, so we do not need to wait for the export to be
3200 Frequency : rare, only if using MMP with Linux RAID
3202 Description: MMP doesn't work with Linux RAID
3203 Details : While using HA for Lustre servers with Linux RAID, it is possible
3204 that MMP will not detect multiple mounts. To make this work we
3205 need to unplug the device queue in RAID when the MMP block is being
3206 written. Also while reading the MMP block, we should read it from
3207 disk and not the cached one.
3209 Severity : enhancement
3211 Description: open file using fid
3212 Details : A file can be opened using just its fid, like
3213 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3216 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3219 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3220 Details : add proper locking for creation of jbd_4k slab cache
3224 Description: MMP check in ext3_remount() fails without displaying any error
3225 Details : When multiple mount protection fails during remount, proper error
3228 Severity : enhancement
3230 Description: Allow stripe size to be up to 4G-64k
3231 Details : Fix math logic to allow large stripe sizes.
3235 Description: add check for >8TB ldiskfs filesystems
3236 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3237 Don't allow >8TB ldiskfs filesystems to be mounted without
3238 force_over_8tb mount option
3243 Descriptoin: OSS complains "too many blocks" and crashes
3244 Details : In filter_commitrw_write():
3245 1. lquota_chkquota();
3246 2. obdo_from_inode();
3247 3. lquota_pending_commit();
3248 "2." has changed uid and gid in this case, then caused
3249 lquota_chkquota() and lquota_pending_commit() to use
3250 different uid and gid. That is the root of the bug.
3254 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3255 Details : Detect and print out the 'stripe count = -1' case properly.
3259 Description: Out or order replies might be lost on replay
3260 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3261 request with tid smaller than one currently being added, add it
3262 to the start, not end of the list.
3264 --------------------------------------------------------------------------------
3266 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3268 * Support for kernels:
3269 2.6.5-7.283 (SLES 9),
3270 2.6.9-55.EL (RHEL 4),
3271 2.6.16.46-0.14 (SLES 10),
3272 2.6.18.8 vanilla (kernel.org)
3273 * Client support for unpatched kernels:
3274 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3275 2.6.16 - 2.6.22 vanilla (kernel.org)
3276 * Due to recently discovered recovery problems, we do not recommend
3277 using patchless RHEL 4 clients with this or any earlier release.
3278 * Recommended e2fsprogs version: 1.39.cfs8
3279 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3280 * Starting with this release, the ldiskfs backing filesystem required
3281 by Lustre is now in its own package, lustre-ldiskfs. This package
3282 should be installed. It is versioned separately from Lustre and
3283 may be released separately in future.
3287 Description: block reactivating mgc import until all deactivates complete
3288 Details : Fix race when failing back MDT/MGS to itself (testing)
3291 Frequency : at statup only
3293 Description: mds_lov_synchronize race leads to various problems
3294 Details : simultaneous MDT->OST connections at startup can cause the
3295 sync to abort, leaving the OSC in a bad state.
3297 Severity : enhancement
3299 Description: add optional extra BUILD_VERSION info
3300 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3301 to override the lustre version.
3304 Frequency : 2.6.18 servers only
3306 Description: ll_kern_mount() doesn't release the module reference
3307 Details : The ldiskfs module reference count never drops down to 0
3308 because ll_kern_mount() doesn't release the module reference.
3313 Description: server LBUG when using old ost_num_threads parameter
3314 Details : Accept the old ost_num_threads parameter but warn that it
3315 is deprecated, and fix an off-by-one error that caused an LBUG.
3320 Description: Transient SCSI error results in persistent IO issue
3321 Details : iobuf->dr_error is not reinitialized to 0 between two
3325 Frequency : sometimes when underlying device returns I/O errors
3327 Description: OSTs not going read-only during write failures
3328 Details : OSTs are not remounted read-only when the journal commit threads
3329 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3330 instead of the ext3 wrappers.
3334 Description: poor connect scaling with increasing client count
3335 Details : Don't run filter_grant_sanity_check for more than 100 exports
3336 to improve scaling for large numbers of clients.
3339 Frequency : SLES10 only
3341 Description: sanity-quota.sh quotacheck failed: rc = -22
3342 Details : Quotas cannot be enabled on SLES10.
3345 Frequency : liblustre clients only
3347 Description: getdirentries does not give error when run on compute nodes
3348 Details : getdirentries does not fail when the size specified as an argument
3349 is too small to contain at least one entry
3351 Severity : enhancement
3353 Description: Add LNET router traceability for debug purposes
3354 Details : If a checksum failure occurs with a router as part of the
3355 IO path, the NID of the last router that forwarded the bulk data
3356 is printed so it can be identified.
3361 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3362 Details : Due to a race condition, liblustre clients were occasionally
3363 evicted incorrectly.
3365 Severity : enhancement
3367 Description: lfs setstripe use optional parameters instead of postional
3370 Severity : enhancement
3372 Description: Nanosecond timestamp support for ldiskfs
3373 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3374 resolution timestamps. There is not yet support for this at
3375 the Lustre filesystem level.
3378 Frequency : during server recovery
3380 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3381 Details : request with rq_no_resend flag not awake l_wait_event if they get
3385 Frequency : nfs export on patchless client
3387 Description: connectathon hang when test nfs export over patchless client
3388 Details : Disconnected dentry cannot be found with lookup, so we do not need
3389 to unhash it or make it invalid
3392 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3393 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3394 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3399 Description: open req refcounting wrong on reconnect
3400 Details : If reconnect happened between getting open reply from server and
3401 call to mdc_set_replay_data in ll_file_open, we will schedule
3402 replay for unreferenced request that we are about to free.
3403 Subsequent close will crash in variety of ways.
3404 Check that request is still eligible for replay in
3405 mdc_set_replay_data().
3410 Description: disable writes to filesystem when reading health_check file
3411 Details : the default for reading the health_check proc file has changed
3412 to NOT do a journal transaction and write to disk, because this
3413 can cause reads of the /proc file to hang and block HA state
3414 checking on a healthy but otherwise heavily loaded system. It
3415 is possible to return to the previous behaviour during configure
3416 with --enable-health-write.
3418 Severity : enhancement
3420 Description: 64-bit inode version
3421 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3422 to the inode. This will be required for version-based recovery.
3427 Description: MDS fails to start if a duplicate client export is detected
3428 Details : in some rare cases it was possible for a client to connect to
3429 an MDS multiple times. Upon recovery the MDS would detect this
3430 and fail during startup. Handle this more gracefully.
3432 Severity : enhancement
3434 Description: Add -o localflock option to simulate old noflock
3436 Details : This will achieve local-only flock/fcntl locks
3442 Description: log_commit_thread vs filter_destroy race leads to crash
3443 Details : Take import reference before releasing llog record semaphore
3448 Description: Wrong request locking in request set processing
3449 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3450 imp_delayed_list, in this place should be used imp_lock.
3453 Frequency : when reconnection
3455 Description: Grant Leak when osc reconnect to OST
3456 Details : When osc reconnect ost, OST(filter) should check whether it
3457 should grant more space to client by comparing fed_grant and
3458 cl_avail_grant, and return the granted space to client instead
3459 of "new granted" space, because client will call osc_init_grant
3460 to update the client grant space info.
3463 Frequency : when client reconnect to OST
3465 Description: Grant Leak when osc do resend and replay bulk write
3466 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3467 bulk write request, because the grant info while be sync between
3468 OSC and OST when reconnect, and we should ignore the grant info
3469 these of resend/replay write req.
3474 Description: Grant space more than available left space sometimes.
3475 Details : When then OST is about to be full, if two bulk writing from
3476 different clients came to OST. Accord the avaliable space of the
3477 OST, the first req should be permitted, and the second one
3478 should be denied by ENOSPC. But if the seconde arrived before
3479 the first one is committed. The OST might wrongly permit second
3480 writing, which will cause grant space > available space.
3483 Frequency : when client is evicted
3485 Description: Grant might be wrongly erased when osc is evicted by OST
3486 Details : when the import is evicted by server, it will fork another
3487 thread ptlrpc_invalidate_import_thread to invalidate the
3488 import, where the grant will be set to 0. While the original
3489 thread will update the grant it got when connecting. So if
3490 the former happened latter, the grant will be wrongly errased
3491 because of this race.
3496 Description: Checking Stale with correct fid
3497 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3498 to check whether it is stale, because sometimes, we want the
3499 enqueue happened anyway, and op_data.fid2 will not be initialized.
3501 Severity : enhancement
3503 Description: update patchless client
3504 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3507 Frequency : only with 2.4 kernel
3509 Description: random memory corruption
3510 Details : size of struct ll_inode_info is to big for union inode.u and this
3511 can be cause of random memory corruption.
3516 Description: Memory leak in recovery
3517 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3518 It should also check obd_fail before fsfilt_start, otherwise if
3519 fsfilt_start return -EROFS,(failover mds during mds recovery).
3520 then the req will return with repmsg->transno = 0 and rc = EROFS.
3521 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3522 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3523 be freed no matter whether fsfilt_commit success or not.
3526 Frequency : only with huge count clients
3528 Description: Prevents from taking the superblock lock in llap_from_page for
3530 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3531 for llap_from_page call in ll_removepage prevents from taking the
3532 superblock lock for a soon died page.
3537 Description: Not check open intent error before release open handle
3538 Details : in some rare cases, the open intent error is not checked before
3539 release open handle, which may cause
3540 ASSERTION(open_req->rq_transno != 0), because it tries to release
3541 the failed open handle.
3546 Description: Set cat log bitmap only after create log success.
3547 Details : in some rare cases, the cat log bitmap is set too early. and it
3548 should be set only after create log success.
3552 Description: Accessing a block bevice can re-enable I/O when Lustre is
3553 tearing down a device.
3554 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3559 Description: service threads may hog cpus when there are a lot of requests
3560 Details : Insert cond_resched to give other threads a chance to use some CPU
3565 Description: the cat log was not initialized in recovery
3566 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3567 unlink log on mds will not be initialized until mds post
3568 recovery. And also in mds post recovery, the unlink log will
3569 initialization will be done asynchronausly, so there will be race
3570 between add unlink log and unlink log initialization.
3574 Description: brw_stats were being printed incorrectly
3575 Details : brw_stats were being printed as log2 but all of them were not
3576 recorded as log2. Also remove some code duplication arising from
3577 filter_tally_{read,write}.
3581 Frequency : rare, only in recovery.
3582 Description: ASSERTION(req->rq_type != LI_POISON) failed
3583 Details : imp_lock should be held while iterating over imp_sending_list for
3584 prevent destroy request after get timeout in ptlrpc_queue_wait.
3588 Description: replay-single.sh test 52 fails
3589 Details : A lock's skiplist need to be cleanup when it being unlinked
3590 from its resource list.
3594 Description: Short directio read returns full requested size rather than
3596 Details : Direct I/O operations should return actual amount of bytes
3597 transferred rather than requested size.
3599 Severity : enhancement
3601 Description: metadata RPC reduction (e.g. for rm performance)
3602 Details : decrease the amount of synchronous RPC between clients and servers
3603 by canceling conflicing lock before the operation on the client
3604 and packing thier handles into the main operation RPC to server.
3606 Severity : enhancement
3608 Description: Async OSC create to avoid the blocking unnecessarily.
3609 Details : If an OST has no remaining object, system will block on the
3610 creation when it needs to create a new object on this OST. Now,
3611 ways use pre-created objects when available, instead of blocking on
3612 an empty osc while others are not empty. If we must block, we
3613 block for the shortest possible period of time.
3617 Description: improve handling recoverable errors
3618 Details : if request processig with error which can be recoverable on server
3619 request should be resend, otherwise page released from cache and
3622 Severity : enhancement
3624 Description: refine locking for avoid write wrong info into lov_objid file
3625 Details : fix possible races with add new target and write/update data in
3628 --------------------------------------------------------------------------------
3630 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3635 Frequency : on some architectures
3637 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3638 Details : Uninitialized flags sometimes cause configuration commands to
3642 Frequency : patchless clients only
3644 Description: missing __iget() symbol export
3645 Details : The __iget() symbol export is missing. To avoid the need for
3646 this on patchless clients the deathrow inode reaper is turned
3647 off, and we depend on the VM to clean up old inodes. This
3648 dependency was during via the fix for bug 12181.
3650 --------------------------------------------------------------------------------
3652 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3654 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3655 INTEROPERATE with older versions automatically. In many cases a
3656 special upgrade step is needed. Please read the
3657 user documentation before upgrading any part of a live system.
3658 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3659 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3660 * WARNING: Lustre configuration and startup changes are required with
3661 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3665 Severity : enhancement
3667 Description: MountConf
3668 Details : Lustre configuration is now managed via mkfs and mount
3669 commands instead of lmc and lconf. New obd types (MGS, MGC)
3670 are added for dynamic configuration management. See
3671 https://mail.clusterfs.com/wikis/lustre/MountConf for
3674 Severity : enhancement
3676 Description: dynamic OST addition
3677 Details : OSTs can now be added to a live filesystem
3679 Severity : enhancement
3681 Description: startup order invariance
3682 Details : MDTs and OSTs can be started in any order. Clients only
3683 require the MDT to complete startup.
3685 Severity : enhancement
3687 Description: parallel, asynchronous orphan cleanup
3688 Details : orphan cleanup is now performed in separate threads for each
3689 OST, allowing parallel non-blocking operation.
3691 Severity : enhancement
3693 Description: optimized stripe assignment
3694 Details : stripe assignments are now made based on ost space available,
3695 ost previous usage, and OSS previous usage, in order to try
3696 to optimize storage space and networking resources.
3698 Severity : enhancement
3700 Description: Permanently set tunables
3701 Details : All writable /proc/fs/lustre tunables can now be permanently
3702 set on a per-server basis, at mkfs time or on a live system.
3704 Severity : enhancement
3706 Description: Lustre message v2
3707 Details : Add lustre message format v2.
3709 Severity : enhancement
3711 Description: client OST exclusion list
3712 Details : Clients can be started with a list of OSTs that should be
3713 declared "inactive" for known non-responsive OSTs.
3717 Description: SPEC SFS validation failure on NFS v2 over lustre.
3718 Details : Changes the blocksize for regular files to be 2x RPC size,
3719 and not depend on stripe size.
3721 Severity : enhancement
3723 Description: Multiple MD RPCs in flight.
3724 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3725 To avoid overly-overloading MDS, introduce a limit on number of
3726 MDS RPCs in flight for a single client and add /proc controls
3727 to adjust this limit.
3729 Severity : enhancement
3731 Description: client read/write statistics
3732 Details : Add client read/write call usage stats for performance
3733 analysis of user processes.
3734 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3735 file access. extents_stats shows chunk size distribution.
3736 extents_stats_per_process show chunk size distribution per
3739 Severity : enhancement
3741 Description: mds statistics
3742 Details : Add detailed mds operations statistics in
3743 /proc/fs/lustre/mds/*/stats.
3747 Description: Failure of copying files with lustre special EAs.
3748 Details : Client side always return success for setxattr call for lustre
3749 special xattr (currently only "trusted.lov").
3754 Description: Refcount LNET uuids
3755 Details : The global LNET uuid list grew linearly with every startup;
3756 refcount repeated list entries instead of always adding to
3759 Severity : enhancement
3761 Description: Dynamic service threads
3762 Details : Within a small range, start extra service threads
3763 automatically when the request queue builds up.
