3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 4.18.0-372.9.1.el8 (RHEL8.6)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 3.10.0-1062.18.1.el7 (RHEL7.7)
11 3.10.0-1127.19.1.el7 (RHEL7.8)
12 3.10.0-1160.66.1.el7 (RHEL7.9)
13 4.18.0-80.11.2.el8 (RHEL8.0)
14 4.18.0-147.8.1.el8 (RHEL8.1)
15 4.18.0-193.28.1.el8 (RHEL8.2)
16 4.18.0-240.22.1.el8 (RHEL8.3)
17 4.18.0-305.25.1.el8 (RHEL8.4)
18 4.18.0-348.23.1.el8 (RHEL8.5)
19 4.4.120-92.70 (SLES12 SP2)
20 4.4.180-94.100 (SLES12 SP3)
21 4.4.0-131 (Ubuntu 16.04)
22 4.15.0-32 (Ubuntu 18.04)
23 5.4.0-48 (Ubuntu 20.04)
24 vanilla linux 5.4.0 (ZFS + ldiskfs)
25 vanilla linux 5.4.21 (ZFS + ldiskfs)
26 vanilla linux 5.4.136 (ZFS + ldiskfs)
27 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
28 * Client primary kernels built and tested during release cycle:
29 4.18.0-372.9.1.el8 (RHEL8.6)
30 5.4.0-37 (Ubuntu 20.04)
31 5.3.18-59.27 (SLES15 SP3)
32 * Other clients known to build on these kernels at some point (others may also work):
33 3.10.0-862.14.4.el7 (RHEL7.5)
34 3.10.0-957.27.2.el7 (RHEL7.6)
35 4.14.0-49.13.1.el7a (RHEL7.5)
36 4.14.0-115.2.2.el7a (RHEL7.6)
37 3.10.0-1062.18.1.el7 (RHEL7.7)
38 3.10.0-1127.19.1.el7 (RHEL7.8)
39 3.10.0-1160.66.1.el7 (RHEL7.9)
40 4.18.0-80.11.2.el8 (RHEL8.0)
41 4.18.0-147.8.1.el8 (RHEL8.1)
42 4.18.0-193.28.1.el8 (RHEL8.2)
43 4.18.0-240.22.1.el8 (RHEL8.3)
44 4.18.0-305.25.1.el8 (RHEL8.4)
45 4.18.0-348.23.1.el8 (RHEL8.5)
46 4.4.120-92.70 (SLES12 SP2)
47 4.4.180-94.100 (SLES12 SP3)
48 4.12.14-95.48 (SLES12 SP4)
49 4.12.14-122.116 (SLES12 SP5)
50 4.12.14-197.75 (SLES15 SP1)
51 5.3.18-24.96 (SLES15 SP2)
52 4.4.0-131 (Ubuntu 16.04)
53 4.15.0-48 (Ubuntu 18.04)
54 5.8.0-53 (Ubuntu 20.04.2 HWE)
55 5.11.0-31 (Ubuntu 20.04.3 HWE)
56 5.11.0 (vanilla kernel.org)
57 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
58 * Recommended ZFS version: 2.1.2
59 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
60 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
61 stack overflow. For more information, please refer to bugzilla 17630.
62 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
63 the re-exporting nodes to be mounted with "32bitapi" mount option
65 --------------------------------------------------------------------------------
70 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
71 for currently supported client and server kernel versions.
72 * Server primary kernels built and tested during release cycle:
73 4.18.0-348.2.1.el8 (RHEL8.5)
74 * Other server kernels known to build and work at some point (others may also work):
75 3.10.0-862.14.4.el7 (RHEL7.5)
76 3.10.0-957.27.2.el7 (RHEL7.6)
77 3.10.0-1062.18.1.el7 (RHEL7.7)
78 3.10.0-1127.19.1.el7 (RHEL7.8)
79 3.10.0-1160.49.1.el7 (RHEL7.9)
80 4.18.0-80.11.2.el8 (RHEL8.0)
81 4.18.0-147.8.1.el8 (RHEL8.1)
82 4.18.0-193.28.1.el8 (RHEL8.2)
83 4.18.0-240.22.1.el8 (RHEL8.3)
84 4.18.0-305.25.1.el8 (RHEL8.4)
85 4.4.120-92.70 (SLES12 SP2)
86 4.4.180-94.100 (SLES12 SP3)
87 4.4.0-131 (Ubuntu 16.04)
88 4.15.0-32 (Ubuntu 18.04)
89 5.4.0-48 (Ubuntu 20.04)
90 vanilla linux 5.4.0 (ZFS + ldiskfs)
91 vanilla linux 5.4.21 (ZFS + ldiskfs)
92 vanilla linux 5.4.136 (ZFS + ldiskfs)
93 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
94 * Client primary kernels built and tested during release cycle:
95 4.18.0-348.2.1.el8 (RHEL8.5)
96 5.4.0-37 (Ubuntu 20.04)
97 5.3.18-59.27 (SLES15 SP3)
98 * Other clients known to build on these kernels at some point (others may also work):
99 3.10.0-862.14.4.el7 (RHEL7.5)
100 3.10.0-957.27.2.el7 (RHEL7.6)
101 4.14.0-49.13.1.el7a (RHEL7.5)
102 4.14.0-115.2.2.el7a (RHEL7.6)
103 3.10.0-1062.18.1.el7 (RHEL7.7)
104 3.10.0-1127.19.1.el7 (RHEL7.8)
105 3.10.0-1160.49.1.el7 (RHEL7.9)
106 4.18.0-80.11.2.el8 (RHEL8.0)
107 4.18.0-147.8.1.el8 (RHEL8.1)
108 4.18.0-193.28.1.el8 (RHEL8.2)
109 4.18.0-240.22.1.el8 (RHEL8.3)
110 4.18.0-305.25.1.el8 (RHEL8.4)
111 4.4.120-92.70 (SLES12 SP2)
112 4.4.180-94.100 (SLES12 SP3)
113 4.12.14-95.48 (SLES12 SP4)
114 4.12.14-122.91 (SLES12 SP5)
115 4.12.14-197.75 (SLES15 SP1)
116 5.3.18-24.96 (SLES15 SP2)
117 4.4.0-131 (Ubuntu 16.04)
118 4.15.0-48 (Ubuntu 18.04)
119 5.8.0-53 (Ubuntu 20.04.2 HWE)
120 5.11.0-31 (Ubuntu 20.04.3 HWE)
121 5.11.0 (vanilla kernel.org)
122 * Recommended e2fsprogs version: 1.46.2.wc5 or newer
123 * Recommended ZFS version: 2.0.7
124 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
125 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
126 stack overflow. For more information, please refer to bugzilla 17630.
127 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
128 the re-exporting nodes to be mounted with "32bitapi" mount option
130 --------------------------------------------------------------------------------
134 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
135 for currently supported client and server kernel versions.