3766 Frequency : mixed-endian client/server environments
3768 Description: mixed-endian crashes
3769 Details : The new msg_v2 system had some failures in mixed-endian
3772 Severity : enhancement
3774 Description: Easy OST removal
3775 Details : OSTs can be permanently deactivated with e.g. 'lctl
3776 conf_param lustre-OST0001.osc.active=0'
3778 Severity : enhancement
3780 Description: MGS proc entries
3781 Details : Added basic proc entries for the MGS showing what filesystems
3784 Severity : enhancement
3786 Description: provide MGS failover
3787 Details : Added config lock reacquisition after MGS server failover.
3789 Severity : enhancement
3791 Description: add Linux 2.4 support
3792 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3796 Description: a large application tries to do I/O to the same resource and dies
3797 in the middle of it.
3798 Details : Check the req->rq_arrival time after the call to
3799 ost_brw_lock_get(), but before we do anything about
3800 processing it & sending the BULK transfer request. This
3801 should help move old stale pending locks off the queue as
3802 quickly as obd_timeout.
3805 Frequency : when an incorrect nid is specified during startup
3807 Description: ptlrpc connect to non-existant node causes kernel crash
3808 Details : LNET can't be re-entered from an event callback, which
3809 happened when we expire a message after the export has been
3810 cleaned up. Instead, hand the zombie cleanup off to another
3813 Severity : enhancement
3815 Description: plain/inodebits lock performance improvement
3816 Details : Grouping plain/inodebits in granted list by their request modes
3817 and bits policy, thus improving the performance of search through
3821 Frequency : only if OST filesystem is corrupted
3823 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3824 Details : for a short time RPCs with bulk IO are in the replay list,
3825 but replay of bulk IOs is unimplemented. If the OST filesystem
3826 is corrupted due to disk cache incoherency and then replay is
3827 started it is possible to trip an assertion. Avoid putting
3828 committed RPCs into the replay list at all to avoid this issue.
3831 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3833 Description: System hang on startup
3834 Details : This bug allowed the liblustre (e.g. catamount) client to
3835 return to the app before handling all startup RPCs. This
3836 could leave the node unresponsive to lustre network traffic
3837 and manifested as a server ptllnd timeout.
3839 Severity : enhancement
3841 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3842 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3843 boolean to control whether to print peer debug info when a
3844 client's RPC times out.
3847 Frequency : only for kernels with patches from Lustre below 1.4.3
3849 Description: Remove old rdonly API
3850 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3853 Frequency : only for devices with external journals
3855 Description: Set external device read-only also
3856 Details : During a commanded failover stop, we set the disk device
3857 read-only while the server shuts down. We now also set any
3858 external journal device read-only at the same time.
3861 Frequency : when upgrading from 1.4 while trying to change parameters
3863 Description: The wrong (new) MDC name was used when setting parameters for
3864 upgraded MDT's. Also allows changing of OSC (and MDC)
3865 parameters if --writeconf is specified at tunefs upgrade time.
3868 Frequency : when setting specific ost indicies
3870 Description: QOS code breaks on skipped indicies
3871 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3872 created with --index need not be sequential.
3874 Severity : enhancement
3876 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3877 Details : The uninit_groups feature works in conjunction with the kernel
3878 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3879 up the pass1 processing of e2fsck. This is a read-only feature
3880 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3881 mount filesystems that have had this feature enabled.
3883 Severity : enhancement
3885 Description: Improve multi-block allocation algorithm to avoid fragmentation
3886 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3887 allocation locality and avoid filesystem fragmentation.
3889 ------------------------------------------------------------------------------
3891 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3893 * Support for kernels:
3894 2.4.21-47.0.1.EL (RHEL 3)
3895 2.6.5-7.283 (SLES 9)
3896 2.6.9-42.0.10.EL (RHEL 4)
3897 2.6.12.6 vanilla (kernel.org)
3898 2.6.16.27-0.9 (SLES 10)
3899 * Recommended e2fsprogs version: 1.39.cfs5
3901 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3905 Frequency : occasional, depends on client load and configuration
3906 Bugzilla : 12181, 12203
3907 Description: data loss for recently-modified files
3909 Details : In some cases it is possible that recently written or created
3910 files may not be written to disk in a timely manner (this should
3911 normally be within 30s unless client IO load is very high).
3912 The problem appears as zero-length files or files that are a
3913 multiple of 1MB in size after a client crash or client eviction
3914 that are missing data at the end of the file.
3916 This problem is more likely to be hit on clients where files are
3917 repeatedly created and unlinked in the same directory, clients
3918 have a large amount of RAM, have many CPUs, the filesystem has
3919 many OSTs, the clients are rebooted frequently, and/or the files
3920 are not accessed by other nodes after being written.
3922 The presence of the problem can be detected by looking at
3923 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3924 smaller than the second (nr_unused) then dirty files will not
3925 be flushed automatically to disk. "sync; sleep 10" should be
3926 run several times on the node before unmounting it to update
3927 Lustre (this is also safe to run on nodes without this problem).
3929 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3930 that can cause this same problem, so customers using that kernel
3931 also need to update the kernel in addition to Lustre. In order
3932 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3934 It is normal that files written just before a client crash (less
3935 than 30s) may not yet have been flushed to disk, even for local
3939 Frequency : frequent on thin XT3 nodes
3941 Description: UUID collision on thin XT3 Linux nodes
3942 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3943 randomly, since we relied on an insufficiently-seeded PRNG.
3948 Description: OSS hangs after "All ost request buffers busy"
3949 Details : A deadlock between quota and journal operations caused OSS
3950 hangs after printing "All ost request buffers busy."
3953 Frequency : always on liblustre builds
3955 Description: Cleanup compiler warnings on liblustre
3958 Frequency : always on liblustre builds on XT3
3960 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3964 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3965 > 32000 subdirectories
3966 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3967 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3968 subdirectory count crosses 32000. This will aid e2fsck to
3969 correctly handle more than 32000 subdirectories.
3972 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3974 Description: System hang on startup
3975 Details : This bug allowed the liblustre (e.g. catamount) client to
3976 return to the app before handling all startup RPCs. This
3977 could leave the node unresponsive to lustre network traffic
3978 and manifested as a server ptllnd timeout.
3980 Severity : enhancement
3982 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3983 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3984 boolean to control whether to print peer debug info when a
3985 client's RPC times out.
3990 Description: make O_SYNC working on 2.6 kernels
3991 Details : 2.6 kernels use different method for mark pages for write,
3992 so need add a code to lustre for O_SYNC work.
3997 Description: Failure to close file and release space on NFS
3998 Details : Put inode details into lock acquired in ll_intent_file_open.
3999 Use mdc_intent_lock in ll_intent_open to properly
4000 detect all kind of errors unhandled by mdc_enqueue.
4005 Description: proc file read during shutdown sometimes raced obd removal,
4007 Details : Add lock to prevent obd access after proc file removal.
4010 Frequency : Only for files larger than 4GB on 32-bit clients.
4012 Description: improperly doing page alignment of locks
4013 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
4014 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
4019 Description: RPCs being resent when they shouldn't be
4020 Details : Some RPCs that should not be resent are being resent. This
4021 can cause inconsistencies in the RPC state machine. Do not
4022 resend such requests.
4025 Frequency : rare, only with NFS export
4027 Description: Crash on NFS re-export node
4028 Details : under very unusual load conditions an assertion is hit in
4029 ll_intent_file_open()
4032 Frequency : only if OST filesystem is corrupted
4034 Description: client incorrectly hits assertion in ptlrpc_replay_req()
4035 Details : for a short time RPCs with bulk IO are in the replay list,
4036 but replay of bulk IOs is unimplemented. If the OST filesystem
4037 is corrupted due to disk cache incoherency and then replay is
4038 started it is possible to trip an assertion. Avoid putting
4039 committed RPCs into the replay list at all to avoid this issue.
4044 Description: large O_DIRECT requests fail under memory pressure/fragmentation
4045 Details : Large single O_DIRECT read and write calls can fail to allocate
4046 a sufficiently large buffer to process the request. In case of
4047 allocation failure the allocation is retried with a smaller
4048 buffer and broken into smaller requests.
4050 Severity : enhancement
4052 Description: Add -o localflock option to simulate old noflock behaviour.
4053 Details : This will achieve local-only flock/fcntl locks coherentness.
4058 Description: versioning check is incomplete
4059 Details : Checking the version difference of client vs. server, report
4060 error if the gap is too big.
4065 Description: add support PG_writeback bit
4066 Details : add support for PG_writeback bit for Lustre, for more carefull
4067 work with page cache in 2.6 kernel. This also fix some deadlocks
4068 and remove hack for work O_SYNC with 2.6 kernel.
4070 Severity : enhancement
4072 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
4073 Details : The uninit_groups feature works in conjunction with the kernel
4074 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
4075 up the pass1 processing of e2fsck. This is a read-only feature
4076 in ldiskfs2 only, so older kernels and current ldiskfs cannot
4077 mount filesystems that have had this feature enabled.
4079 Severity : enhancement
4081 Description: Improve multi-block allocation algorithm to avoid fragmentation
4082 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
4083 allocation locality and avoid filesystem fragmentation.
4085 ------------------------------------------------------------------------------
4087 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
4089 * Support for kernels:
4090 2.6.9-42.0.3.EL (RHEL 4)
4091 2.6.5-7.276 (SLES 9)
4092 2.4.21-47.0.1.EL (RHEL 3)
4093 2.6.12.6 vanilla (kernel.org)
4094 2.6.16.21-0.8 (SLES10)
4095 * Recommended e2fsprogs version: 1.39.cfs2-0
4097 * The backwards-compatible /proc/sys/portals symlink has been removed
4098 in this release. Before upgrading, please ensure that you change
4099 any configuration scripts or /etc/sysctl.conf files that access
4100 /proc/sys/portals/* or sysctl portals.* to use the corresponding
4101 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
4102 in advance of the upgrade on any system running Lustre 1.4.6 or
4103 newer, since /proc/sys/lnet was added in that version.
4104 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
4108 Frequency : only when quota is used
4110 Description: avoid scanning export list for quota master
4111 Details : Change the algorithms to avoid scanning export list in order
4112 to improve the efficiency.
4115 Frequency : MDS failover only, very rarely
4117 Description: "went back in time" messages on mds failover
4118 Details : The greatest transno may be lost when the current operation
4119 finishes with an error (transno==0) and the client's last_rcvd
4120 record is over-written. Save the greatest transno in the
4121 mds_last_transno for this case.
4124 Frequency : always for specific kernels and striping counts
4126 Description: client may get "Matching packet too big" without ACL support
4127 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
4128 when trying to access files in certain configurations. The
4129 clients should in fact be denied when mounting because they do
4130 not understand ACLs.
4133 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4135 Description: many clients connecting with IO in progress causes connect timeouts
4136 Details : Avoid synchronous journal commits to avoid delays caused by many
4137 clients connecting/disconnecting when bulk IO is in progress.
4138 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4139 OST_IO_PORTAL to avoid delays behind potentially many pending
4143 Frequency : occasionally with multiple writers to a single file
4145 Description: shared writes to file may result in wrong size reported by stat()
4146 Details : Allow growing of kms when extent lock is cancelled
4149 Frequency : always with random mmap IO to multi-striped file
4151 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4152 Details : Only if the hole is at the end of OST object so that kms is too
4153 small. Fix is to increase kms accordingly in ll_nopage.
4156 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4158 Description: writes to a missing object would leak memory on the OST
4159 Details : If there is an inconsistency between the MDS and OST filesystems,
4160 such that the MDS references an object that doesn't exist, writes
4161 to that object will leak memory due to incorrect cleanup in the
4162 error handling path, eventually running out of memory on the OST.
4167 Description: Creating too long symlink causes lustre errors
4168 Details : Check symlink and name lengths before sending requests to MDS.
4171 Frequency : only if flock is enabled (not on by default)
4173 Description: posix locks not released on fd closure on 2.6.9+
4174 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4175 kernels, this caused such locks not to be released on fd close and
4176 then assertions on fs unmount about still used locks.
4179 Frequency : MDS failover only, very rarely
4181 Description: clients may get ASSERTION(granted_lock != NULL)
4182 Details : When request was taking a long time, and a client was resending
4183 a getattr by name lock request. The were multiple lock requests
4184 with the same client lock handle and
4185 mds_getattr_name->fixup_handle_for_resent_request found one of the
4186 lock handles but later failed with ASSERTION(granted_lock != NULL).
4191 Description: handle->h_buffer_credits > 0, assertion failure
4192 Details : h_buffer_credits is zero after truncate, causing assertion
4193 failure. This patch extends the transaction or creates a new
4197 Frequency : NFS re-export or patchless client
4198 Bugzilla : 11179, 10796
4199 Description: Crash on NFS re-export node (__d_move)
4200 Details : We do not want to hash the dentry if we don't have a lock.
4201 But if this dentry is later used in d_move, we'd hit uninitialised
4202 list head d_hash, so we just do this to init d_hash field but
4203 leave dentry unhashed.
4206 Frequency : NFS re-export or patchless client
4208 Description: NFS exports has problem with symbolic link
4209 Details : lustre client didn't properly install dentry when re-exported
4210 to NFS or running patchless client.
4213 Frequency : NFS re-export or patchless client
4215 Description: Various nfs/patchless fixes.
4216 Details : fixes reuse disconected alias for lookup process - this fixes
4217 warning "find_exported_dentry: npd != pd",
4218 fix permission error with open files at nfs.
4219 fix apply umask when do revalidate.
4222 Frequency : occasional
4224 Description: Crash on NFS re-export node
4225 Details : calling clear_page() on the wrong pointer triggered oops in
4226 generic_mapping_read().
4229 Frequency : rarely, using O_DIRECT IO
4231 Description: unaligned directio crashes client with LASSERT
4232 Details : check for unaligned buffers before trying any requests.
4235 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4237 Description: stale data returned from RAID cache
4238 Details : If only a small amount of IO is done to the RAID device before
4239 reading it again it is possible to get stale data from the RAID
4240 cache instead of reading it from disk.
4243 Frequency : always for sles10 kernel
4245 Description: sles10 support
4246 Details : ll_follow_link: compile fixes and using of nd_set_link
4247 under newer kernels.
4250 Frequency : depends on arch, kernel and compiler version, always on sles10
4253 Description: recursive or deep enough symlinks cause stack overflow
4254 Details : getting rid of large stack-allocated variable in
4258 Frequency : depends on hardware
4260 Description: lustre write performance loss in the SLES10 kernel
4261 Details : the performance loss is caused by using of write barriers in the
4262 ext3 code. The SLES10 kernel turns barrier support on by
4263 default. The fix is to undo that change for ldiskfs.
4265 ------------------------------------------------------------------------------
4267 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4269 * Support for kernels:
4270 2.6.9-42.0.3EL (RHEL 4)
4271 2.6.5-7.276 (SLES 9)
4272 2.4.21-47.0.1.EL (RHEL 3)
4273 2.6.12.6 vanilla (kernel.org)
4277 Frequency : quota enabled and large files being deleted
4279 Description: releasing more than 4GB of quota at once hangs OST
4280 Details : If a user deletes more than 4GB of files on a single OST it
4281 will cause the OST to spin in an infinite loop. Release
4282 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4287 Description: statfs data retrieved from /proc may be stale or zero
4288 Details : When reading per-device statfs data from /proc, in the
4289 {kbytes,files}_{total,free,avail} files, it may appear
4290 as zero or be out of date.
4293 Frequency : systems with MD RAID1 external journal devices
4295 Description: lconf's call to blkid is confused by RAID1 journal devices
4296 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4297 of returning all block devices that match the journal UUID.
4300 Frequency : always, for aggregate stripe size over 4GB
4302 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4303 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4304 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4307 Frequency : always if "lfs find" used on a local file/directory
4309 Description: "lfs find" segfaults if used on a local file/directory
4310 Details : The case where a directory component was not specified wasn't
4311 handled correctly. Handle this properly.