136 * Server primary kernels built and tested during release cycle:
137 3.10.0-1062.18.1.el7 (RHEL7.7)
138 3.10.0-1127.19.1.el7 (RHEL7.8)
139 3.10.0-1160.6.1.el7 (RHEL7.9)
140 * Other server kernels known to build and work at some point (others may also work):
141 3.10.0-862.14.4.el7 (RHEL7.5)
142 3.10.0-957.27.2.el7 (RHEL7.6)
143 4.18.0-80.11.2.el8 (RHEL8.0)
144 4.18.0-147.8.1.el8 (RHEL8.1)
145 4.18.0-193.28.1.el8 (RHEL8.2)
146 4.18.0-240.1.1.el8 (RHEL8.3)
147 4.4.120-92.70 (SLES12 SP2)
148 4.4.180-94.100 (SLES12 SP3)
149 4.4.0-131 (Ubuntu 16.04)
150 4.15.0-32 (Ubuntu 18.04)
151 5.4.0-48 (Ubuntu 20.04)
152 vanilla linux 5.4.0 (ZFS + ldiskfs)
153 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
154 * Client primary kernels built and tested during release cycle:
155 3.10.0-1062.18.1.el7 (RHEL7.7)
156 3.10.0-1127.19.1.el7 (RHEL7.8)
157 3.10.0-1160.6.1.el7 (RHEL7.9)
158 4.12.14-95.48 (SLES12 SP4)
159 4.12.14-122.20 (SLES12 SP5)
160 4.12.14-197.75 (SLES15 SP1)
161 5.3.18-24.24 (SLES15 SP2)
162 4.15.0-48 (Ubuntu 18.04)
163 * Other clients known to build on these kernels at some point (others may also work):
164 3.10.0-862.14.4.el7 (RHEL7.5)
165 3.10.0-957.27.2.el7 (RHEL7.6)
166 4.14.0-49.13.1.el7a (RHEL7.5)
167 4.14.0-115.2.2.el7a (RHEL7.6)
168 4.18.0-80.11.2.el8 (RHEL8.0)
169 4.18.0-147.8.1.el8 (RHEL8.1)
170 4.18.0-193.28.1.el8 (RHEL8.2)
171 4.18.0-240.1.1.el8 (RHEL8.3)
172 4.4.120-92.70 (SLES12 SP2)
173 4.4.180-94.100 (SLES12 SP3)
174 4.4.0-131 (Ubuntu 16.04)
175 5.4.0-37 (Ubuntu 20.04)
176 5.4.0 (vanilla kernel.org)
177 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
178 * Recommended ZFS version: 2.0.0
179 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
180 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
181 stack overflow. For more information, please refer to bugzilla 17630.
182 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
183 the re-exporting nodes to be mounted with "32bitapi" mount option
185 --------------------------------------------------------------------------------
189 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
190 for currently supported client and server kernel versions.
191 * Server primary kernels built and tested during release cycle:
192 3.10.0-1062.1.1.el7 (RHEL7.7)
193 * Other server kernels known to build and work at some point (others may also work):
194 3.10.0-862.14.4.el7 (RHEL7.5)
195 3.10.0-957.27.2.el7 (RHEL7.6)
196 4.18.0-80.7.1.el8 (RHEL8.0)
197 4.4.120-92.70 (SLES12 SP2)
198 4.4.180-94.100 (SLES12 SP3)
199 4.4.0-131 (Ubuntu 16.04)
200 4.15.0-32 (Ubuntu 18.04)
201 vanilla linux 4.6.7 (ZFS only)
202 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
203 * Client primary unpatched kernels built and tested during release cycle:
204 3.10.0-1062.1.1.el7 (RHEL7.7)
205 4.12.14-95.29 (SLES12 SP4)
206 4.15.0-48 (Ubuntu 18.04)
207 * Other clients known to build on unpatched kernels at some point (others may also work):
208 3.10.0-862.14.4.el7 (RHEL7.5)
209 3.10.0-957.27.2.el7 (RHEL7.6)
210 4.14.0-49.13.1.el7a (RHEL7.5)
211 4.14.0-115.2.2.el7a (RHEL7.6)
212 4.18.0-80.7.1.el8 (RHEL8.0)
213 4.4.120-92.70 (SLES12 SP2)
214 4.4.180-94.100 (SLES12 SP3)
215 4.4.0-131 (Ubuntu 16.04)
216 vanilla linux 4.15.0 (ZFS only)
217 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
218 * Recommended ZFS / SPL version: 0.7.13
219 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
220 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
221 stack overflow. For more information, please refer to bugzilla 17630.
222 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
223 the re-exporting nodes to be mounted with "32bitapi" mount option
224 * Note for Ubuntu 14.04.5:
225 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
226 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
228 --------------------------------------------------------------------------------
232 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
233 for currently supported client and server kernel versions.
234 * Server known to build on the following kernels (others may also work):
235 2.6.32-696.18.7.el6 (RHEL6.9)
236 3.10.0-862.14.4.el7 (RHEL7.5)
237 3.10.0-957.el7 (RHEL7.6)
238 4.4.120-92.70 (SLES12 SP2)
239 4.4.162-94.69 (SLES12 SP3)
240 4.4.0-131 (Ubuntu 16.04)
241 4.15.0-32 (Ubuntu 18.04)
242 vanilla linux 4.6.7 (ZFS only)
243 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
244 * Client known to build on unpatched kernels (others may also work):
245 2.6.32-696.18.7.el6 (RHEL6.9)
246 2.6.32-754.6.3.el6 (RHEL6.10)
247 3.10.0-862.14.4.el7 (RHEL7.5)
248 3.10.0-957.el7 (RHEL7.6)
249 4.14.0-49.13.1.el7a (RHEL7.5)
250 4.14.0-115.2.2.el7a (RHEL7.6)
251 4.4.120-92.70 (SLES12 SP2)
252 4.4.162-94.69 (SLES12 SP3)
253 4.4.0-131 (Ubuntu 16.04)
254 4.15.0-32 (Ubuntu 18.04)
255 vanilla linux 4.15.0 (ZFS only)
256 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
257 * Recommended ZFS / SPL version: 0.7.9
258 * Tested with ZFS / SPL version: 0.7.9
259 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
260 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
261 stack overflow. For more information, please refer to bugzilla 17630.
262 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
263 the re-exporting nodes to be mounted with "32bitapi" mount option
264 * Note for Ubuntu 14.04.5:
265 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
266 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
268 --------------------------------------------------------------------------------
270 2018-04-03 Intel Corporation
272 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
273 for currently supported client and server kernel versions.
274 * Server known to build on patched kernels:
275 2.6.32-431.29.2.el6 (RHEL6.5)
276 2.6.32-504.30.3.el6 (RHEL6.6)
277 2.6.32-573.26.1.el6 (RHEL6.7)
278 2.6.32-642.15.1.el6 (RHEL6.8)
279 2.6.32-696.18.7.el6 (RHEL6.9)
280 3.10.0-693.21.1.el7 (RHEL7.4)
281 3.0.101-0.47.71 (SLES11 SP3)
282 3.0.101-107 (SLES11 SP4)
283 3.12.74-60.64.40 (SLES12 SP1)
284 4.4.114-92.64 (SLES12 SP2)
285 4.4.114-94.11 (SLES12 SP3)
286 3.13.0-101 (Ubuntu 14.04, ZFS only)
287 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
288 4.4.0-116 (Ubuntu 16.04)
289 vanilla linux 4.6.7 (ZFS only)
290 * Client known to build on unpatched kernels:
291 2.6.32-431.29.2.el6 (RHEL6.5)
292 2.6.32-504.30.3.el6 (RHEL6.6)
293 2.6.32-573.26.1.el6 (RHEL6.7)
294 2.6.32-642.15.1.el6 (RHEL6.8)
295 2.6.32-696.18.7.el6 (RHEL6.9)
296 3.10.0-693.21.1.el7 (RHEL7.4)
297 3.0.101-0.47.71 (SLES11 SP3)
298 3.0.101-107 (SLES11 SP4)
299 3.12.74-60.64.40 (SLES12 SP1)
300 4.4.114-92.64 (SLES12 SP2)
301 4.4.114-94.11 (SLES12 SP3)
302 3.13.0-101 (Ubuntu 14.04)
303 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
304 4.4.0-116 (Ubuntu 16.04)
306 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
307 * Recommended ZFS / SPL version: 0.7.6
308 * Tested with ZFS / SPL version: 0.7.6
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
314 * Note for Ubuntu 14.04.5:
315 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
316 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
318 --------------------------------------------------------------------------------
320 2017-07-05 Intel Corporation
322 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
323 for currently supported client and server kernel versions.
324 * Server known to build on patched kernels:
325 2.6.32-431.29.2.el6 (RHEL6.5)
326 2.6.32-504.30.3.el6 (RHEL6.6)
327 2.6.32-573.26.1.el6 (RHEL6.7)
328 2.6.32-642.15.1.el6 (RHEL6.8)
329 2.6.32-696.3.1.el6 (RHEL6.9)
330 3.10.0-514.21.1.el7 (RHEL7.3)
331 3.0.101-0.47.71 (SLES11 SP3)
332 3.0.101-100 (SLES11 SP4)
333 3.12.74-60.64.40 (SLES12 SP1)
334 4.4.59-92.17 (SLES12 SP2)
335 vanilla linux 4.6.7 (ZFS only)
336 * Client known to build on unpatched kernels:
337 2.6.32-431.29.2.el6 (RHEL6.5)
338 2.6.32-504.30.3.el6 (RHEL6.6)
339 2.6.32-573.26.1.el6 (RHEL6.7)
340 2.6.32-642.15.1.el6 (RHEL6.8)
341 2.6.32-696.3.1.el6 (RHEL6.9)
342 3.10.0-514.21.1.el7 (RHEL7.3)
343 3.0.101-0.47.71 (SLES11 SP3)
344 3.0.101-100 (SLES11 SP4)
345 3.12.74-60.64.40 (SLES12 SP1)
346 4.4.59-92.17 (SLES12 SP2)
348 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
349 * Recommended ZFS / SPL version: 0.6.5.9
350 * Tested with ZFS / SPL version: 0.6.5.9
351 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
352 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
353 stack overflow. For more information, please refer to bugzilla 17630.
354 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
355 the re-exporting nodes to be mounted with "32bitapi" mount option
357 --------------------------------------------------------------------------------
359 12-07-2016 Intel Corporation
361 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
362 for currently supported client and server kernel versions.
363 * Server known to build on patched kernels:
364 2.6.32-431.29.2.el6 (RHEL6.5)
365 2.6.32-504.30.3.el6 (RHEL6.6)
366 2.6.32-573.26.1.el6 (RHEL6.7)
367 2.6.32-642.11.1.el6 (RHEL6.8)
368 3.10.0-514.el7 (RHEL7.3)
369 3.0.101-0.47.71 (SLES11 SP3)
370 3.0.101-80 (SLES11 SP4)
371 3.12.62-60.64.8 (SLES12 SP1)
372 vanilla linux 4.5.7 (ZFS only)
373 * Client known to build on unpatched kernels:
374 2.6.32-431.29.2.el6 (RHEL6.5)
375 2.6.32-504.30.3.el6 (RHEL6.6)
376 2.6.32-573.26.1.el6 (RHEL6.7)
377 2.6.32-642.11.1.el6 (RHEL6.8)
378 3.10.0-514.el7 (RHEL7.3)
379 3.0.101-0.47.71 (SLES11 SP3)
380 3.0.101-80 (SLES11 SP4)
381 3.12.62-60.64.8 (SLES12 SP1)
383 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
384 * Recommended ZFS / SPL version: 0.6.5.8
385 * Tested with ZFS / SPL version: 0.6.5.8
386 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
387 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
388 stack overflow. For more information, please refer to bugzilla 17630.
389 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
390 the re-exporting nodes to be mounted with "32bitapi" mount option
392 --------------------------------------------------------------------------------
394 02-29-2016 Intel Corporation
396 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
397 for currently supported client and server kernel versions.
398 * Server known to build on patched kernels:
399 2.6.32-431.29.2.el6 (RHEL6.5)
400 2.6.32-504.30.3.el6 (RHEL6.6)
401 2.6.32-573.12.1.el6 (RHEL6.7)
402 3.10.0-327.3.1.el7 (RHEL7.2)
403 3.0.101-0.47.71 (SLES11 SP3)
404 3.0.101-68 (SLES11 SP4)
405 vanilla linux 4.2.1 (ZFS only)
406 * Client known to build on unpatched kernels:
407 2.6.32-431.29.2.el6 (RHEL6.5)
408 2.6.32-504.30.3.el6 (RHEL6.6)
409 2.6.32-573.12.1.el6 (RHEL6.7)
410 3.10.0-327.3.1.el7 (RHEL7.2)
411 3.0.101-0.47.71 (SLES11 SP3)
412 3.0.101-68 (SLES11 SP4)
415 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
416 * Recommended ZFS / SPL version: 0.6.4.2
417 * Tested with ZFS / SPL version: 0.6.4.2
418 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
419 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
420 stack overflow. For more information, please refer to bugzilla 17630.
421 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
422 the re-exporting nodes to be mounted with "32bitapi" mount option
424 --------------------------------------------------------------------------------
426 03-10-2015 Intel Corporation
428 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
429 for currently supported client and server kernel versions.
430 * Server known to build on patched kernels:
431 2.6.32-431.29.2.el6 (RHEL6.5)
432 2.6.32-504.8.1.el6 (RHEL6.6)
433 3.0.101-0.46 (SLES11 SP3)
434 * Client known to build on unpatched kernels:
435 2.6.32-431.29.2.el6 (RHEL6.5)
436 2.6.32-504.8.1.el6 (RHEL6.6)
437 3.10.0-123.20.1.el7 (RHEL7)
438 3.0.101-0.46 (SLES11 SP3)
439 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
440 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
441 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
442 stack overflow. For more information, please refer to bugzilla 17630.
443 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
444 the re-exporting nodes to be mounted with "32bitapi" mount option
446 Severity : enhancement
448 Description: control OST-index in IDIF via ROCOMPAT flag.
449 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
450 last_rcvd file. For new formatted OST device, it will be auto set;
451 for the case of upgrading from old OST device, you can enable it
452 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
453 enabled, for new created OST-object, its IDIF-in-LMA will contain
454 the OST-index; for the existing OST-object, the OSD will convert
455 old format IDIF as new format IDIF with OST-index stored in the
456 LMA EA when accessing such OST-object or via OI scrub. Once such
457 flag is enabled, it cannot be reverted back, so the system cannot
458 be downgraded to the orignal incompatible version.