4314 Frequency : always on ppc64
4316 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4317 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4318 perform endianness conversion.
4321 Frequency : rarely (truncate to non-zero file size after write under load)
4322 Bugzilla : 10730, 10687
4323 Description: Files padded with zeros to next 4K multiple
4324 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4325 it is possible that files that are truncated to a non-zero size
4326 immediately after being written are filled with zero bytes beyond
4327 the truncated size. No file data is lost.
4329 Severity : enhancement
4331 Description: Allow recovery/failover for liblustre clients.
4332 Details : liblustre clients were unaware of failover configurations until
4335 Severity : enhancement
4337 Description: user file locks should fail when not mounting with flock option
4338 Details : Set up an error-returning stub in ll_file_operations.lock field
4339 to prevent incorrect behaviour when client is mounted without
4340 flock option. Also, set up properly f_op->flock field for
4344 Frequency : always on ia64
4346 Description: "lfs df" loops on printing out MDS statfs information
4347 Details : The obd_ioctl_data was not initialized and in some systems
4348 this caused a failure during the ioctl that did not return
4349 an error. Initialize the struct and return an error on failure.
4352 Frequency : SLES 9 only
4354 Description: Error of copying files with lustre special EAs as root
4355 Details : Client side always return success for setxattr call for lustre
4356 special xattr (currently only "trusted.lov").
4359 Frequency : rarely on clusters with both ia64+i386 clients
4361 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4362 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4363 clients are doing concurrent writes to the same file it is
4364 possible that the ia64 clients may LASSERT because the OST
4365 extent locks are not PAGE_SIZE aligned. Ensure that grown
4366 locks are always aligned on the request boundary.
4369 Frequency : specific use, occasional
4371 Description: Overwriting in use executable truncates on-disk binary image
4372 Details : If one node attempts to overwrite an executable in use by
4373 another node, we now correctly return ETXTBSY instead of
4374 truncating the file.
4376 Severity : enhancement
4378 Description: Async OSC create to avoid the blocking unnecessarily.
4379 Details : If an OST has no remaining object, system will block on the creating
4380 when need to create a new object on this OST. Now, ways use
4381 pre-created objects when available, instead of blocking on an
4382 empty osc while others are not empty. If we must block, we block
4383 for the shortest possible period of time.
4388 Description: chmod on Lustre root is propagated to other clients
4389 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4390 invalid by the follow_mount time.
4393 Frequency : liblustre clients only
4395 Description: Race in 'instant cancel' lock handling could lead to such locks
4396 never to be granted in case of SMP MDS
4397 Details : Do not destroy not yet granted but cbpending locks in
4401 Frequency : replay/resend of open
4403 Description: non null lock assetion failure in mds_intent_policy
4404 Details : Trying to replay/resend lockless open requests resulted in
4405 mds_open() returning 0 with no lock. Now it sets a flag if
4406 a lock is going to be returned.
4408 Severity : enhancement
4410 Description: Checksum enhancements
4411 Details : New checksum enhancements allow for resending RPCs that failed
4414 Severity : enhancement
4416 Description: Tunables on number of dirty pages in cacche
4417 Details : Allow to set limit on number of dirty pages cached.
4422 Description: client crash on unmount - lock still has references
4423 Details : In some error handling cases it was possible to leak a lock
4424 reference on a client while accessing a file. This was not
4425 harmful to the client during operation, but would cause the
4426 client to crash when the filesystem is unmounted.
4429 Frequency : specific case, rare
4431 Description: ETXTBSY on mds though file not in use by client
4432 Details : ETXTBSY is no longer incorrectly returned when attempting to
4433 chmod or chown a directory that the user previously tried to
4434 execute or a currently-executing binary.
4437 Frequency : extremely rare except on liblustre-based clients
4439 Description: Lustre space not freed when files are deleted
4440 Details : Clean up open-unlinked files after client eviction. Previously
4441 the unlink was skipped and the files remained as orphans.
4446 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4447 Details : In some cases it was possible to send a blocking callback to
4448 a client doing a glimpse, even though that client didn't get
4449 a lock granted. When the glimpse lock is cancelled on the OST
4450 the freed lock is left on the waiting list and corrupted the list.
4453 Frequency : all core dumps
4455 Description: Broke core dumps to lustre
4456 Details : Negative dentry may be unhashed if parent does not have UPDATE
4457 lock, but some callers, e.g. do_coredump, expect dentry to be
4458 hashed after successful create, hash it in ll_create_it.
4460 ------------------------------------------------------------------------------
4462 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4464 * Support for kernels:
4465 2.6.9-42.0.2.EL (RHEL 4)
4466 2.6.5-7.276 (SLES 9)
4467 2.4.21-40.EL (RHEL 3)
4468 2.6.12.6 vanilla (kernel.org)
4472 Frequency : always on RHEL 3
4474 Description: Number of open files grows over time
4475 Details : The number of open files grows over time, whether or not
4476 Lustre is started. This was due to a filp leak introduced
4477 by one of our kernel patches.
4479 ------------------------------------------------------------------------------
4481 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4483 * Support for kernels:
4484 2.6.9-42.EL (RHEL 4)
4485 2.6.5-7.276 (SLES 9)
4486 2.4.21-40.EL (RHEL 3)
4487 2.6.12.6 vanilla (kernel.org)
4492 Bugzilla : 5719, 9635, 9792, 9684
4493 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4494 Details : If a server is under heavy load and cannot reply to new
4495 connection requests before the client resends the (re)connect,
4496 the connection handling code can behave badly if two service
4497 threads are concurrently handing separate (re)connections from
4498 the same client. Add better locking to the connection handling
4499 code, and ensure that only a single connection will be processed
4500 for a given client UUID, even if the lock is dropped.
4502 Severity : enhancement
4504 Description: add TCP zero-copy support to kernel
4505 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4506 sends if the hardware supports scatter-gather and checksumming.
4507 This allows socklnd to do client-write and server-read more
4508 efficiently and reduce CPU utilization from skbuf copying.
4511 Frequency : only if NFS exporting from client
4513 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4514 Details : If setting attributes on a file created under NFS that had
4515 never been opened it would be possible to oops the client
4516 if the file had no objects.
4519 Frequency : always for liblustre
4521 Description: liblustre client does MDS+OSTs setattr RPC for each write
4522 Details : When doing a write from a liblustre client, the client
4523 incorrectly issued an RPC to the MDS and each OST the file was
4524 striped over in order to update the timestamps. When writing
4525 with small chunks and many clients this could overwhelm the MDS
4526 with RPCs. In all cases it would slow down the write because
4527 these RPCs are unnecessary.
4529 Severity : enhancement
4531 Description: allow number of MDS service threads to be changed at module load
4532 Details : It is now possible to change the number of MDS service threads
4533 running. Adding "options mds mds_num_threads={N}" to the MDS's
4534 /etc/modprobe.conf will set the number of threads for the next
4535 time Lustre is restarted (assuming the "mds" module is also
4536 reloaded at that time). The default number of threads will
4537 stay the same, 32 for most systems.
4542 Description: OST crash if filesystem is unformatted or corrupt
4543 Details : If an OST is started on a device that has never been formatted
4544 or if the filesystem is corrupt and cannot even mount then the
4545 error handling cleanup routines would dereference a NULL pointer.
4550 Description: NULL pointer deref in llap_from_page.
4551 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4552 incorrect) mapping if the page was truncated/reclaimed while it was
4553 searched for. Check for this condition and skip such pages when
4554 doing readahead. Introduce extra check to llap_from_page() to
4555 verify page->mapping->host is non-NULL (so page is not anonymous).
4558 Frequency : Sometimes when using sys_sendfile
4560 Description: "page not covered by a lock" warnings from ll_readpage
4561 Details : sendfile called ll_readpage without right page locks present.
4562 Now we introduced ll_file_sendfile that does necessary locking
4563 around call to generic_file_sendfile() much like we do in
4567 Frequency : with certain MDS communication failures at client mount time
4569 Description: NULL pointer deref after failed client mount
4570 Details : a client connection request may delayed by the network layer
4571 and not be sent until after the PTLRPC layer has timed out the
4572 request. If the client fails the mount immediately it will try
4573 to clean up before the network times out the request. Add a
4574 reference from the request import to the obd device and delay
4575 the cleanup until the network drops the request.
4578 Frequency : occasionally during client (re)connect
4580 Description: assertion failure during client (re)connect
4581 Details : processing a client connection request may be delayed by the
4582 client or server longer than the client connect timeout. This
4583 causes the client to resend the connection request. If the
4584 original connection request is replied in this interval, the
4585 client may trip an assertion failure in ptlrpc_connect_interpret()
4586 which thought it would be the only running connect process.
4589 Frequency : only with obd_echo servers and clients that are rebooted
4591 Description: kernel BUG accessing uninitialized data structure
4592 Details : When running an obd_echo server it did not start the ping_evictor
4593 thread, and when a client was evicted an uninitialized data
4594 structure was accessed. Start the ping_evictor in the RPC
4595 service startup instead of the OBD startup.
4597 Severity : enhancement
4598 Bugzilla : 10193 (patchless)
4599 Description: Remove dependency on various unexported kernel interfaces.
4600 Details : No longer need reparent_to_init, exit_mm, exit_files,
4601 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4604 Frequency : rare (only users of deprecated and unsupported LDAP config)
4606 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4607 Details : LDAP apparently contains 'lustreName' attributes instead of
4608 'name'. A simple remapping of the name is sufficient.
4611 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4613 Description: waiting_locks_callback trips kernel BUG if client is evicted
4614 Details : Running with the dump_on_timeout debug flag turned on makes
4615 it possible that the waiting_locks_callback() can try to dump
4616 the Lustre kernel debug logs from an interrupt handler. Defer
4617 this log dumping to the expired_lock_main() thread.
4619 Severity : enhancement
4621 Description: Support NFS exporting on 2.6 kernels.
4622 Details : Implement non-rawops metadata methods for NFS server to use without
4623 changing NFS server code.
4626 Frequency : very rare (synthetic metadata workload only)
4628 Description: two racing renames might cause an MDS thread to deadlock
4629 Details : Running the "racer" program may cause one MDS thread to rename
4630 a file from being the source of a rename to being the target of
4631 a rename at exactly the same time that another thread is doing
4632 so, and the second thread has already enqueued these locks after
4633 doing a lookup of the target and is trying to relock them in
4634 order. Ensure that we don't try to re-lock the same resource.
4637 Frequency : only very large systems with liblustre clients
4639 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4640 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4641 all OSTs in parallel. This allows the eviction of stale liblustre
4642 clients to proceed much faster than if they were done in series,
4643 and also offers similar improvements for other set_info RPCs.
4648 Description: excessive CPU usage during initial read phase on client
4649 Details : During the initial read phase on a client, it would agressively
4650 retry readahead on the file, consuming too much CPU and impacting
4651 performance (since 1.4.5.8). Improve the readahead algorithm
4652 to avoid this, and also improve some other common cases (read
4653 of small files in particular, where "small" is files smaller than
4654 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4659 Description: MDS crash when receiving packet with unknown intent.
4660 Details : Do not LBUG in unknown intent case, just return -EFAULT
4662 Severity : enhancement
4663 Bugzilla : 9293, 9385
4664 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4665 Details : Do not serialize getattr (non-intent version) and statfs.
4668 Frequency : occasional, when OST network is overloaded/intermittent
4670 Description: client evicted by OST after bulk IO timeout
4671 Details : If a client sends a bulk IO request (read or write) the OST
4672 may evict the client if it is unresposive to its data GET/PUT
4673 request. This is incorrect if the network is overloaded (takes
4674 too long to transfer the RPC data) or dropped the OST GET/PUT
4675 request. There is no need to evict the client at all, since
4676 the pinger and/or lock callbacks will handle this, and the
4677 client can restart the bulk request.
4680 Frequency : Always when mmapping file with no objects
4682 Description: client crashes when mmapping file with no objects
4683 Details : Check that we actually have objects in a file before doing any
4684 operations on objects in ll_vm_open, ll_vm_close and
4690 Description: Request leak when working with deleted CWD
4691 Details : Introduce advanced request refcount tracking for requests
4692 referenced from lustre intent.
4694 Severity : Enhancement
4696 Description: Cache open file handles on client.
4697 Details : MDS now will return special lock along with openhandle, if
4698 requested and client is allowed to hold openhandle, even if unused,
4699 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4700 closing files for every read/write openration.
4702 Severity : Enhancement
4704 Description: Cache open negative dentries on client when possible.
4705 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4706 negative dentries on lock revocation.
4711 Description: Remounting a client read-only wasn't possible with a zconf mount
4712 Details : It wasn't possible to remount a client read-only with llmount.
4714 Severity : enhancement
4715 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4716 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4717 MPICH is included to add Lustre-specific ADIO interfaces.
4718 This is based closely on the UFS ADIO layer and only differs
4719 in file creation, in order to allow the OST striping to be set.
4720 This is user-contributed code and not supported by CFS.
4725 Description: extended inode attributes (immutable, append-only) work improperly
4726 when 2.4 and 2.6 kernels are used on client/server or vice versa
4727 Details : Introduce kernel-independent values for these flags.
4729 Severity : enhancement
4732 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4733 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4734 granularity. Now, Fractional MB granularity is be supported,
4735 this is very useful for low memory system.
4737 Severity : enhancement
4739 Description: Getattr by fid
4740 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4741 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4742 have locks and avoids getattr rpc requests for it on every stat.
4745 Frequency : Always, for filesystems larger than 2TB
4747 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4748 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4749 filesystems larger than 2TB to be mounted with mballoc enabled.
4752 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4754 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4755 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4756 filesystems larger than 2TB is mounted then the kernel will
4757 silently corrupt the start of the filesystem. CONFIG_LBD is
4758 enabled for all CFS-supported kernels, but the possibility of
4759 this happening with a modified kernel config exists.
4761 Severity : enhancement
4763 Description: add client O_DIRECT support for 2.6 kernels
4764 Details : It is now possible to do O_DIRECT reads and writes to files
4765 in the Lustre client mountpoint on 2.6 kernel clients.
4767 Severity : enhancement
4769 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4770 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4771 parallel, not waiting for response from every OST before sending
4772 a rpc to the next OST.
4777 Description: setattr vs write race when updating file timestamps
4778 Details : Client processes that update a file timestamp into the past
4779 right after writing to the file (e.g. tar) it is possible that
4780 the updated file modification time can be reset to the current
4781 time due to a race between processing the setattr and write RPC.
4783 Severity : enhancement
4785 Description: Bring 'lfs find' closer in line with regular Linux find.
4786 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4787 -print0 options and obtains all the needed info through the lustre
4790 Severity : enhancement
4792 Description: support up to 1024 configured devices on one node
4793 Details : change obd_dev array from statically allocated to dynamically
4794 allocated structs as they are first used to reduce memory usage
4799 Description: Flush dirty partially truncated pages during truncate
4800 Details : Immediatelly flush partially truncated pages in filter_setattr,
4801 this way we completely avoid having any pages in page cache on OST
4802 and can retire ugly workarounds during writes to flush such pages.
4807 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4808 Details : protect inode from truncation within vfs_unlink() context
4809 just take a reference before calling vfs_unlink() and release it
4810 when parent's i_sem is free.
4815 Description: last_id value checked outside lock on OST caused LASSERT failure
4816 Details : If there were multiple MDS->OST object precreate requests in
4817 flight, it was possible that the OST's last object id was checked
4818 outside a lock and incorrectly tripped an assertion. Move checks
4819 inside locks, and discard old precreate requests.