460 --------------------------------------------------------------------------------
462 07-30-2014 Intel Corporation
464 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
465 for currently supported client and server kernel versions.
466 * Server known to build on patched kernels:
467 2.6.32-431.20.3.el6 (RHEL6.5)
468 3.0.101-0.31 (SLES11 SP3)
469 * Client known to build on unpatched kernels:
470 2.6.32-431.20.3.el6 (RHEL6.5)
471 3.0.101-0.31 (SLES11 SP3)
474 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
475 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
476 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
477 stack overflow. For more information, please refer to bugzilla 17630.
478 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
479 the re-exporting nodes to be mounted with "32bitapi" mount option
481 --------------------------------------------------------------------------------
483 10-31-2013 Intel Corporation
485 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
486 * Server builds for kernels:
487 2.6.32-358.18.1.el6 (RHEL6.4)
488 * Client builds for unpatched kernels:
489 2.6.32-358.18.1.el6 (RHEL6.4)
490 3.0.92-0.8 (SLES11 SP3)
492 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
493 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
494 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
495 stack overflow. For more information, please refer to bugzilla 17630.
496 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
497 the re-exporting nodes to be mounted with "32bitapi" mount option
499 --------------------------------------------------------------------------------
501 05-30-2013 Intel Corporation
503 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
504 * Server builds for kernels:
505 2.6.32-358.6.2.el6 (RHEL6.4)
506 * Client builds for unpatched kernels:
507 2.6.32-358.6.2.el6 (RHEL6.4)
508 3.0.74-0.6.6 (SLES11 SP2)
510 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
511 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
512 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
513 stack overflow. For more information, please refer to bugzilla 17630.
514 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
515 the re-exporting nodes to be mounted with "32bitapi" mount option
517 --------------------------------------------------------------------------------
519 9-30-2012 Whamcloud, Inc.
521 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
522 * Server support for kernels:
523 2.6.32-279.5.1.el6 (RHEL6)
524 * Client support for unpatched kernels:
525 2.6.32-279.5.1.el6 (RHEL6)
527 2.6.18-238.19.1.el5 (RHEL5)
528 2.6.16.60-0.69.1 (SLES10)
529 * Recommended e2fsprogs version: 1.41.90.wc4
530 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
531 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
532 stack overflow. For more information, please refer to bugzilla 17630.
533 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
534 the re-exporting nodes to be mounted with "32bitapi" mount option
536 Severity : enhancement
538 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
540 Severity : enhancement
542 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
544 Severity : enhancement
546 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
548 Severity : enhancement
550 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
552 Severity : enhancement
554 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
556 --------------------------------------------------------------------------------
558 04-30-2012 Whamcloud, Inc.
560 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
561 * Server support for kernels:
562 2.6.32-220.4.2.el6 (RHEL6)
563 * Client support for unpatched kernels:
564 2.6.32-220.4.2.el6 (RHEL6)
566 2.6.18-238.19.1.el5 (RHEL5)
567 2.6.16.60-0.69.1 (SLES10)
568 * Recommended e2fsprogs version: 1.41.90.wc4
569 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
570 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
571 stack overflow. For more information, please refer to bugzilla 17630.
572 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
573 the re-exporting nodes to be mounted with "32bitapi" mount option
575 Severity : enhancement
577 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
579 Severity : enhancement
581 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
583 Severity : enhancement
585 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
587 Severity : enhancement
589 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
591 --------------------------------------------------------------------------------
593 09-30-2011 Whamcloud, Inc.
595 * Server support for kernels:
596 2.6.18-238.19.1.el5 (RHEL5)
597 2.6.32-131.6.1.el6 (RHEL6)
598 * Client support for unpatched kernels:
599 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
600 2.6.18-238.19.1.el5 (RHEL5)
601 2.6.32-131.6.1.el6 (RHEL6)
602 2.6.29.4-167.fc11 (FC11)
603 2.6.16.60-0.69.1 (SLES10)
605 * Recommended e2fsprogs version: 1.41.90.wc3
606 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
607 Lustre filesystem with 4K stack may cause a stack overflow. For more
608 information, please refer to bugzilla 17630.
609 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
610 mounted with "32bitapi" mount option
612 Severity : enhancement
613 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
615 Severity : enhancement
616 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
618 Severity : enhancement
619 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
622 Frequency : only with SLES10
624 Description: Use OFED "KMP" provided by Novell
625 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
626 Lustre is now built against this package. That means you need to
627 install the ofed-kmp package from Novell for the patchless client.
628 Note that the ofed-kmp that Novell ships may not exactly match the
629 kernel version but should still be compatible.
631 Severity : enhancement
633 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
634 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
636 --------------------------------------------------------------------------------
638 2010-07-15 Oracle, Inc.
640 * Server support for kernels:
641 2.6.18-164.11.1.el5 (RHEL 5)
642 2.6.18-164.11.1.0.1.el5 (OEL 5)
643 * Client support for unpatched kernels:
644 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
645 2.6.18-164.11.1.el5 (RHEL 5),
646 2.6.18-164.11.1.0.1.el5 (OEL 5)
647 2.6.16.60-0.42.8 (SLES 10),
649 2.6.29.4-167.fc11 (FC11)
650 * Recommended e2fsprogs version: 1.41.10-sun2
652 Severity : enhancement
654 Description: Distribute lfs_migrate script with lustre
655 Details : The lfs_migrate mentioned in the manual is now included in the
656 lustre tree in lustre/scripts and will distributed in lustre RPMs.
658 Severity : enhancement
660 Description: Failover nid list added to import proc file.
662 Severity : enhancement
664 Description: "som_preview" mount option required to use SOM
665 Details : The clients will need to use som_preview mount option
669 Description: Lustre does not do 1MB IOs to HW RAID
670 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
672 Severity : enhancement
674 Description: lfs quota output is not very convenient for awk/sed-parsing
675 Details : Some positions in lfs quota output table could be empty or
676 non-empty which made it hard to parse it with scripts, now
677 a dash is put instead of space where there is not supposed
682 Description: Update to OFED 1.5.1
683 Details : OFED 1.5.1 will be used in environments that don't use
684 vendor supplied OFED.
688 Description: Remove lnet. prefix from lctl params display
689 Details : Also renamed lnet.memused to lnet_memused to
694 Description: Use conf_param -d to remove permanant settings.
698 Description: Add list_param -R
699 Details : lctl list_param now supports recursive mode.
702 Bugzilla : 21829, 22312
703 Description: Fix llobdstat.
704 Details : 1. Now updating the write_bytes counter
705 2. Port script fixes from 1.8.
709 Description: decrease the usage of memory on clients.
710 Details : 1. On clients, recycle dentries and inodes unused.
711 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
716 Description: llobdstat fix and enhancement
717 Details : add a counter to set a limit to how many samples will be returned
718 fix a wildcard in the path to limit to obdfilter stats only
722 Description: lfs check only allowed for root
723 Details : Code cleanups and added a regresion test of non-root use of
728 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
729 Details : Fix hr->hr_index overload.
733 Description: Fix FIEMAP functionality.
737 Description: FC11 patchless client support added
741 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
742 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
746 Description: New acc-sm script and various test-framework enhancments.
750 Description: Previously landed patch was reverted.
754 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
758 Description: Add ldiskfs maxdirsize mount option.
762 Description: Fix mkfs.lustre support for 16TB LUNs.
766 Description: Print mmp_check_interval
767 Details : print mmp_check_interval and make it possible to abort
768 mount operation in case it takes too long.
772 Description: Add cascading_rw to lustre mpi tests.
776 Description: Store SOM data in EA.
777 Details : Also several SOM related cleanups.
781 Description: Add refcount to md_open_data
785 Description: Use refcount on lov_request_set to prevent memory corruption.
789 Description: Statahead: use parent dir's i_mutex for synchronization.
793 Description: loadgen improvements.
797 Description: Enable peer health detection in ksocklnd and ko2iblnd.
801 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
805 Description: Fix race on obd_devcie:md_stats
809 Description: Fix SLES 11 client builds.
810 Details : Resolved uninitalized pointer gcc warnings.
814 Description: Allow non-root access for "lfs check".
815 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
819 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
821 Severity : enhancement
823 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
824 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
828 Description: Sanity test_27p fails.
829 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
830 Sanity tests 27 were cleaned up and improved.
834 Description: Unify req format on client/servers
835 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
836 buffers and swabbers approach..
841 Description: drop join file stuff.
843 Severity : enhancement
845 Description: Add 2.6.27 and SLES11 (patchless client) support
849 Description: Fix in ptlrpc_expire_one_request() to print the signed time
852 Severity : enhancement
854 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
857 Severity : enhancement
859 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
861 Severity : enhancement
863 Description: correctly handle big reply message.
864 Details : send LNet event if reply is bigger than buffer and adjust this buffer
869 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
871 Severity : enhancement
873 Description: Add support for OEL5.
877 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
879 Severity : enhancement
881 Description: "weak-modules" support
882 Details : Implement "weak-modules" support which enables kernel modules
883 to be used with any kernel that implements the same kABI. In
884 order to achieve this modules are now installed in
885 /lib/modules/$(uname -r)/updates/kernel on all distributions.
889 Description: client could not reconnect to OST because of an active request.
890 Details : The client now retries to reconnect to the same server, if a
891 connect request failed with EBUSY or -EAGAIN. Test case 35b
892 conf-sanity.sh has been created.
896 Description: Don't allow make backward step on assignin osc next id.
897 Details : race between allocation next id and ll_sync thread can be cause
898 of set wrong osc next id and can be kill valid ost objects.
902 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
906 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
910 Description: Error handling in osc_statfs_interpret() has been improved.
911 Details : Check in osc_statfs_interpret() for EBADR.
915 Description: Do not send statfs() requests to OSTs disabled by administrator.
916 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
918 Severity : enhancement
920 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
922 Severity : enhancement
924 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
926 Severity : enhancement
928 Description: Enhancement for lfs(1) command to use numeric uid/gid.
932 Description: Improved cleanup procedure for conf-sanity test_46a.
936 Description: Do not start bulk transfer, if client tried to reconnect.
937 Details : Do not start bulk transfer, if client tried to reconnect,
938 but failed due to active rpcs. Check for exp_abort_active_req
939 has been implemented.
943 Description: Prevent inconsistences between linux and lustre mount structures.
944 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
945 Make the sleep interruptible.
947 Severity : enhancement
949 Description: provide server to client comms path
950 Details : server to client communications path via new LDLM_SET_INFO rpc
954 Description: truncate starts GFP_FS allocation under transaction causing deadlock
955 Details : ldiskfs_truncate calls grab_cache_page which may start page
956 allocation under an open transaction. This may lead to
957 calling prune_icache with consequent lustre reentrance.
961 Description: send correctly lsm on open replay
962 Details : MDS is trust to LSM size on replay open, but client can set wrong size
967 Description: enable client interface failover
968 Details : When a child reconnects from another NID, properly update export
969 nid hash position and ldlm reverse import.
971 Severity : enhancement
973 Description: Communicate OST degraded/readonly state via statfs to MDS
974 Details : Flags in the statfs returned from OSTs indicate whether the
975 OST is in a degraded RAID state, or if the filesystem has
976 turned read-only after a filesystem error is detected.
981 Description: don't panic if EPROTO was hit when reading symlink
982 Details : correctly handling request reference in error cases.
984 Severity : enhancement
986 Description: Add LustreNetLink, a kernel-userspace communcation path.
988 Severity : enhancement
990 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
995 Description: access to llog context before init.
996 Details : move handling CATALOGS file at osc layer and forbid access to llog
1000 Frequency : rare, only when client is writing to OST on local node
1002 Description: Avoid deadlock for local client writes
1003 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
1004 memory freeing context. This allows OST threads to set the
1005 PF_MEMALLOC flag on task structures in order to allocate memory
1006 from reserved pools and complete IO.
1007 Use GFP_HIGHUSER for OST allocations for non-local client writes,
1008 so that the OST threads generate memory pressure and allow
1009 inactive pages to be reclaimed.
1011 Severity : enhancement
1013 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
1018 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
1019 Details : move ldlm namespace creation in setup phase to avoid grab
1020 _lprocfs_lock with cli_sem held.
1024 Description: Temporarily disable grant shrink.
1025 Details : Disable the feature for debugging.
1028 Frequency : only when formatting test filesystems
1030 Description: Unable to run several mkfs.lustre on loop devices at the same time
1031 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1032 formatting. The solution is to proper handle the error.
1034 Severity : enhancement
1036 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1038 Severity : enhancement
1040 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1045 Description: resolve race between obd_disconnect and class_disconnect_exports
1046 Details : if obd_disconnect will be called to already disconnected export he
1047 forget release one reference and osc module can't unloaded.
1049 Severity : enhancement
1051 Description: Allow tuning service thread via /proc
1052 Details : For each service a new
1053 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1054 created that can be used to set min/max thread counts, and get the
1055 current number of running threads.
1060 Descriptoin: don't return error if have partially created objects for file.
1061 Details : lov_update_create_set uses set->set_success as index for created
1062 objects, so if some requests will be failed, they will have hole
1063 at end of array and qos_shrink_lsm for allocate correct lsm.
1065 Severity : enhancement
1067 Description: Update OFED support to 1.4.1
1072 Description: Reduce small size read RPC
1073 Details : Set read-ahead limit for every file and only do read-ahead when
1074 available read-ahead pages are bigger than 1M to avoid small size
1077 Severity : enhancement
1079 Description: move AT tunable parameters for more consistent usage
1080 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1082 Severity : enhancement
1084 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1085 Details : allow skip disconnected ost for send statfs request and hide error
1091 Description: fix racy locking of mballoc block bitmaps causing BUG
1092 Details : The locking of the mballoc buddy bitmap and the in-memory
1093 block bitmap was using two different spin locks in some
1094 cases. This made it possible to incorrectly access the
1095 mballoc bitmap while another process was modifying it,
1096 causing a sanity assertion to fail. While no on-disk corruption
1097 was reported, there was some risk of this happening.