4822 Frequency : always, if extents are used on OSTs
4824 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4825 Details : OSTs using the extents format would not zero the high 16 bits of
4826 the index physical block number. This is not a problem for any
4827 OST filesystems smaller than 16TB, and no kernels support ext3
4828 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4829 new/modified files) and can be fixed for existing filesystems
4830 with e2fsprogs-1.39-cfs1.
4835 Description: import connection selection may be incorrect if timer wraps
4836 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4837 import connections to be ignored if the 32-bit jiffies counter
4838 wraps. Use a 64-bit jiffies counter.
4841 Frequency : very large clusters immediately after boot
4843 Description: LNET request buffers exhausted under heavy short-term load
4844 Details : If a large number of client requests are generated on a service
4845 that has previously never seen so many requests it is possible
4846 that the request buffer growth cannot keep up with the spike in
4847 demand. Instead of dropping incoming requests, they are held in
4848 the LND until the RPC service can accept more requests.
4851 Frequency : Sometimes during replay
4853 Description: Assertion failure in ll_local_open after replay.
4854 Details : If replay happened on an open request reply before we were able
4855 to set replay handler, reply will become not swabbed tripping the
4856 assertion in ll_local_open. Now we set the handler right after
4857 recognising of open request
4860 Frequency : very rare
4862 Description: kernel reports "badness in vsnprintf"
4863 Details : Reading from the "recovery_status" /proc file in small chunks
4864 may cause a negative length in lprocfs_obd_rd_recovery_status()
4865 call to vsnprintf() (which is otherwise harmless). Exit early
4866 if there is no more space in the output buffer.
4868 Severity : enhancement
4870 Description: clear OBD RPC statistics by writing to them
4871 Details : It is now possible to clear the OBD RPC statistics by writing
4872 to the "stats" file.
4877 Description: Client mtime is not the same on different clients after utimes
4878 Details : In some cases, the client was using the utimes() syscall on
4879 a file cached on another node. The clients now validate the
4880 ctime from the MDS + OSTs to determine which one is right.
4885 Description: Inability to activate failout mode
4886 Details : lconf script incorrectly assumed that in pythong string's numeric
4887 value is used in comparisons.
4890 Frequency : always with multiple stripes per file
4892 Description: Inefficient object allocation for mutli-stripe files
4893 Details : When selecting which OSTs to stripe files over, for files with
4894 a stripe count that divides evenly into the number of OSTs,
4895 the MDS is always picking the same starting OST for each file.
4896 Return the OST selection heuristic to the original design.
4901 Description: mount failures may take full timeout to return an error
4902 Details : Under some heavy load conditions it is possible that a
4903 failed mount can wait for the full obd_timeout interval,
4904 possibly several minutes, before reporting an error.
4905 Instead return an error as soon as the status is known.
4907 Frequency : quota enabled and large files being deleted
4909 Description: releasing more than 4GB of quota at once hangs OST
4910 Details : If a user deletes more than 4GB of files on a single OST it
4911 will cause the OST to spin in an infinite loop. Release
4912 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4917 Description: statfs data retrieved from /proc may be stale or zero
4918 Details : When reading per-device statfs data from /proc, in the
4919 {kbytes,files}_{total,free,avail} files, it may appear
4920 as zero or be out of date.
4923 Frequency : systems with MD RAID1 external journal devices
4925 Description: lconf's call to blkid is confused by RAID1 journal devices
4926 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4927 of returning all block devices that match the journal UUID.
4930 Frequency : always, for aggregate stripe size over 4GB
4932 Description: assertion fails when trying to use 4GB stripe size
4933 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4934 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4937 Frequency : always on ppc64
4939 Description: the first write on an ext3 filesystem with mballoc got stuck
4940 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4941 perform endianness conversion.
4943 ------------------------------------------------------------------------------
4945 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4947 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4948 INTEROPERATE with older versions automatically. Please read the
4949 user documentation before upgrading any part of a live system.
4950 * WARNING: Lustre networking configuration changes are required with
4951 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4954 * Support for kernels:
4955 2.6.9-22.0.2.EL (RHEL 4)
4956 2.6.5-7.244 (SLES 9)
4957 2.6.12.6 vanilla (kernel.org)
4960 Severity : enhancement
4961 Bugzilla : 7981/8208
4962 Description: Introduced Lustre Networking (LNET)
4963 Details : LNET is new networking infrastructure for Lustre, it includes
4964 a reorganized network configuration mode (see the user
4965 documentation for full details) as well as support for routing
4966 between different network fabrics. Lustre Networking Devices
4967 (LNDs) for the supported network fabrics have also been
4968 created for this new infrastructure.
4970 Severity : enhancement
4971 Description: Introduced Access control lists
4972 Details : clients can set ACLs on files and directories in order to have
4973 more fine-grained permissions than the standard Unix UGO+RWX.
4974 The MDS must be started with the "-o acl" mount option.
4976 Severity : enhancement
4977 Description: Introduced filesystem quotas
4978 Details : Administrators may now establish per-user quotas on the
4981 Severity : enhancement
4983 Description: Configuration change for the XT3
4984 The PTLLND is now used to run Lustre over Portals on the XT3
4985 The configure option(s) --with-cray-portals are no longer used.
4986 Rather --with-portals=<path-to-portals-includes> is used to
4987 enable building on the XT3. In addition to enable XT3 specific
4988 features the option --enable-cray-xt3 must be used.
4993 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4994 Details : A race between allocating a new llcd and re-getting the llcd_lock
4995 allowed another thread to grab newly-allocated llcd.
4997 Severity : enhancement
4999 Description: 2.6 OST async journal commit and locking fix to improve performance
5000 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
5001 2.6 kernels are now async as they already were in 2.4 kernels so
5002 that they can commit concurrently with the network bulk transfer.
5003 For block-allocated files the filter allocation semaphore is held
5004 to avoid filesystem fragmentation during allocation. BKL lock
5005 removed for 2.6 xattr operations where it is no longer needed.
5010 Description: lconf incorrectly determined whether two IP networks could talk
5011 Details : In some more complicated routing and multiple-network
5012 configurations, lconf will avoid trying to make a network
5013 connection to a disjoint part of the IP space. It was doing the
5014 math incorrectly for one set of cases.
5019 Description: Fix for potential infinite loop processing records in an llog.
5020 Details : If an llog record is corrupted/zeroed, it is possible to loop
5021 forever in llog_process(). Validate the llog record length
5022 and skip the remainder of the block on error.
5025 Frequency : occasional (liblustre only)
5027 Description: liblustre could not open files whose last component is a symlink
5028 Details : sysio_path_walk() would incorrectly pass the open intent to
5029 intermediate path components.
5032 Frequency : rare (liblustre only with non-standard tuning)
5033 Bugzilla : 7201 (7350)
5034 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
5035 Details : llu_lookup_finish_locks() tries to set lock data on a lock
5036 after it has been released, only do this for referenced locks
5038 Severity : enhancement
5040 Description: specifying an (invalid) directory default stripe_size of -1
5041 would reset the directory default striping
5042 Details : stripe_size -1 was used internally to signal directory stripe
5043 removal, now use "all default" to signal dir stripe removal
5044 as a directory striping of "all default" is not useful
5047 Frequency : common for large clusters running liblustre clients
5049 Description: doing an ls when liblustre clients are running is slow
5050 Details : sending a glimpse AST to a liblustre client waits for every AST
5051 to time out, as liblustre clients will not respond. Since they
5052 cannot cache data we refresh the OST lock LVB from disk instead.
5054 Severity : enhancement
5056 Description: doing an ls at the same time as file IO can be slow
5057 Details : enqueue and other "small" requests can be blocked behind many
5058 large IO requests. Create a new OST IO portal for non-IO
5059 requests so they can be processed faster.
5062 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
5064 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
5065 created file with mode 000
5066 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
5067 allow an ftruncate() as a non-root user to a file with mode 000.
5068 Lustre now respects this flag to disable mode checks when
5069 truncating a file owned by the user
5072 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
5075 Description: Revoking locks from clients that went dead or catatonic might take
5077 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
5078 cancellation of such locks instant on servers without waiting for
5079 any reply from clients. Clients drops these locks when cancel
5080 notification from server is received without replying.
5083 Frequency : liblustre-only, when liblustre client dies or becomes busy
5085 Description: Doing ls on Linux clients can take a long time with active
5087 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
5088 granting such locks to it in the first place if possible. Locks
5089 are taken by proxy on the OST during the read or write and
5090 dropped immediately afterward. Add connect flags handling, do
5091 not grant locks to liblustre clients for glimpse ASTs.
5093 Severity : enhancement
5095 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
5096 Details : Existing read-ahead algorithm is tuned for the case of streamlined
5097 sequential reads and behaves badly with applications doing random
5098 reads. Improve it by reading ahead at least read region, and
5099 avoiding excessive large RPC for small reads.
5101 Severity : enhancement
5103 Description: Creating more than 1000 files for a single job may cause a load
5104 imbalance on the OSTs if there are also a large number of OSTs.
5105 Details : qos_prep_create() uses an OST index reseed value that is an
5106 even multiple of the number of available OSTs so that if the
5107 reseed happens in the middle of the object allocation it will
5108 still utilize the OSTs as uniformly as possible.
5113 Description: OST or MDS may oops in ping_evictor_main()
5114 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
5115 but doesn't restart at beginning of obd_exports_timed list
5118 Severity : enhancement
5120 Description: improve by-nid export eviction on the MDS and OST
5121 Details : allow multiple exports with the same NID to be evicted at one
5122 time without re-searching the exports list.
5125 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
5127 Description: MDS may oops in groups_free()
5128 Details : in rare race conditions a newly allocated group_info struct is
5129 freed again, and this can be NULL. The 2.4 compatibility code
5130 for groups_free() checked for a NULL pointer, but 2.6 did not.
5133 Frequency : common for liblustre clients doing little filesystem IO
5134 Bugzilla : 9352, 7313
5135 Description: server may evict liblustre clients accessing contended locks
5136 Details : if a client is granted a lock or receives a completion AST
5137 with a blocking AST already set it would not reply to the AST
5138 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5141 Frequency : lfs setstripe, only systems with more than 160 OSTs
5143 Description: unable to set striping with a starting offset beyond OST 160
5144 Details : llapi_create_file() incorrectly limited the starting stripe
5145 index to the maximum single-file stripe count.
5148 Frequency : LDAP users only
5150 Description: lconf did not handle in-kernel recovery with LDAP properly
5151 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5153 Severity : enhancement
5155 Description: bind OST threads to NUMA nodes to improve performance
5156 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5157 node and do their allocations there to localize memory access
5159 Severity : enhancement
5161 Description: llmount can determine client NID directly from Myrinet (GM)
5162 Details : the client NID code from gmnalnid was moved directly into
5163 llmount, removing the need to use this or specifying the
5164 client NID explicitly when mounting GM clients with zeroconf
5167 Frequency : if client is started with down MDS
5169 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5170 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5171 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5172 ptlrpc_import_delayed_req() because we didn't check if the
5173 request was interrupted, and we also didn't break out of the
5174 event loop if there was no timeout
5179 Description: data loss during non-page-aligned writes to a single file from
5180 both multiple nodes and multiple threads on one node at same time
5181 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5182 inconsistency led to false short-reads, that were cached and later
5183 used by ->prepare_write() to fill in partially written page,
5184 leading to data loss.
5187 Frequency : always, if lconf --abort_recovery used
5189 Description: lconf --abort_recovery fails with 'Operation not supported'
5190 Details : lconf was attempting to abort recovery on the MDT device and not
5193 Severity : enhancement
5195 Description: remove cleanup logs
5196 Details : replace lconf-generated cleanup logs with lustre internal
5197 cleanup routines. Eliminates the need for client-cleanup and
5200 Severity : enhancement
5202 Description: add support for EAs (user and system) on lustre filesystems
5203 Details : it is now possible to store extended attributes in the Lustre
5204 client filesystem, and with the user_xattr mount option it
5205 is possible to allow users to store EAs on their files also
5207 Severity : enhancement
5209 Description: Add possibility (config option) to show minimal available OST free
5211 Details : When compiled with --enable-mindf configure option, statfs(2)
5212 (and so, df) will return least minimal free space available from
5213 all OSTs as amount of free space on FS, instead of summary of
5214 free spaces of all OSTs.
5216 Severity : enhancement
5218 Description: do not expand extent locks acquired on OST-side
5219 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5220 by server: they are not cached anyway.
5223 Frequency : when mmap is used/binaries executed from Lustre
5225 Description: Unmmap pages before throwing them away from read cache.
5226 Details : llap_shrink cache now attempts to unmap pages before discarding
5227 them (if unmapping failed - do not discard). SLES9 kernel has
5228 extra checks that trigger if this unmapping is not done first.
5233 Description: lconf didn't resolve symlinks before checking to see whether a
5234 given mountpoint was already in use
5237 Frequency : when migrating failover services
5238 Bugzilla : 6395, 9514
5239 Description: When migrating a subset of services from a node (e.g. failback
5240 from a failover service node) the remaining services would
5241 time out and evict clients.
5242 Details : lconf --force (implied by --failover) sets the global obd_timeout
5243 to 5 seconds in order to quickly disconnect, but this caused
5244 other RPCs to time out too quickly. Do not change the global
5245 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5247 Severity : enhancement
5248 Frequency : if MDS is started with down OST
5249 Bugzilla : 9439,5706
5250 Description: Allow startup/shutdown of an MDS without depending on the
5251 availability of the OSTs.
5252 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5253 Add appropriate locking and lov-osc refcounts for safe
5254 cleaning. Add osc abort_inflight calls in case the
5255 synchronize never started.
5258 Frequency : occasional (Cray XT3 only)
5260 Description: root not authorized to access files in CRAY_PORTALS environment
5261 Details : The client process capabilities were not honoured on the MDS in
5262 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5263 been accessed by an authorized user then root was able to access
5264 the file on the local client also. The root user capabilities
5265 are now allowed on the MDS, as this environment has secure UID.
5268 Frequency : occasional
5270 Description: ldiskfs "too long searching" message happens too often
5271 Details : A debugging message (otherwise harmless) prints too often on
5272 the OST console. This has been reduced to only happen when
5273 there are fragmentation problems on the filesystem.
5278 Description: Division by zero in statfs when all OSCs are inactive
5279 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5280 lov_statfs divides by value returned by lov_get_stripecnt().
5284 Bugzilla : 9489, 3273
5285 Description: First write from each client to each OST was only 4kB in size,
5286 to initialize client writeback cache, which caused sub-optimal
5287 RPCs and poor layout on disk for the first written file.
5288 Details : Clients now request an initial cache grant at (re)connect time
5289 and so that they can start streaming writes to the cache right
5290 away and always do full-sized RPCs if there is enough data.
5291 If the OST is rebooted the client also re-establishes its grant
5292 so that client cached writes will be honoured under the grant.
5297 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5298 Details : Now I/O RPCs go to different portal number and (presumably) fast
5299 lock requests (and glimses) and other RPCs get their own service
5300 threads pool that should be able to service those RPCs
5303 Severity : enhancement
5305 Description: Ability to exchange lustre version between client and servers and
5306 issue warnings at client side if client is too old. Also for
5307 liblustre clients there is ability to refuse connection of too old
5309 Details : New 'version' field is added to connect data structure that is
5310 filled with version info. That info is later checked by server and
5314 Frequency : rare, liblustre only.
5315 Bugzilla : 9296, 9581
5316 Description: Two simultaneous writes from liblustre at offset within same page
5317 might proceed at the same time overwriting eachother with stale
5319 Details : I/O lock withing llu_file_prwv was released too early, before data
5320 actually was hitting the wire. Extended lock-holding time until
5321 server acknowledges receiving data.