1100 Frequency : rare, on llog test 6
1102 Descriptoin: don't allow connect to already connected import
1103 Details : allowing connect to already connected import is hide connecting problem.
1108 Descriptoin: allow kill process which wait statahead result
1109 Details : for some reasons 'ls' can stick in waiting result from statahead,
1110 in this case need way for kill this process.
1112 Severity : enhancement
1114 Description: Add state history info file, enhance import info file
1115 Details : Track import connection state changes in a new osc/mdc proc file;
1116 add overview-type data to the osc/mdc import proc file.
1118 Severity : enhancement
1120 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1122 Severity : enhancement
1124 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1129 Descriptoin: OST grant too much space to client even there are not enough space.
1130 Details : Client will shrink its grant cache to OST if there are no write
1131 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1132 this grant cache if there are already not enough avaible space
1133 (left_space < total_clients * 32M).
1136 Frequency : start MDS on uncleanly shutdowned MDS device
1138 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1139 Details : stay in waiting mds<>ost recovery finished produce random bugs
1140 due race between two ll_sync thread for one lov target. send
1141 ACTIVATE event only if connect realy finished and import have
1145 Frequency : rare, connect and disconnect target at same time
1147 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1148 Details : don't call obd_disconnect under lov_lock. this long time
1149 operation and can block ptlrpcd which answer to connect request.
1154 Descriptoin: don't lose wakeup for imp_recovery_waitq
1155 Details : recover_import_no_retry or invalidate_import and import_close can
1156 both sleep on imp_recovery_waitq, but we was send only one wakeup
1160 Frequency : always with long access acl
1162 Descriptoin: mds can't pack reply with long acl.
1163 Details : mds don't control size of acl but they limited by reint/getattr
1166 Severity : enhancement
1168 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1170 Severity : enhancement
1172 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1175 Frequency : start MDS on uncleanly shutdowned MDS device
1177 Descriptoin: aborting recovery hang on MDS
1178 Details : don't throttle destroy RPCs for the MDT.
1181 Frequency : on remount
1183 Description: external journal device not working after the remount
1184 Details : clear dev_rdonly flag for external journal devices in
1190 Description: shutdown vs evict race
1191 Details : client_disconnect_export vs connect request race.
1192 if client will evicted at this time - we start invalidate
1193 thread without referece to import and import can be freed
1197 Frequency : rare, need acl's on inode.
1199 Description: client can't handle ost additional correctly
1200 Details : if ost was added after client connected to mds client can have
1201 hit lnet_try_match_md ... to big messages to wide striped files.
1202 in this case need teach client to handle config events about add
1203 lov target and update client max ea size at that event.
1205 Severity : enhancement
1207 Description: Changelogs
1208 Details : Changelogs are a lightweight mechanism to track filesystem
1209 metadata and namespace changes. The changelog is recorded
1210 permanently on the MDTs, and is periodically "consumed" / purged
1211 when records are no longer needed.
1213 Severity : enhancement
1215 Description: compact fld format with extents
1216 Details : Store range of seq rather than every seq in FLD. Seq
1217 controller update FLD rather than clients. In Case of CMD, mdt0
1218 has FLD, all other metadata server act as non persistent proxy
1219 for FLD queries and cache fld entries in fld cache.
1224 Description: don't skip ost target if they assigned to file
1225 Details : Drop slow OSCs if we can, but not for requested start idx.
1226 This means "if OSC is slow and it is not the requested
1227 start OST, then it can be skipped, otherwise skip it only
1228 if it is inactive/recovering/out-of-space.
1232 Description: more cleanup in mds_lov
1233 Details : not send LOV EA under replay, we can't know about they size at this
1234 time. Don't allow client connect to mds before any ost connected,
1235 for avoid problems with LOV EA size and returning EIO to client.
1237 Severity : enhancement
1239 Description: Interoperability at server side (Disk interoperability)
1241 Severity : enhancement
1243 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1245 Severity : enhancement
1247 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1249 Severity : enhancement
1251 Description: New client IO stack (CLIO).
1253 Severity : enhancement
1255 Description: Commit on sharing. Eliminate inter-client dependencies between
1256 uncommitted transactions by doing transaction commits.
1257 Thereby clients may recovery independently.
1260 Frequency : Create a symlink file with a very long name
1262 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1263 Details : If there is no extra space in the request for early cancels,
1264 ldlm_req_handles_avail() returns 0 instead of a negative value.
1266 Severity : enhancement
1268 Description: Add /proc entry for import status
1269 Details : The mdc, osc, and mgc import directories now have
1270 an import directory that contains useful import data for debugging
1271 connection problems.
1273 Severity : enhancement
1275 Description: Re-disable certain /proc logging
1276 Details : Enable and disable client's offset_stats, extents_stats and
1277 extents_stats_per_process stats logging on the fly.
1280 Frequency : Only on FC kernels 2.6.22+
1282 Description: oops in statahead
1283 Details : Do not drop reference count for the dentry from VFS when lookup,
1284 VFS will do that by itself.
1286 Severity : enhancement
1288 Description: Generic /proc file permissions
1289 Details : Set /Proc file permissions in a more generic way to enable non-
1290 root users operate on some /proc files.
1294 Description: Hitting mdc_commit_close() ASSERTION
1295 Details : Properly handle request reference release in
1296 ll_release_openhandle().
1300 Description: quota recovery deadlock during mds failover
1301 Details : This patch includes att18982, att18236, att18237 in bz14840.
1303 1. fix osts hang when mds does failover with quotaon
1304 2. prevent watchdog storm when osts threads wait for the
1309 Frequency : only patchless client
1310 Description: add workaround for race between add/remove dentry from hash
1312 Severity : enhancement
1314 Description: Allow OST glimpses to return PW locks
1318 Description: LBUG when llog conf file is full
1319 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1323 Description: Prevent import from entering FULL state when server in recovery
1327 Description: service mount cannot take device name with ":"
1328 Details : Only when device name contains ":/" will mount treat it as
1334 Description: replace ptlrpcd with the statahead thread to interpret the async
1335 statahead RPC callback
1339 Frequency : on recovery
1340 Description: I/O failures after umount during fail back
1341 Details : if client reconnected to restarted server we need join to recovery
1342 instead of find server handler is changed and process self
1343 eviction with cancel all locks.
1345 Severity : enhancement
1347 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1349 Severity : enhancement
1351 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1353 Severity : enhancement
1355 Description: Upcall on Lustre log has been dumped
1356 Details : Allow for a user mode script to be called once a Lustre log has
1357 been dumped. It passes the filename of the dumped log to the
1358 script, the location of the script can be specified via
1359 /proc/sys/lnet/debug_log_upcall.
1364 Description: avoid idr_remove called for id which is not allocated.
1365 Details : Move assigment s_dev for clustered nfs to end of initialization,
1366 to avoid problem with error handling.
1371 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1372 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1373 into CONN_USED_HASH and this prodice warning when put connection
1374 again in unused hash.