5324 Frequency : extremely rare. Never observed in practice.
5326 Description: avoid generating lustre_handle cookie of 0.
5327 Details : class_handle_hash() generates handle cookies by incrementing
5328 global counter, and can hit 0 occasionaly (this is unlikely, but
5329 not impossible, because initial value of cookie counter is
5330 selected randonly). Value of 0 is used as a sentinel meaning
5331 "unassigned handle" --- avoid it. Also coalesce two critical
5332 sections in this function into one.
5334 Severity : enhancement
5336 Description: allow liblustre clients to delegate truncate locking to OST
5337 Details : To avoid overhead of locking, liblustre client instructs OST to
5338 take extent lock in ost_punch() on client's behalf. New connection
5339 flag is added to handle backward compatibility.
5341 Severity : enhancement
5342 Bugzilla : 4928, 7341, 9758
5343 Description: allow number of OST service threads to be specified
5344 Details : a module parameter allows the number of OST service threads
5345 to be specified via "options ost ost_num_threads={N}" in the
5346 OSS's /etc/modules.conf or /etc/modprobe.conf.
5350 Bugzilla : 6146, 9635, 9895
5351 Description: servers crash with bad pointer in target_handle_connect()
5352 Details : In rare cases when a client is reconnecting it was possible that
5353 the connection request was the last reference for that export.
5354 We would temporarily drop the export reference and get a new
5355 one, but this may have been the last reference and the export
5356 was just destroyed. Get new reference before dropping old one.
5358 Severity : enhancement
5359 Frequency : if client is started with failover MDS
5361 Description: Allow multiple MDS hostnames in the mount command
5362 Details : Try to read the configuration from all specified MDS
5363 hostnames during a client mount in case the "primary"
5366 Severity : enhancement
5368 Description: Stop sending data to evicted clients as soon as possible.
5369 Details : Check if the client we are about to send or are sending data to
5370 was evicted already. (Check is done every second of waiting,
5371 for which l_wait_event interface was extended to allow checking
5372 of exit condition at specified intervals).
5375 Frequency : rare, normally only when NFS exporting is done from client
5377 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5379 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5382 Frequency : always, if srand() is called before liblustre initialization
5384 Description: Liblustre uses system PRNG disturbing its usage by user application
5385 Details : Introduce internal to lustre fast and high-quality PRNG for
5386 lustre usage and make liblustre and some other places in generic
5387 lustre code to use it.
5389 Severity : enhancement
5390 Bugzilla : 9477, 9557, 9870
5391 Description: Verify that the MDS configuration logs are updated when xml is
5392 Details : Check if the .xml configuration logs are newer than the config
5393 logs stored on the MDS and report an error if this is the case.
5394 Request --write-conf, or allow starting with --old_conf.
5396 Severity : enhancement
5398 Description: Handle symlinks in the path when checking if Lustre is mounted.
5399 Details : Resolve intermediate symlinks when checking if a client has
5400 mounted a filesystem to avoid duplicate client mounts.
5405 Description: lconf can hit an error exception but still return success.
5406 Details : The lconf command catches the Command error exception at the top
5407 level script context and will exit with the associated exit
5408 status, but doesn't ensure that this exit status is non-zero.
5413 Description: failure of ptlrpc thread startup can cause oops
5414 Details : Starting a ptlrpc service thread can fail if there are a large
5415 number of threads or the server memory is very fragmented.
5416 Handle this without oopsing.
5419 Frequency : always, only if liblustre and non-default acceptor port was used
5421 Description: liblustre cannot connect to servers with non-default acceptor port
5422 Details : tcpnal_set_default_params() was not called and was therefore
5423 ignoring the environment varaible TCPNAL_PORT, as well as other
5424 TCPNAL_ environment variables
5429 Description: two objects could be created on the same OST for a single file
5430 Details : If an OST is down, in some cases it was possible to create two
5431 objects on a single OST for a single file. No problems other
5432 than potential performance impact and spurious error messages.
5436 Bugzilla : 5681, 9562
5437 Description: Client may oops in ll_unhash_aliases
5438 Details : Client dcache may become inconsistent in race condition.
5439 In some cases "getcwd" can fail if the current directory is
5445 Description: Inode refcounting problems in NFS export code
5446 Details : link_raw functions used to call d_instantiate without obtaining
5447 extra inode reference first.
5451 Bugzilla : 9942, 9903
5452 Description: Referencing freed requests leading to crash, memleaks with NFS.
5453 Details : We used to require that call to ll_revalidate_it was always
5454 followed by ll_lookup_it. Also with revalidate_special() it is
5455 possible to call ll_revalidate_it() twice for the same dentry
5456 even if first occurence returned success. This fix changes semantic
5457 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5458 reference on a request referred from the intent.
5459 ll_intent_release() then releases such a request.
5462 Frequency : rare, normally benchmark loads only
5464 Description: unlinked inodes were kept in memory on the client
5465 Details : If a client is repeatedly creating and unlinking files it
5466 can accumulate a lot of stale inodes in the inode slab cache.
5467 If there is no other client load running this can cause the
5468 client node to run out of memory. Instead flush old inodes
5469 from client cache that have the same inode number as a new inode.
5472 Frequency : SLES9 2.6.5 kernel and long filenames only
5473 Bugzilla : 9969, 10379
5474 Description: utime reports stale NFS file handle
5475 Details : SLES9 uses out-of-dentry names in some cases, which confused
5476 the lustre dentry revalidation. Change it to always use the
5480 Frequency : rare, unless heavy write-truncate concurrency is continuous
5481 Bugzilla : 4180, 6984, 7171, 9963, 9331
5482 Description: OST becomes very slow and/or deadlocked during object unlink
5483 Details : filter_destroy() was holding onto the parent directory lock
5484 while truncating+unlinking objects. For very large objects this
5485 may block other threads for a long time and slow overall OST
5486 responsiveness. It may also be possible to get a lock ordering
5487 deadlock in this case, or run out of journal credits because of
5488 the combined truncate+unlink. Solution is to do object truncate
5489 first in one transaction without parent lock, and then do the
5490 final unlink in a new transaction with the parent lock. This
5491 reduces the lock hold time dramatically.
5494 Frequency : rare, 2.4 kernels only
5496 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5497 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5498 before unmounting their filesystem in order to avoid deadlock.
5499 The kernel_locked() function in 2.4 kernels only checks whether
5500 the kernel lock is held, not whether it is this process that is
5501 holding it as 2.6 kernels do.
5506 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5507 Details : The client ptlrpc code may be trying to reconnect to a down
5508 server before a previous connection attempt has timed out.
5509 Increase the reconnect interval to be longer than the connection
5510 timeout interval to avoid sending duplicate connections to
5514 Frequency : echo_client brw_test command
5516 Description: fix echo_client to work with OST preallocated code
5517 Details : OST preallocation code (5137) didn't take echo_client IO path
5518 into account: echo_client calls filter methods outside of any
5519 OST thread and, hence, there is no per-thread preallocated
5520 pages and buffers to use. Solution: hijack pga pages for IO. As
5521 a byproduct, this avoids unnecessary data copying.
5525 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5526 Description: Client can oops in mdc_commit_close() after open replay
5527 Details : It was possible for the MDS to return an open request with no
5528 transaction number in mds_finish_transno() if the client was
5529 evicted, but without actually returning an error. Clients
5530 would later try to replay that open and may trip an assertion
5531 Simplify the client close codepath, and always return an error
5532 from the MDS in case the open is not successful.
5535 Frequency : rare, 2.6 OSTs only
5537 Description: OST may deadlock under high load on fragmented files
5538 Details : If there was a heavy load and highly-fragmented OST filesystems
5539 it was possible to have all the OST threads deadlock waiting on
5540 allocation of biovecs, because the biovecs were not released
5541 until the entire RPC IO was completed. Instead, release biovecs
5542 as soon as they are complete to ensure forward IO progress.
5544 Severity : enhancement
5546 Description: Support for specifying external journal device at mount
5547 Details : If an OST or MDS device is formatted with an external journal
5548 device, this device major/minor is stored in the ext3 superblock
5549 and may not be valid for failover. Allow detecting and
5550 specifying the external journal at mount time.
5555 Description: Mounting an MDS with pending unlinked files may cause oops
5556 Details : target_finish_recovery() calls mds_postrecov() which returned
5557 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5558 considers this an error and immediately begins cleaning up the
5559 lov, just after starting the mds_lov process
5561 Severity : enhancement
5563 Description: Implement 'lfs df' to report actual free space on per-OST basis
5564 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5565 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5566 usage of objects; '-h' to report in human readable format.
5568 ------------------------------------------------------------------------------
5570 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5577 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5578 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5579 at startup during orphan unlinking. Instead, skip buddy update
5580 before setup, as it will be regenerated after recovery is complete.
5585 Description: If an OST is inactive, its locks might reference stale inodes.
5586 Details : lov_change_cbdata() must iterate over all namespaces, even if
5587 they are inactive to clear inode references from the lock.
5589 Severity : enhancement
5590 Frequency : occasional, if non-standard max_dirty_mb used
5592 Description: Client will block write RPCs if not enough grant
5593 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5594 then the client will block writes while waiting for another RPC
5595 to complete instead of consuming its dirty limit. With change
5596 we get improved performance when max_dirty_mb is small.
5598 Severity : enhancement
5599 Bugzilla : 3389, 6253
5600 Description: Add support for supplementary groups on the MDS.
5601 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5602 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5603 lookups for user supplementary groups into a cache.
5607 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5608 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5609 for liblustre O_EXCL is enforced by MDS.
5613 Description: readdir never returns NULL in liblustre.
5614 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5615 dentry in struct dirent.
5619 Frequency : liblustre only, depends on application IO pattern
5620 Description: liblustre clients evicted if not contacting servers
5621 Details : Don't put liblustre clients into the ping_evictor list, so
5622 they will not be evicted by the pinger ever.
5624 Severity : enhancement
5626 Description: Add ability to evict clients by NID from MDS.
5627 Details : By echoing "nid:$NID" string into
5628 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5629 $NID would be instantly evicted from this MDS and from all active
5630 OSTs connected to it.
5634 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5635 Details : lookup_it_finish() used to query file size from OSTs that was not
5640 Description: service threads change working directory to that of init
5641 Details : Starting lustre service threads may pin the working directory
5642 of the parent thread, making that filesystem busy. Threads
5643 now change to the working directory of init to avoid this.
5647 Frequency : during shutdown only
5648 Description: shutdown with a failed MDS or OST can cause unmount to hang
5649 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5654 Frequency : 2.6 only
5655 Description: chmod/chown may include an extra supplementary group
5656 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5657 supplementary group and if none is specified this is used.
5660 Bugzilla : 5479 (6816)
5662 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5663 Details : If lookup is in progress on a file that is unlinked we might try
5664 to revalidate the inode and fail in revalidate after lookup is
5665 complete and ll_file_open() enqueues the open again but
5666 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5669 Frequency : always, if lconf --abort_recovery used
5671 Description: lconf --abort_recovery fails with 'Operation not supported'
5672 Details : lconf was attempting to abort recovery on the MDT device and not
5675 ------------------------------------------------------------------------------
5677 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5682 Frequency : rare (only unsupported configurations with a node running as an
5684 Bugzilla : 6514, 5137
5685 Description: Mounting a Lustre file system on a node running as an OST could
5687 Details : OSTs now preallocates memory needed to write out data at
5688 startup, instead of when needed, to avoid having to
5689 allocate memory in possibly low memory situations.
5690 Specifically, if the file system is mounted on on OST,
5691 memory pressure could force it to try to write out data,
5692 which it needed to allocate memory to do. Due to the low
5693 memory, it would be unable to do so and the node would
5694 become unresponsive.
5696 Severity : enhancement
5698 Description: Addition of lconf --service command line option
5699 Details : lconf now accepts a '--service <arg>' option, which is
5700 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5702 Severity : enhancement
5704 Description: Failover mode is now the default for OSTs.
5705 Details : By default, OSTs will now run in failover mode. To return to
5706 the old behaviour, add '--failout' to the lmc line for OSTs.
5708 Severity : enhancement
5710 Description: Health checks are now provided for MDS and OSTs
5711 Details : Additional detailed health check information on MSD and OSTs
5712 is now provided through the procfs health_check value.
5715 Frequency : occasional, depends on IO load
5717 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5718 after numerous create/delete cycles
5719 Details : The ext3 inode allocation policy would not allocate new inodes
5720 very well on the OSTs because there are no new directories
5721 being created. Instead we look for groups with free space if
5722 the parent directories are nearly full.
5727 Description: Network or server problems during mount may cause partially
5728 mounted clients instead of returning an error.
5729 Details : The config llog parsing code may overwrite the error return
5730 code during mount error handling, returning success instead
5736 Description: MDS can fail to allocate large reply buffers
5737 Details : After long uptimes the MDS can fail to allocate large reply
5738 buffers (e.g. zconf client mount config records) due to memory
5739 fragmentation or consumption by the buffer cache. Preallocate
5740 some large reply buffers so that these replies can be sent even
5741 under memory pressure.
5745 Frequency : rare (liblustre)
5746 Description: fsx running with liblustre complained that using truncate() to
5747 extend the file doesn't work. This patch corrects that issue.
5748 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5749 ATTR_SIZE and lsm use in llu_setattr_raw.
5753 Frequency : rare, only 2.6 kernels
5754 Description: Unusual file access patterns on the MDS may result in inode
5755 data being lost in very rare circumstances.
5756 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5757 code in the 2.6 kernel caused the inode and/or EA data not to
5758 be read from disk, causing single-file corruption.
5762 Frequency : rare, only 2.6 filesystems using extents
5763 Description: Heavy concurrent write and delete load may cause data corruption.
5764 Details : It was possible under high-load situations to have an extent
5765 metadata block in the block device cache from a just-unlinked
5766 file overwrite a newly-allocated data block. We now unmap any
5767 metadata buffers that alias just-allocated data blocks.
5771 Frequency : filesystems with default stripe_count larger than 77
5772 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5773 Details : lconf specifies an inode size of 4096 bytes when the default
5774 stripe_count is larger than 77. This conflicts with the default
5775 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5776 this case to avoid pinning too much memory for large EAs.
5778 ------------------------------------------------------------------------------
5780 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5785 Frequency : rare (extremely heavy IO load with hundreds of clients)
5787 Description: Client is evicted, gets IO error writing to file
5788 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5789 caused clients to be evicted for AST timeouts. The fixes in
5790 bug 5192 mean we no longer need to have such short AST timeouts
5791 so ldlm_timeout has been increased.
5794 Frequency : occasional during --force or --failover shutdown under load
5795 Bugzilla : 5949, 4834
5796 Description: Server oops/LBUG if stopped with --force or --failover under load
5797 Details : a collection of import/export refcount and cleanup ordering
5798 issues fixed for safer force cleanup
5801 Frequency : only filesystems larger than 120 OSTs
5802 Bugzilla : 5990, 6223
5803 Description: lfs getstripe would oops on a very large filesystem
5804 Details : lov_getconfig used kfree on vmalloc'd memory
5807 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5808 Bugzilla : 6242, 6243
5809 Description: reading from files that had been truncated to a non-zero size
5810 but never opened returned no data
5811 Details : ll_file_read() reads zeros from no-object files to EOF
5816 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5817 Details : There's little harm in aborting MDS/OSS recovery and letting it
5818 try again, so I removed the LASSERT and return an error instead.