1377 Severity : enhancement
1379 Description: File striping can now be set to use an arbitrary pool of OSTs.
1381 Severity : enhancement
1383 Description: Export bytes_read/bytes_write count on OSC/OST.
1387 Description: Early reply size mismatch, MGC loses connection
1388 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1389 the connect flags are properly negotiated.
1394 Description: quotas are not honored with O_DIRECT
1395 Details : all writes with the flag O_DIRECT will use grants which leads to
1396 this problem. Now using OBD_BRW_SYNC to guard this.
1400 Description: add quota statistics
1401 Details : 1. sort out quota proc entries and proc code.
1402 2. add quota statistics
1404 Severity : enhancement
1406 Description: enable quota support for HEAD.
1410 Description: Properly propagate oinfo flags from lov to osc for statfs
1411 Details : restore missing copy oi_flags to lov requests.
1413 Severity : enhancement
1415 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1416 : l_getgroups(8), lst(8), routerstat(8)
1417 Details : included man pages for llobdstat(8), llstat(8),
1418 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1420 Severity : enhancement
1422 Description: configure's --enable-quota should check the
1423 : kernel .config for CONFIG_QUOTA
1424 Details : configure is terminated if --enable-quota is passed but
1425 : no quota support is in kernel
1429 Description: Remove portals compatibility
1430 Details : Remove portals compatibility, not interoperable with releases
1435 Description: Resolve device initialization race
1436 Details : Prevent proc handler from accessing devices added to the
1437 obd_devs array but yet be intialized.
1439 Severity : enhancement
1441 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1443 Severity : enhancement
1445 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1450 Description: Using wrong pointer in osc_brw_prep_request
1451 Details : Access to array[-1] can produce panic if kernel compiled with
1452 CONFIG_PAGE_ALLOC enabled
1456 Description: Client runs out of low memory
1457 Details : Consider only lowmem when counting initial number of llap pages
1461 Description: *optional* service tags registration
1462 Details : if the "service tags" package is installed on a Lustre node
1463 When the filesystem is mounted, a local-node service tag will
1464 be created. See http://inventory.sun.com/ for more information
1465 about the Service Tags asset management system.
1469 Description: Kernel BUG tries to release flock
1470 Details : Lustre does not destroy flock lock before last reference goes
1471 away. So always drop flock locks when client is evicted and
1472 perform unlock regardless of successfulness of speaking to MDS.
1476 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1480 Description: sanity test 65a fails if stripecount of -1 is set
1481 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1486 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1487 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1490 Severity : enhancement
1492 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1496 Description: do not process already freed flock
1497 Details : flock can possibly be freed by another thread before it reaches
1498 to ldlm_flock_completion_ast.
1502 Description: LBUG during stress test
1503 Details : Need properly lock accesses the flock deadlock detection list.
1507 Description: oops in page fault handler
1508 Details : kernel page fault handler can return two special 'pages' in error
1509 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1513 Description: timeout with invalidate import.
1514 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1515 should be handled by ptlrpcd. This produce long age waiting and
1516 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1518 Severity : enhancement
1520 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1524 Description: enable MGS and MDT services start separately
1525 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1526 a co-located MGS without starting the MGS, which is a complement
1527 to 'nosvc' mount option.
1531 Frequency : after recovery
1532 Description: precreate to many object's after del orphan.
1533 Details : del orphan st in oscc last_id == next_id and this triger growing
1534 count of precreated objects. Set flag LOW to skip increase count
1535 of precreated objects.
1539 Frequency : rare, on clear nid stats
1540 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1541 Details : when clean nid stats sometimes try destroy live entry,
1542 and this produce panic in free.
1546 Description: Stack overflow during MDS log replay
1547 ease stack pressure by using a thread dealing llog_process.
1551 Description: wait until IO finished before start new when do lock cancel.
1552 Details : VM protocol want old IO finished before start new, in this case
1553 need wait until PG_writeback is cleared until check dirty flag and
1554 call writepages in lock cancel callback.
1556 Severity : enhancement
1558 Description: using special macro for print time and cleanup in includes.
1562 Description: mds_mfd_close() ASSERTION(rc == 0)
1563 Details : In mds_mfd_close(), we need protect inode's writecount change
1564 within its orphan write semaphore to prevent possible races.
1568 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1572 Frequency : rare, on shutdown ost
1573 Description: don't hit live lock with umount ost.
1574 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1575 use shrink_dcache_sb instead.
1579 Description: don't panic with use echo client
1580 Details : echo client pass NULL as client nid pointer and this produce null
1581 pointer dereference.
1585 Description: fix build on ppc32
1586 Details : compile code with -m64 flag produce wrong object file for ppc32.
1590 Description: add message levels for liblustreapi
1594 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1595 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1596 will result in an -EAGAIN and in the exit call path we call
1597 alloc_rr() which will with increasing aggressiveness attempt to
1598 aquire precreated objects on the minimum number of required OSCs.
1602 Description: Use old size assignment to avoid deadlock
1603 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1604 the scheduling while holding a spinlock. We do not need locking
1605 for size in ll_update_inode() because size is only updated from
1606 the MDS for directories or files without objects, so there is no
1607 other place to do the update, and concurrent access to such inodes
1608 are protected by the inode lock.
1612 Description: resolve "_IOWR redefined" build error on SLES10
1616 Description: dump the memory debugging after all modules are unloaded to
1617 suppress false negative in conf_sanity test 39
1619 Severity : enhancement
1621 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1626 Description: fix SLES kernel versioning
1627 Details : the kernel version for our SLES 10 kernel did not include a "-"
1628 before the "smp" at the end. while this was not a problem in
1629 general, it did mean that software trying to use the kernel
1630 version to try to detect a vendor specific kernel would fail.
1631 this was most evident by the OFED build scripts.
1635 Description: Don't update lov_desc members until making sure they are valid
1636 Details : When updating lov_desc members via proc fs, need fix their
1637 validities before doing the real update.
1641 Description: don't put request into delay list while invalidate in flight.
1642 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1643 import in flight. this produce timeout for invalidate and sometimes
1644 can cause stale data.
1648 Frequency : on ppc only
1649 Description: not convert ost objects for directory because it's not exist.
1650 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1654 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1656 Severity : enhancement
1658 Description: Update to support 2.6.22.14 vanilla kernel.
1662 Frequency : rare, on recovery
1663 Description: read procfs can produce deadlock in some situation
1664 Details : Holding lprocfs lock which send rpc can produce block for destroy
1665 obd objects and this also block reconnect with -EALREADY. This
1666 isn't fix all lprocfs bugs - but make it rare.
1668 Severity : enhancement
1670 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1673 Frequency : frequent on X2 node
1675 Description: mdc_set_open_replay_data LBUG
1676 Details : Set replay data for requests that are eligible for replay.
1680 Description: lustre_mgs: operation 101 on unconnected MGS
1681 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1682 MGC, and late on MGC cannot successfully connect to MGS and a lot
1683 of the error messages complaining that MGS is not connected.