5820 Severity : enhancement
5822 Description: New debugging infrastructure for tracking down data corruption
5823 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5824 (b) cover more of the client-side code path, and (c) try to narrow
5825 down where problems occurred
5829 Bugzilla : 3819, 4364, 4397, 6313
5830 Description: Racing close and eviction MDS could cause assertion in mds_close
5831 Details : It was possible to get multiple mfd references during close and
5832 client eviction, leading to one thread referencing a freed mfd.
5834 Severity: : enhancement
5835 Bugzilla : 3262, 6359
5836 Description: Attempts to reconnect to servers are now more aggressive.
5837 Details : This builds on the enhanced upcall-less recovery that was added
5838 in 1.4.2. When trying to reconnect to servers, clients will
5839 now try each server in the failover group every 10 seconds. By
5840 default, clients would previously try one server every 25 seconds.
5845 Description: After recovery, certain operations trigger a failed
5846 assertion on a client.
5847 Details : Failing over an mds, using lconf -d --failover, while a
5848 client was doing a readdir() call would cause the client to
5849 LBUG after recovery completed and the readdir() was resent.
5851 Severity : enhancement
5853 Description: Default groups are now added by lconf
5854 Details : You can now run lconf --group <servicename> without having to
5855 manually add groups with lmc.
5858 Frequency : occasional
5860 Description: Nodes with an elan id of 0 trigger a failed assertion
5863 Frequency : always when accessing e.g. tty/console device nodes
5865 Description: tty and some other devices nodes cannot be used on lustre
5866 Details : file's private_data field is used by device data and lustre
5867 values in there got lost. New field was added to struct file to
5868 store fs-specific private data.
5871 Frequency : when exporting Lustre via NFS
5873 Description: NFSD failed occasionally when looking up a path component
5874 Details : NFSD is looking up ".." which was broken in ext3 directories
5875 that had grown large enough to become hashed.
5878 Frequency : Clusters with multiple interfaces not on the same subnet
5880 Description: Nodes will repeatedly try to reconnect to an interface which it
5881 cannot reach and report an error to the log.
5882 Details : Extra peer list entries will be created by lconf with some peers
5883 unreachable. lconf now validates the peer before adding it.
5886 Frequency : Only if a default stripe is set on the filesystem root.
5888 Description: Setting a default stripe on the filesystem root prevented the
5889 filesystem from being remounted.
5890 Details : The client was sending extra request flags in the root getattr
5891 request and did not allocate a reply buffer for the dir EA.
5894 Frequency : occasional, higher if lots of files are accessed by one client
5895 Bugzilla : 6159, 6097
5896 Description: Client trips assertion regarding lsm mismatch/magic
5897 Details : While revalidating inodes the VFS looks up inodes with ifind()
5898 and in rare cases can find an inode that is being freed.
5899 The ll_test_inode() code will free the lsm during ifind()
5900 when it finds an existing inode and then the VFS later attaches
5901 this free lsm to a new inode.
5905 Bugzilla : 6422, 7030
5906 Description: MDS deadlock between mkdir and client eviction
5907 Details : Creating a new file via mkdir or mknod (starting a transaction
5908 and getting the ns lock) can deadlock with client eviction
5909 (gets ns lock and trying to finish a synchronous transaction).
5912 Frequency : occasional
5913 Description: While starting a server, the fsfilt_ext3 module could not be
5915 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5916 kernels. Previously, lconf would still use the ext3 name
5917 when trying to load modules. Now, it will correctly use
5918 ext3 on 2.4 and ldiskfs on 2.6.
5920 Severity : enhancement
5921 Description: The default stripe count has been changed to 1
5922 Details : The interpretation of the default stripe count (0, to lfs
5923 or lmc) has been changed to mean striping across a single
5924 OST, rather than all available. For general usage we have
5925 found a stripe count of 1 or 2 works best.
5927 Severity : enhancement
5928 Description: Add support for compiling against Cray portals.
5929 Details : Conditional compiling for some areas that are different
5933 Frequency : occasional
5934 Bugzilla : 6409, 6834
5935 Description: Creating files with an explicit stripe count may lead to
5936 a failed assertion on the MDS
5937 Details : If some OSTs are full or unavailable, creating files may
5938 trigger a failed assertion on the MDS. Now, Lustre will
5939 try to use other servers or return an error to the
5943 Frequency : occasional
5945 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5946 caused lock timeout during readahead (since 1.4.2).
5947 Details : Processes doing readahead might match a lock that hasn't been
5948 granted yet if there are overlapping and conflicting lock
5949 requests. The readahead process waits on ungranted lock
5950 (original lock is CBPENDING), while OST waits for that process
5951 to cancel CBPENDING read lock and eventually evicts client.
5953 Severity : enhancement
5955 Description: Initial enabling of flock support for clients
5956 Details : Implements fcntl advisory locking and file status functions.
5957 This feature is provided as an optional mount flag (default
5958 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5959 locking are implemented yet, and those that are are not guaranteed
5960 to be completely correct in production environments.
5961 mount -t lustre -o [flock|noflock] ...
5964 Frequency : occasional
5966 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5967 assertion in the ext3 JBD (journaling) layer.
5968 Details : The b_committed_data struct is protected by the big kernel lock
5969 in 2.4 kernels, serializing journal_commit_transaction() and
5970 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5971 access to this struct. In 2.6 kernels there is finer grained
5972 locking to improve SMP performance of the JBD layer.
5976 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5981 Description: merge and cleanup kernel patches.
5984 -----------------------------------------------------------------------------
5986 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5988 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5989 versions of Lustre. Please update all servers and clients to
5990 version 1.4.2 or later at the same time. You must also run
5991 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5992 to update the configuration logs.
5994 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5995 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5996 - swab llog records of type '0' so we get proper header size/idx (5861)
5997 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5998 - fix rename of one directory over another leaking an inode (5953)
5999 - avoid SetPageDirty on 2.6 (5981)
6000 - don't re-add just-being-destroyed locks to the waiting list (5653)
6001 - when creating new directories, inherit the parent's custom
6002 striping settings if present parent (3048)
6003 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
6004 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
6005 - don't hold client locks on temporary worklist from l_lru (5666)
6006 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
6007 - automatically evict dead clients (5921)
6008 - Update file size properly in create+truncate+fstat case (6196)
6009 - Do not unhash mountpoint dentries, do not allow removal of
6011 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
6012 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
6013 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
6014 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
6015 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
6016 - Correctly handle reads of files with no objects (6243)
6017 - lctl recover will also mark a device active if deactivate used (5933)
6019 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
6020 - allow --write-conf on an MDS with different nettype than client (5619)
6021 - don't write config llogs to MDS for mounts not from that MDS (5617)
6022 - lconf should create multiple TCP connections from a client (5201)
6023 - init scripts are now turned off by default; run chkconfig --on
6024 lustre and chkconfig --on lustrefs to use them
6025 - upcalls are no longer needed for clients to recover to failover
6027 - add --abort-recovery option to lconf to abort recovery on device
6029 - add support for an arbitrary number of OSTs (3026)
6030 - Quota support protocol changes.
6031 - forward compatibility changes to wire structs (6007)
6032 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
6033 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
6034 - improved "lustre status" script
6035 - initialize blocksize for non-regular files (6062)
6036 - added --disable-server and --disable-client configure options (5782)
6037 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
6038 - Vanilla 2.4.29 support
6039 - increase maximum number of obd devices to 520 (6242)
6040 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
6041 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
6042 SLES 9 2.6.5 kernels
6043 - limit stripes per file to 160 (the maximum EA size) (6093)
6045 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6048 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
6049 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
6050 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
6051 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
6052 - small ext3 extents cleanups and fixes (5733)
6053 - improved mballoc code, several small races and bugs fixed (5733, 5638)
6054 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
6055 - avoid needless client->OST connect, fix handle mismatch (5317)
6056 - fix DLM error path that led to out-of-sync client, long delays (5779)
6057 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
6058 - fix several locking issues related to i_size (5492,5624,5654,5672)
6059 - don't move pending lock onto export if it is already evicted (5683)
6060 - fix kernel oops when creating .foo in unlinked directory (5548)
6061 - fix deadlock in obdfilter statistics vs. object create (5811)
6062 - use time_{before,after} to avoid timer jiffies wrap (5882)
6063 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
6064 - Do not leak request if server was not able to process it (5154)
6065 - If mds_open unable to find parent dir, make that negative lookup(5154)
6066 - don't create new directories with extent-mapping (5909, 5936)
6068 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
6069 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
6070 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
6071 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
6072 - support for RHEL4 kernel on i686 (5773)
6073 - provide error messages when incompatible logs are encountered (5898)
6075 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
6076 * version 1.4.0.10 (1.4.1 release candidate 1)
6078 - don't keep a lock reference when lock is not granted (4238)
6079 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6080 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
6081 - drop import inflight refcount on signal_completed_replay error (5255)
6082 - unlock page after async write error during send (3677)
6083 - handle missing objects in filter_preprw_read properly (5265)
6084 - no transno return for symlink open, don't save no-trasno open (3440)
6085 - don't try to complete elan receive that already failed (4012)
6086 - free RPC server reply state on error (5406)
6087 - clean up thread from ptlrpc_start_thread() on error (5160)
6088 - readahead could read extra page into cache that wasn't ejected (5388)
6089 - prevent races in class_attach/setup/cleanup/detach (5260)
6090 - don't dereference de->d_inode after l_dput of de (5458)
6091 - use "int" for stripe value returned from lock_to_stripe (5544)
6092 - mballoc allocation and error-checking fixes in 2.6 (5504)
6093 - block device patches to fix I/O request sizes in 2.6 (5482)
6094 - look up hostnames for IB nals (5602)
6095 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
6096 - don't start multiple acceptors for the same port (5277)
6097 - fix incorrect LASSERT in mds_getattr_name (5635)
6098 - export a proc file for general "ping" checking (5628)
6099 - fix "lfs check" to not block when the MDS is down (5628)
6101 - service request history (4965)
6102 - put {ll,lov,osc}_async_page structs in a single slab (4699)
6103 - create an "evict_client" /proc entry on OSTs, like the MDS has
6104 - fix mount usage message, return errors per mount(8) (5168)
6105 - change grep [] to grep "[]" in tests so they work in more UMLs
6106 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
6107 - remove ancient LOV_MAGIC_V0 EA support (5047)
6108 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
6109 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
6110 - allow readahead allocations to fail when low on memory (5383)
6111 - mmap locking landed again, after considerable improvement (2828)
6112 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
6114 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
6117 - send OST transaction number in read/write reply to free req (4966)
6118 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6119 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6120 - the watchdog thread now runs as interruptible (5246)
6121 - flock/lockf fixes (but it's still disabled, pending 5135)
6122 - don't use EXT3 constants in llite code (5094)
6123 - memory shortage at startup could cause assertion (5176)
6125 - reorganization of lov code
6126 - single portals codebase
6128 - add extents/mballoc support (5025)
6129 - direct I/O reads in the obdfilter (4048)
6130 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6132 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6135 - send OST transaction number in read/write reply to free req (4966)
6136 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6137 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6138 - the watchdog thread now runs as interruptible (5246)
6139 - handle missing objects in filter_preprw_read properly (5265)
6140 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6141 - drop import inflight refcount on signal_completed_replay error (5255)
6142 - unlock page after async write error during send (3677)
6143 - return original error code on reconstructed replies (3761)
6144 - no transno return for symlink open, don't save no-trasno open (3440)
6146 - add pid to ldlm debugging output (4922)
6147 - bump the watchdog timeouts -- we can't handle 30sec yet
6148 - extra debugging for orphan dentry/inode bug (5259)
6150 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6153 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6154 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6155 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6156 - fix NULL dereference in /proc/sys/portals/routes (4827)
6157 - allow failed mdc_close() operations to be interrupted (4561)
6158 - stop precreate on OST before MDS would time out on it (4778)
6159 - don't send partial-page writes before EOF from client (4410)
6160 - discard client grant for sub-page writes on large-page clients (4520)
6161 - don't free dentries not owned by NFS code, check generation (4806)
6162 - fix lsm leak if mds_create_objects() fails (4801)
6163 - limit debug_daemon file size, always print CERROR messages (4789)
6164 - use transno after validating reply (3892)
6165 - process timed out requests if import state changes (3754)
6166 - update mtime on OST during writes, return in glimpse (4829)
6167 - add mkfsoptions to LDAP (4679)
6168 - use ->max_readahead method instead of zapping global ra (5039)
6169 - don't interrupt __l_wait_event() during strace
6171 - add software watchdogs to catch hung threads quickly (4941)
6172 - make lustrefs init script start after nfs is mounted
6173 - fix CWARN/ERROR duplication (4930)
6174 - return async write errors to application if possible (2248)
6175 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6176 - print NAL number in %x format (4645)
6177 - update barely-supported suse-2.4.21-171 series (4842)
6178 - support for sles 9 %post scripts
6179 - support for building 2.6 kernel-source packages
6180 - support for sles km_* packages
6182 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6185 - ignore -ENOENT errors in osc_destroy (3639)
6186 - notify osc create thread that OSC is being cleaned up (4600)
6187 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6188 - reconstruct ost_handle() like mds_handle() (4657)
6189 - create a new thread to do import eviction to avoid deadlock (3969)
6190 - let lconf resolve symlinked-to devices (4629)
6191 - don't unlink "objects" from directory with default EA (4554)
6192 - hold socknal file ref over connect in case target is down (4394)
6193 - allow more than 32000 subdirectories in a single directory (3244)
6194 - fix blocks count for O_DIRECT writes (3751)
6195 - OST returns ENOSPC from object create when no space left (4539)
6196 - don't send truncate RPC if file size isn't changing (4410)
6197 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6198 - bind to privileged port in socknal and tcpnal (3689)
6200 - rate limit CERROR/CWARN console message to avoid overload (4519)
6201 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6202 - basic mmap support (3918)
6203 - kernel patch series update from b1_4 (4711)
6205 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6208 - avoid crash during MDS cleanup with OST shut down (2775)
6209 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6210 - don't use bad inodes on the MDS (3744)
6211 - dynamic object preallocation to improve recovery speed (4236)
6212 - don't hold spinlock over lock dumping or change debug flags (4401)
6213 - don't zero obd_dev when it is force cleaned (3651)
6214 - print grants to console if they go negative (4431)
6215 - "lctl deactivate" will stop automatic recovery attempts (3406)
6216 - look for existing locks in ldlm_handle_enqueue() (3764)
6217 - don't resolve lock handle twice in recovery avoiding race (4401)
6218 - revalidate should check working dir is a directory (4134)
6220 - don't always mark "slow" obdfilter messages as errors (4418)
6222 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6225 - don't close LustreDB during write_conf until it is done (3860)
6226 - fix typo in lconf for_each_profile (3821)
6227 - allow dumping logs from multiple threads at one time (3820)
6228 - don't allow multiple threads in OSC recovery (3812)
6229 - fix debug_size parameters (3864)
6230 - fix mds_postrecov to initialize import for llog ctxt (3121)
6231 - replace config semaphore with spinlock (3306)
6232 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6233 - don't allow enqueue to complete on a destroyed export (3822)
6234 - down write_lock before checking llog header bitmap (3825)
6235 - recover from lock replay timeout (3764)
6236 - up llog sem before sending rpc (3652)
6237 - reduce ns lock hold times when setting kms (3267)
6238 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6239 - fix NULL deref and obd_dev leak on setup error (3312)
6240 - replace some LBUG about llog ops with error handling (3841)
6241 - don't match INVALID dentries from d_lookup and spin (3784)
6242 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6243 - fix invalid assertion in ptlrpc_set_wait (3880)
6245 - add libwrap support for the TCP acceptor (3996)
6246 - add /proc/sys/portals/routes for non-root route listing (3994)
6247 - allow setting MDS UUID in .xml (2580)
6248 - print the stack of a process that LBUGs (4228)
6250 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6253 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6254 - make sure to unset replay flag from failed open requests (3440)
6255 - if default stripe count is 0, use OST count for inode size (3636)
6256 - update parent mtime/ctime on client for create/unlink (2611)
6257 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6258 - free recovery state on server during a forced cleanup (3571)
6259 - unregister_reply for resent reqs (3063)
6260 - loop back devices mounting and status check on 2.6 (3563)
6261 - fix resource-creation race that can provoke i_size == 0 (3513)
6262 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6263 - more debugging for page-accounting assertion (3746)
6264 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6265 - avoid deadlock after precreation failure (3758)
6266 - fix race and lock order deadlock in orphan handling (3450, 3750)
6267 - add validity checks when grabbing inodes from l_ast_data (3599)
6269 - add /proc/.../recovery_status to obdfilter (3428)
6270 - lightweight CDEBUG infrastructure, debug daemon (3668)
6271 - change default OSC RPC parameters to be better on small clusters
6272 - turn off OST read cache for files smaller than 32MB
6273 - install man pages and include them in rpms (3100)
6274 - add new init script for (un)mounting lustre filesystems (2593)
6275 - run chkconfig in %post for init scripts (3701)
6276 - drop scimac NAL (unmaintained)
6278 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6281 - clean kiobufs before and after use (3485)
6282 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6283 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6284 - change init script to fail more clearly if not run as root (1528)
6285 - allow clients to reconnect during replay (1742)
6286 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6287 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6288 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6289 - fix cond_resched (3554)
6290 - only evict unfinished clients after recovery (3515)
6291 - allow bulk resend, prevent data loss (3570)
6292 - dynamic ptlrpc request buffer allocation (2102)
6293 - don't allow unlinking open directory if it isn't empty (2904)
6294 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6295 - remove extraneous obd dereference causing LASSERT failure (3334)
6296 - don't use get_cycles() when creating temp. files on the mds (3156)
6297 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6298 - handle EEXIST for set-stripe, set proper directory name (3336)
6300 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6301 - fix ksocknal_fmb_callback() error messages (2918)
6303 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6306 - don't copy lvb into (possibly NULL) reply on error (2983)
6307 - don't deref dentry after dput, don't free lvb on error (2922)
6308 - use the kms to determine writeback rpc length (2947)
6309 - increment oti_logcookies when osc is inactive (2948)
6310 - update client's i_blocks count via lvb messages (2543)
6311 - handle intent open/close of special files properly (1557)
6312 - mount MDS with errors=remount-ro, like obdfilter (2009)
6313 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6314 - don't use cancelling-locks' kms values (2947)
6315 - use highest lock extent for kms, not last one (2925)
6316 - don't dereference ERR_PTR() dentry in error handling path (3107)
6317 - fix thread race in portals_debug_dumplog() (3122)
6318 - create lprocfs device entries at setup instead of at attach (1519)
6319 - common AST error handler, don't evict client on completion race (3145)
6320 - zero nameidata in detach_mnt in 2.6 (3118)
6321 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6322 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6323 - initialize RPC timeout timer earlier for 2.6 (3219)
6324 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6325 - print nal/nid for unknown nid (3258)
6326 - additional checks for oscc recovery before doing precreate (3284)
6327 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6328 - don't crash in mdc_close for bad permissions on open (3285)
6329 - zero i_rdev for non-device files (3147)
6330 - clear page->private before handing to FS, better assertion (3119)
6331 - tune the read pipeline (3236)
6332 - fix incorrect decref of invalidated dentry (2350)
6333 - provide read-ahead stats and refine rpc in flight stats (3328)
6334 - don't hold journal transaction open across create RPC (3313)
6335 - update atime on MDS at close time (3265)
6336 - close LDAP connection when recovering to avoid server load (3315)
6337 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6338 - don't leak open file on MDS after open resend (3325)
6339 - serialize filter_precreate and filter_destroy_precreated (3329)
6340 - loop device shouldn't call sync_dev() for nul device (3092)
6341 - clear page cache after eviction (2766)
6342 - resynchronize MDS->OST in background (2824)
6343 - refuse to mount the same filesystem twice on same mountpoint (3394)
6344 - allow llmount to create routes for mounting behind routers (3320)
6345 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6346 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6347 - fix rare problem with rename on htree directories (3417)
6349 - allow default OST striping configuration per directory (1414)
6350 - fix compilation for qswnal for 2.6 kernels (3125)
6351 - increase maximum number of MDS request buffers for large systems
6352 - change liblustreapi to be useful for external progs like lfsck (3098)
6353 - increase local configuration timeout for slow disks (3353)
6354 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6356 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6359 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6360 - don't overwrite extent policy data in reply if lock was blocked (2901)
6361 - drop filter export grants atomically with removal from device (2663)
6362 - del obd_self_export from work_list in class_disconnect_exports (2908)
6363 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6364 - swab reply message in mdc_close, other PPC fixes (2464)
6365 - fix destroying of named logs (2325)
6366 - overwrite old logs when running lconf --write_conf (2264)
6367 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6368 - fix race in target_handle_connect (2898)
6369 - mds_reint_create() should take same inode create lock (2926)
6370 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6371 - don't close files for self_export to avoid uninitialized obd (2936)
6372 - allow MDS with the same name as client node (2939)
6373 - hold dentry reference for closed log files for unlink (2325)
6374 - reserve space for all logs during transactions (2059)
6375 - don't evict page beyond end of stripe extent (2925)
6376 - don't oops on a deleted current working directory (2399)
6377 - handle hard links to targets without a parent properly (2517)
6378 - don't dereference NULL lock when racing during eviction (2867)
6379 - don't grow lock extents when lots of conflicting locks (2919)
6381 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6384 - account for cache space usage on clients to avoid data loss (974)
6385 - lfsck support in lustre kernel code (2349)
6386 - reduce journal credits needed for BRW writes (2370)
6387 - orphan handling to avoid losing space on client/server crashes
6388 - ptlrpcd can be blocked, stopping ALL progress (2477)
6389 - use lock value blocks to assist in proper KMS, faster stat (1021)
6390 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6391 - recovery for initial connections (2355)
6392 - fixes for mds_cleanup_orphans (1934)
6393 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6394 - block all file creations until orphan recovery completes (1901)
6395 - client remove rq_connection from request struct (2423)
6396 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6397 - recovery timer race (2670)
6398 - mdc_close recovey bug (2532)
6399 - ptlrpc cleanup bug (2710)
6400 - mds timeout on local locks (2588)
6401 - namespace lock held during RPCs (2431)
6402 - handle interrupted sync write properly (2503)
6403 - don't try to handle a message that hasn't been replied to (2699)
6404 - client assert failure during cleanup after abort recovery (2701)
6405 - leak mdc device after failed mount (2712)
6406 - ptlrpc_check_set allows timedout requests to complete (2714)
6407 - wait for inflight reqs when ptlrpcd finishes (2710)
6408 - make sure unregistered services are removed from the srv_list
6409 - reset bulk XID's when resending them (caught by 1138 test)
6410 - unregister_bulk after timeout
6411 - fix lconf error (2694)
6412 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6413 - readahead locks pages, leaves pending causing memory pressure (2673)
6414 - increase OST request buffers to 4096 on large machines (2729)
6415 - fix up permission of existing directories in simple_mkdir (2661)
6416 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6417 - don't assign transno to errored transactions (2742)
6418 - don't delete objects on OST if given a bogus objid from MDS (2751)
6419 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6420 - if rq_no_resend, then timeout request after recovery (2432)
6421 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6422 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6423 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6424 - MDS recovery shouldn't delete objects that it has given out (2730)
6425 - if enqueue arrives after completion, don't clobber LVB (2819)
6426 - don't unlock pages twice when trigger_group_io returns error (2814)
6427 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6428 - don't write pages to disk if there was an error (1450)
6429 - don't ping imports that have recovery disabled (2676)
6430 - take buffered bytes into account when balancing socknal conn (2817)
6431 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6432 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6433 - remove little-endian swabbing of llog records (1987)
6434 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6435 - retry reposting request buffers if they fail (1191)
6436 - grow extent at grant time to avoid granting a revoked lock (2809)
6437 - lock revoke doesn't evict page if covered by a second lock (2765)
6438 - disable VM readahead to avoid reading outside lock extents (2805)
6440 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6441 - updated kernel patches for hp-2.4.20 kernel (2681)
6443 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6446 - fix truncated write corruption (2366)
6447 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6449 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6450 - ensure len(uuid) < 37 in lmc (1171)
6451 - fix ia64 OOPS in llog_test (2255)
6452 - zero end of page at obdfilter for partial page writes (2648)
6453 - don't leave stale dentries around after renames (bug 2428)
6454 - fix timeouts when evicting a client with a single lock held (2642)
6455 - set deadline for the initial HELLO message to drain (2634)
6456 - print out dotted-quad IP addresses in the socknal (2302)
6458 - additional debugging for MDS client eviction problem (2443)
6459 - fix mkfsoptions support for osts (2603, 2604)
6461 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6464 - add series for the vanilla 2.6.0 kernel
6465 - add series for the vanilla 2.4.24 kernel
6466 - add series for a cray x86/64 UL kernel drop
6467 - fix xattr patches for the vanilla 2.4.19 series
6469 - generate true UUIDs in lmc (1171)
6470 - have portals stack dumping break in UML (2466)
6471 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6472 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6473 - wake callback waiting threads on client eviction (2460)
6474 - Add --ptldebug and --subsystem to lmc (1719)
6475 - update assertion to allow safe interrupt allocation
6476 - set rq_no_resend for cancel requests (2432)
6477 - recalculate ptlrpcd timeout after resend (2494)
6478 - call vfs_rmdir when removing pending directories (2368)
6479 - fix renaming a file to itself (2429)
6480 - lmc creates a default one-stripe lov (2454)
6481 - expand procfs space to handle large clusters (2326)
6482 - increase UML stack to avoid overflow
6483 - update lconf's list of debug and subsystem masks
6484 - fix lfs find --obd (2510)
6485 - /proc tunable for disabling filter read caching (2591)
6486 - stop rpm packages from altering slapd.conf (2301)
6487 - disable nagle in the socknal under 0conf (2578)
6488 - choose mds inode size based on stripe count (2572)
6489 - fix kernel-source rpm problems (2516)
6491 - add --disable-doc to avoid pdf generation (2421)
6492 - update documentation, tests, type-os, comments
6493 - avoid format warnings on ia64
6494 - remove the TOE NAL
6495 - tiny code cleanups by removing unused fields
6497 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6500 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6501 - remove the most blatant lies from BUILDING file (2371)
6502 - change default debug level to reasonable production setting
6503 - reduce client side cache size to reduce cache flush time
6504 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6505 - make TCP zerocopy and pinger support enabled by default (2476)
6506 - sync writes completed after process exits caused crashes (2319)
6507 - maintain correct mount count on the MDS (2356)
6508 - backout 1557, because 2316 wasn't really fixed
6509 - better file I/O statistics gathering in /proc
6510 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6511 - another kernel patch to fix zero-copy TCP function export
6512 - don't take duplicate lock when processing re-sent getattr (2420)
6513 - lctl uses obd_self_export instead of creating new conn (2353)
6514 - MDS/OST recovery case which requires object creation asserted (2425)
6515 - move lfs from /usr/sbin to /usr/bin in packages
6516 - fix race between mds_client_add and mds_client_free (2417)
6517 - use kmalloc instead of slabs in portals (2430)
6518 - don't create duplicate records when a failover MDS is present (2442)
6519 - remove unnecessary mount age check (2332)
6520 - don't remove directory inodes from locks prematurely (2451)
6521 - don't break if MDS service name is the same as hostname (2103)
6522 - fix races in client write RPC generation when cache full (2482)
6524 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6527 - remove now-unused request->rq_obd (278)
6528 - if an allocation fails, print out how much memory we've used (1933)
6529 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6530 - add the "configurable stack size" patch to most series files (1256)
6531 - ability to write large log records, for 100+ OST configs (2306)
6532 - fix NULL deref when filter_prep fails (2314)
6533 - fix operator precedence error in filter_sync
6534 - dynamic allocation of socknal TX descriptors (2315)
6535 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6536 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6537 - ensure that CWARN messages in Portals always get to the syslog
6538 - __init/__exit are not for prototype decls (ldlm_init/exit)
6539 - x86-64 compile warning fixes
6540 - fix gateway LMC keyword conflict (2318)
6541 - fix MDS lock inversions in getattr/reint paths (1844)
6542 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6543 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6544 - DLM race condition prevented some lock evictions (2328)
6545 - ENOMEM detection and retry on socknal sends (2230)
6546 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6547 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6548 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6549 - better error messages when a client is rejected during recovery (1505)
6550 - avoid cancelling locks which were never granted, after failure (2330)
6551 - fix i_sem/journal inversion in mds_client_add (2333)
6552 - fix truncate/getattr lock cycle deadlock (2334)
6553 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6554 - fix two rare exit paths which could leak an l_lock() ref (2321)
6555 - fencepost error in MDS/OST orphan recovery (2226)
6556 - make log record alignment 8 bytes (1988)
6557 - lstripe now fails when requested offset > ost_count (2237)
6558 - ensure that all kernel series have a complete list.h (1607)
6559 - fix crashes in special-file operations (2316)
6560 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6562 - allow configurable automake binary, for testing new versions
6563 - small update to the lfs documentation
6565 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6567 * fix negative export reference count in fsfilt_sync (2312)
6569 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6570 * release candidate 0.9.1
6572 - orphans are moved into the PENDING directory for possible recovery
6573 - replayed opens now open by fid for orphan/rename safety (1042)
6574 - last close of an orphan inode generates a transno (683)
6575 - chdir() and mount() now pin the directory entry (1020)
6576 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6577 - discard very old requests without processing them (1502)
6578 - remove some common, well-understood CERRORs (1505)
6579 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6580 - clear "grant" flags in OST replies until OST grant code lands (1644)
6581 - fix read performance by not clobbering i_blksize on client (1598)
6582 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6583 - make LRU size a /proc tunable, clears locks when reduced (707)
6584 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6585 - server locks take references on exports now (1558)
6586 - build fixes for 2.4.20-rh trees (1663)
6587 - return an error from lov_create if all OSCs are inactive (1751)
6588 - fix import levels when a reconnect happens without a timeout (1597)
6589 - exit early from mds_open if we get a lookup error (1749)
6590 - partial page read at EOF wouldn't wait for disk before sending (1642)
6591 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6592 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6593 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6594 - write/truncate lock inversion (1639)
6595 - Don't auto-load obdclass, portals modules during cleanup (1495)
6596 - fix timestamps from jumping to "now" (1763)
6597 - extra journal assertions (1648)
6598 - add an extra multiunlink test (1771)
6599 - fix read_record/write_record API (1776)
6600 - fix leak of offset_extent, possible incorrect i_size later (1772)
6601 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6602 - Debugging for the kqswnal_get_idle_tx problems (1820)
6603 - Allow recovery to be attempted multiple times (1536)
6604 - Write out MDS last_rcvd file after it is first created (1600)
6605 - Fix tx_descriptor leak in failed transmit situations (1827)
6606 - ext3 journaling fixes for assertion failure after IO error (1871)
6607 - class_export_put() on freed export after completion AST error (1896)
6608 - Fix revalidate looping in VFS (1322)
6609 - Don't access a freed export during MDS_REINT timeout (1521)
6610 - Add open-unlink recovery support on the MDS (1673,1764)
6611 - Return an error if no MDS data was read from last_rcvd (1946)
6612 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6613 - Don't setup a disk device that doesn't match exported UUID (317)
6614 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6615 - avoid committing NULL handle in force close
6616 - local.sh is now a one-stripe LOV configuration
6617 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6618 - fix ext3 htree duplicate directory entry corruption (1516)
6619 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6620 - update to new LOV EA format (2097)
6621 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6623 2003-06-15 Phil Schwan <phil@clusterfs.com>
6626 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6627 - per-import/export recovery handling (958, 931, 959)
6628 - multiple last-rcvd slots, for serving multiple FSes (949)
6629 - connections are again shared between multiple imp/exports (963, 964)
6630 - "umount -f" would hang if any requests needed to be sent (393, 978)
6631 - avoid pinning large req buffer by copying for queued messages (989)
6632 - add "uuid" to "lctl device" command to help upcalls (991)
6633 - "open" RPCs with transnos would confuse recovery counters (1037)
6634 - do proper endian conversion of all wire messages (288, 340, 891)
6635 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6636 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6637 - fix ptlrpc_connection leak in target_handle_connect (1174)
6638 - fix import refcounting bug in OST and MDS cleanup (1134)
6639 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6640 - fix very unlikely obd_types race condition (501)
6641 - remove osc_open hack for echo_client (1187)
6642 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6643 - a failure in read_inode2 leads to deadlock (1139)
6644 - cancel ack-locks as soon as transaction is committed (1072)
6645 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6646 - make sure to commitrw after any preprw to avoid deadlock (1162)
6647 - failing to execute a file in a lustre FS would lock inode (1203)
6648 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6649 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6650 - obd_uuid2tgt would walk off the end of the list (1255)
6651 - on IA64 the osc would give portals incorrect bulk size (1258)
6652 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6653 - fix lock inversion caused by new llite matching code (1282)
6654 - limit the number of dirty pages on a client to 10MB (1286)
6655 - timed out locks were not being corrected cancelled (1289)
6656 - fix O_DIRECT above 4GB on IA-32 (1292)
6657 * major user-visible changes
6658 - fail out/fail over policy now controlled by the upcall (993)
6660 - add OBD_PING to check server availability and failure (954)
6661 - lustre messages are now sent in sending host order (288, 340, 891)
6662 - add eadatalen to MDS getattr reply (340)
6663 - OST read replies may contain second buffer, with per-page status (593)
6665 2003-03-11 Phil Schwan <phil@clusterfs.com>
6668 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6669 - failed open()s could cause deadlock; fixed (867, 869)
6670 - stop cancelling OST locks when files are closed (481)
6671 - overlapping XID spaces caused network corruption (851, 853)
6672 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6673 - setattr_raw added, to do single-RPC, server-side setattrs
6674 - lmc/lconf syntax change for OST UUIDs
6675 - fix crashy race condition between ptlrpc_free_req and osc_close
6676 - don't use request in mdc_enqueue if we hit a timeout (889)
6677 - don't set the inode i_size for regular files from the MDS (896)
6678 - handle out of order completion AST (842)
6679 - don't LBUG if a lock request times out after receiving AST (913)
6680 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6681 - if a bad lock AST arrives, send an error instead of dropping entirely
6682 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6683 - fix leak in bulk IO when only partially completed (899, 900, 926)
6684 - fix O_DIRECT for ia64 (55)
6685 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6686 - C-z after timeout could hang a process forever; fixed (977)
6688 - client-side I/O cache (678, 924, 929, 941, 970)
6690 - READPAGE and SETATTRs which don't take server-side locks get
6693 2003-02-11 Phil Schwan <phil@clusterfs.com>
6696 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6697 - Fix fsfilt_extN_readpage() to read a full page of directory
6698 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6699 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6700 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6701 - Most intent operations were removed, in favour of a new RPC mode
6702 that does a single RPC to the server and bypasses most of the VFS
6703 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6704 - Aggressively cancel local locks on DLM servers
6705 - mds_reint_unlink sends EA to the client if it's the last nlink.
6706 client uses that EA to unlink OST objects.
6707 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6708 - recursive symlinks were fixed (439)
6709 - fixed NULL deref in DEBUG_REQ
6710 - filter_update_lastobjid no longer calls sync, which annoyed extN
6711 - fixed multi-client small-writes to a single file problem (445)
6712 - fixed mtime updates during file writes (607)
6713 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6714 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6715 - fixed Linux OST size reporting problem (444, 656)
6716 - OST now updates object mtime with writes or setattr (607, 619)
6717 - client verifies file size before zeroing page past EOF (445)
6718 - OST now writes last allocated objid to disk with allocation (108)
6719 - LOV on echo now works (409)
6721 - mds_reint_unlink sends a new buffer, with the EA included. this
6722 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6723 set if a regular file was being unlinked, and it was the last link
6724 - use PtlGet from the target for bulk writes (315)
6725 - OST now updates object mtime with writes or setattr (607, 619)
6726 - LDLM now has a grant-time callback to revalidate locked items, if
6728 - Many MDS operations were reorganized to combat race conditions
6730 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6731 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6733 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6736 - Fully reactivate OST imports after reconnection (512, others)
6737 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6738 - More graceful error handling for truncating on dead OST (515)
6739 - Don't error out unless we're actually accessing dead stripes (474)
6740 - Fix garbage sizes when stripes are missing (410)
6741 - LRU counters were broken, causing constant lock purge (433, 432)
6742 - garbage on read from stripes with failed OSTs (441)
6743 - mark OSCs as active before reconnecting during recovery (438)
6744 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6745 - lfind did not preserve OST order in output (443)
6746 - symlinks cause hung clients, incorrect data (439)
6747 - stop dereferencing request after dropping refcount (457)
6748 - don't LASSERT(spin_is_locked) on non-SMP (455)
6749 - fixes for many rename() bugs
6750 - fstat didn't correctly synchronize attributes (399)
6751 - server must handle lock cancellation during blocking AST prep (487)
6752 - bulk descriptors were free()d too soon (511)
6753 - fix paths in lconf, which would load incorrect modules (451, 507)
6754 - fix confusing lconf 'host not found' error message (386)
6755 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6756 - fix race condition in mdc_blocking_ast() for inode access (526)
6757 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6758 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6759 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6760 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6761 - fix lstripe in several ways (kernel side) (527)
6762 - fix request leak in ldlm_cli_enqueue (262)
6763 - incorrect OSC was marked inactive after OST failure
6764 - call mds_fs_cleanup before unmounting filesystem (524)
6765 - fix races between taking ns_lock and ldlm_lock_change_resource
6766 - fix races updating LOV export open file list
6767 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6768 - fix recovery NULL deref in ldlm_cli_cancel_unused
6769 - fix some DLM races by using new hash table for lock handles (419)
6770 - permit the client to specify desired inodes, at replay
6771 - duplicate requests when we queue them for replay reintegration
6772 - fix last_rcvd offset calculation
6773 - sync after each recovered transaction, so we always make progress
6774 - never, not always, ERESTART requests without transnos
6775 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6777 - skip replay if the MDS says that the client is already connected
6778 - don't check for a recovery-enabled export to match lctl's UUID
6779 - don't INC_USE_COUNT for phantom exports
6780 - don't crash when cleaning up phantom exports (567)
6781 - don't double-finish or set replay data for errored mdc_open requests
6782 - abort requests when they time out, so we don't get old replies
6783 - send/receive replies for AST messages again
6784 - if the client says that it doesn't have the lock, cancel it on the
6786 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6787 mark it as destroyed, it will all work out when decref is called
6788 - fix module use counts (22, 581)
6790 - ASTs now expect a reply (server cancels lock on error reply)
6792 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6795 - fix many simultaneous client startup (392)
6796 - fix dentry->d_it clobbering
6797 - credentials weren't being shipped for readdir/getattr operations
6798 - remove invalid assertions triggered during some concurrent MD
6800 - proper Lustre versions added (336, 389)
6801 - fix memory leak for create error case (398)
6802 - fix LOV locking bug that would get cli/srv out of sync
6803 - fix echo client over LOV (409)
6804 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6805 - fix double-O_EXCL intent crash (424)
6806 - avoid sending multiple lock CANCELs (352)
6808 - MDS can do multi-client recovery (modulo bugs in new code)
6810 - many updates, edits, cleanups
6812 2002-11-18 Phil Schwan <phil@clusterfs.com>
6815 - fix null d_it dereference (346)
6816 - fix full OST/dbench hang (333)
6817 - fix permission problem with file removal (286)
6818 - fix removal of OSCs from LOV when they fail
6819 - fix NULL deref during bulk timeout (214)
6820 - fix problems related to multiple filesystems on one MDS (241)
6821 - fixed serious subtle metadata locking bugs
6822 - free locks on clients when inodes are removed due to memory
6824 - fix inode pointer in lock data (285)
6825 - partial support for multiple MDS on a single host (241)
6826 - data locks weren't cancelled at clear_inode time (290, 311)
6827 - intent locks could lead to unbounded lock growth (205)
6828 - added a maximum lock count, an LRU list, and a flusher
6829 - fix multiple rename (365)
6830 - properly abstracted the echo client
6831 - OSC locked 1 byte too many; fixed
6832 - rewrote brw callback code:
6833 - fixed recovery bugs related to LOVs (306)
6834 - fixed too-many-pages-in-one-write crash (191)
6835 - fixed (again) crash in sync_io_timeout (214)
6836 - probably fixed callback-related race (385)
6838 - Add capability to MDS protocol
6839 - LDLM cancellations and callbacks on different portals
6841 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6844 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6845 - defer open object destruction to close time (601981, 138)
6846 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6847 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6848 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6849 - don't call obd_setattr in ll_file_release for destroyed objects
6850 * protocol change to lustre_msg: move |version| and add |flags|
6851 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6852 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6853 * added replay of create, unlink, link and rename operations during
6854 MDS failover; recovery should be much more robust now
6855 * remove failed OSCs from LOVs (only lov_create uses this so far)
6856 * the lustre-HOWTO was brought (more) up to date (582544)
6858 2002-10-23 Phil Schwan <phil@clusterfs.com>
6861 - in-use dentries weren't being reused properly (617851)
6862 - prevent multiple LDLM setup (599178)
6863 - fix LOV size calculations for truncate (617853)
6864 - fix client handling of MDS intent errors (POSIX)
6865 - fix permission bug in lovstripe.c test (624321)
6866 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6867 - truncate past end of file could corrupt data
6868 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6869 - a race in recovery could return ETIMEDOUT to apps (623947)
6870 - building outside the source directory was fixed
6871 * the lustre-HOWTO was brought (more) up to date (582544)
6872 * major progress was made on recovery functionality
6874 2002-10-10 Phil Schwan <phil@clusterfs.com>
6877 - recovery deadlock fix
6878 - rm -rf causes LBUG fix (617817)
6879 - file open by multiple tasks fix (618962)
6880 - directory permissions bugs (602707 and 620007)
6881 - journal_stop fixed with locking (611313)
6882 - O_APPEND failures resolved (618273, perhaps 614459)
6883 - lconf PATH fix (619770)
6884 - IA64 build fix (621450)
6885 - RPC buffer sizes scale with amount of memory
6887 2002-10-01 Phil Schwan <phil@clusterfs.com>
6890 - locks would be cancelled without throwing away data pages,
6891 resulting in inconsistent data (605627)
6892 - inode attributes were not always being refreshed (605627, 612449)
6893 - lconf now continues to cleanup after lctl reports an error
6894 - MDS now enforces user permissions (602707)
6895 - lprocfs cleanup fixed, but not yet enabled (614157)
6896 - fixed infinite server hang, should a client not respond to an AST
6897 - avoid going into recovery if user calls readlink() with a buffer
6898 that's too small (613941)
6899 - AST RPCs no longer require replies (614867) -- this may be changed
6900 - don't crash server if client sends an IOV that's too big (611336)
6901 - fixed lock conversion deadlock (611892)
6902 - fixed the following of symlinks (614622)
6903 * recovery: the server can remove locks from a client that dies, other
6904 clients can make progress
6905 * more extN patch fixes
6906 * compile-time configurable ptlrpc buffer allocations
6908 - collaborative read cache document
6909 - Lustre Lite Performance CDR document-in-progress
6911 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6914 - fix typo in patch-2.4.18
6916 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6919 - clear ptlrpc request each time in handle_incoming_request()
6920 - unlink of files now destroys the object on the OST
6922 2002-09-19 Peter Braam <braam@clusterfs.com>
6924 * add hard link support
6925 * change obdfile creation method
6926 * kernel patch changed
6928 2002-09-19 Peter Braam <braam@clusterfs.com>
6931 - stack overflow bug in extN fixed
6933 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6935 * documentation updates
6936 - add man pages for config tools
6937 - update tests/README to describe testing with new config tools
6938 - finish metadata API descriptions
6939 * bug fixes and cleanups
6940 - statfs workaround for 16TB limit
6941 - LOV stripe allocation improved, can stripe on subset of OSTs
6942 - LOV file size/IO offset was wrong for files > 4GB in size
6943 - object EA data was being dropped, caused files to be unreadable
6944 - memory overflow with non-LOV OST caused memory corruption
6945 - fixed regression tests to work with new config tools, obdfilter
6946 - fixed bug when directory size became larger than 1 block
6947 - fixed bug (for single client case) when PWD was deleted
6948 - invalidate local directory pages when doing intent-based ops
6949 - avoid LDLM oops when lock callback contained bad data
6951 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6953 * documentation updates
6954 * bug fixes and cleanups
6955 - configuration tools
6958 - 64-bit compile warnings
6959 - 64-bit internal statfs data
6961 * test_brw on persistent OST devices
6963 * lprocfs (disabled)
6965 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6967 * documentation updates
6968 * bug fixes and cleanups
6969 * configuration tools
6971 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6974 * many small fixes to 0_5_4
6975 * io/network handling
6976 * thinkos in MDS operations
6978 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6981 * crucial basic fixes to 0.5.3
6982 * IOR, Iozone work over Elan
6985 2002-08-07 Phil Schwan <phil@clusterfs.com>
6986 * version 0_5_3, our first alpha
6987 * we use the new Portals iovs
6988 * documentation updates
6989 * bug fixes and cleanups
6990 * small changes in the DLM wire protocol
6992 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6993 * version 0_5_1 with some initial stability,
6994 * locking on MD and file I/O.
6995 * documentation updates
6996 * several bug fixes since 0.5.0
6997 * small changes in wire protocol
6999 2002-07-18 Phil Schwan <phil@clusterfs.com>
7001 * delivered as Lustre Light Alpha
7002 * fixed a crash after handling invalid MDS requests
7003 * fixed directory pages for architectures with non-4k pages sizes
7005 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
7006 * release version v0_4_4
7007 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
7009 * Fixes a number of interruption problems with OST operations.
7010 * Update documentation for portals header changes
7011 * Move all wire protocol structs/defines to lustre_idl.h
7012 * Fixes symlink length bug.
7013 * Add tcpdump to repository.
7015 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
7016 * release version v0_4_3
7017 * Fixes statfs for inodes on extN.
7018 * Fixes bug in runtests which would delete /etc/hosts.
7019 * Use 64-bit object IDs wherever possible (not into VFS though)
7020 Remove ost_get_info, which is unused by lustre, and out of date.
7022 2002-07-03 Peter Braam <braam@clusterfs.com>
7023 * release version v0_4_2 Fixes a lookup error (type not passed)
7024 * move forward to head of Portals
7025 * move forward to latest Lustre kernel
7027 2002-06-25 Peter Braam <braam@clusterfs.com>
7028 * release version v0_4_1. Hopefully stable on single node use.