1687 Frequency : on network error
1688 Description: panic with double free request if network error
1689 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1690 but it's true only for synchronus enqueue, for async enqueue
1691 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1694 Severity : enhancement
1696 Description: client-side metadata stat-ahead during readdir(directory readahead)
1697 Details : perform client-side metadata stat-ahead when the client detects
1698 readdir and sequential stat of dir entries therein
1701 Frequency : on start mds
1703 Description: Implement get_info(last_id) in obdfilter.
1706 Frequency : occasional
1708 Description: Correctly check stale fid, not start epoch if ost not support SOM
1709 Details : open with flag O_CREATE need set old fid in op_fid3 because
1710 op_fid2 was overwritten with new generated fid, but mds can answer
1711 with one of these two fids and both is not stale. Setattr
1712 incorrectly started an epoch and assume will be called
1713 done_writing, but without SOM done_writing ever being called.
1716 Frequency : rare, depends on device drivers and load
1718 Description: MDS or OSS nodes crash due to stack overflow
1719 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1720 In some cases, in conjunction with device drivers that use a lot
1721 of stack the MDS (or possibly OSS) service threads could overflow
1722 the stack. One change which was identified to consume additional
1723 stack has been reworked to avoid the extra stack usage.
1726 Frequency : occasional
1728 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1729 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1730 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1731 oscc_has_objects and see OSC has no objests and call
1732 oscc_internal_create to resend create request.
1734 Severity : enhancement
1736 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1738 Severity : enhancement
1740 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1743 Frequency : very rare
1745 Description: Fix replay if there is an un-replied request and open
1746 Details : In some cases, older replay request will revert the
1747 mcd->mcd_last_xid on MDS which is used to record the client's
1748 latest sent request.
1750 Severity : enhancement
1752 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1754 Severity : enhancement
1756 Description: Add rhel5 support to HEAD.
1758 Serverity : enhancement
1760 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1765 Description: Don't allow skipping OSTs if index has been specified.
1766 Details : Don't allow skipping OSTs if index has been specified, make
1767 locking in internal create lots better.
1771 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1772 Details : access to bitfield in structure is always rounded to long
1773 and this produce problem with not atomic change any bit.
1777 Description: Lustre make rpms failed.
1778 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1779 builds Lustre rpms from tarball.
1782 Frequency : rare on shutdown ost
1784 Description: If llog cancel was not send before clean_exports phase, this can
1785 produce deadlock in llog code.
1786 Details : If llog thread has last reference to obd and call class_import_put
1787 this produce deadlock because llog_cleanup_commit_master wait when
1788 last llog_commit_thread exited, but this never success because was
1789 called from llog_commit_thread.
1793 Description: allow userland application know is lost one of stripes.
1794 Details : fill lvb_blocks with error code on ost and return it to
1795 application if error flag found.
1799 Description: NULL lov_tgts causing MDS oops
1800 Details : more safe checks for NULL lov_tgts for avoid oops.
1802 Severity : enhancement
1804 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1808 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1809 Details : If an OST is down the MDS will hang indefinitely in
1810 obd_statfs_rqset() waiting for the statfs data. While for
1811 MDS QOS usage of statfs, it should not stuck in waiting.
1813 Severity : enhancement
1815 Description: remote_acl support
1816 Details : Support ACL-based permission check for remote user.
1817 Support setfacl/getfacl for remote user with the utils
1818 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1819 the system "{s,g}etfacl" utils.
1821 Severity : enhancement
1823 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1825 Severity : enhancement
1827 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1831 Description: make mgs_setparam() handle fsname containing dash
1832 Details : fsname containing a dash does not work with lctl conf_param
1834 Severity : enhancement
1836 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1838 Severity : enhancement
1840 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1844 Frequency : rare, at shutdown
1845 Description: access already free / zero obd_namespace.
1846 Details : if client_disconnect_export was called without force flag set,
1847 and exist connect request in flight, this can produce access to
1848 NULL pointer (or already free pointer) when connect_interpret
1849 store ocd flags in obd_namespace.
1853 Frequency : only at startup
1854 Description: not alloc memory with spinlock held.
1855 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1856 if any spinlock held.
1858 Severity : enhancement
1860 Description: make lustre randomly fail allocating memory
1861 Details : Make lustre randomly failed allocating memory for testing purpose.
1863 Severity : enhancement
1865 Description: lost problems with lov objid file
1866 Details : Fixes some scability and access to not inited memory problems
1867 in work with lov objdid file.
1872 Description: lfs find does not continue on file error
1873 Details : Continue other files processing when a file/dir is absent.
1877 Description: Inconsistent usage of lustre_pack_reply()
1878 Details : Standardize the usage of lustre_pack_reply() such that it
1879 always generate a CERROR on failure.
1882 Frequency : occasional
1884 Description: MDS hang or stay in waiting lock
1885 Details : If client receive lock with CBPENDING flag ldlm need send lock
1886 cancel as separate rpc, to avoid situation when cancel request
1887 can't processed due all i/o threads stay in wait lock.
1891 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1895 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1899 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1903 Description: Update to RHEL4 latest kernel.
1905 Severity : enhancement
1907 Description: Build SLES10 patchless client fails
1908 Details : The configure was broken by run ./configure with
1909 --with-linux-obj=.... argument for patchless client. When the
1910 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1911 can't search header adequately. Use absolute path such as
1912 -I($LINUX)/include instead.
1916 Description: interrupt oig_wait produce painc on resend.
1917 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1918 private set, and this produce situation when rq_ptlrpcd_data not
1919 copyed to new allocated request and triggered LBUG on assert
1920 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1921 ptlrpcd set if request is changed and can be safety changed to use
1926 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1927 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1928 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1929 these macros and replace them with their difinitions instead.
1933 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1934 Details : Update which_patch & target file for SLES10 latest kernel.
1937 Description: Remove client patches from SLES 10 kernel.
1938 Details : This causes SLES 10 clients to behave as patchless clients
1939 even on a Lustre-patched (server) kernel.
1941 Severity : enhancement
1943 Description: self-adjustable client's lru lists
1944 Details : use adaptive algorithm for managing client cached locks lru
1945 lists according to current server load, other client's work
1946 pattern, memory activities, etc. Both, server and client
1947 side namespaces provide number of proc tunables for controlling
1950 Severity : enhancement
1952 Description: light-weight GSS support
1953 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1954 and reduce performance overhead.
1956 Severity : enhancement
1958 Description: Linux keyring support
1959 Details : Support using service of Linux keyring for Lustre GSS internal
1960 context refresh/cache mechanism.
1964 Description: Fix errors in lfs documentation
1965 Details : Fixes man pages
1969 Description: don't use GFP_* in generic Lustre code.
1970 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1974 Description: obdclass is limited by single OBD_ALLOC(idarray)
1975 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1979 Description: warnings with build patchless client with vanila 2.6.19 and up
1980 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1981 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1985 Frequency : Only for SLES
1986 Description: sanity_quota fail test_1
1987 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1988 file, which makes TSTID[2] inunique.
1992 Frequency : Only for SLES9
1993 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1994 Details : Update target/ChangeLog/which_patch .
1998 Frequency : Only for test_10 in sanity-quota.sh
1999 Description: a bug in quota test code
2000 Details : checking a test flag in a wrong place
2004 Frequency : Only for test_13 in sanity-quota.sh
2005 Description: a bug in quota test script
2006 Details : shouldn't check overfull allocated quota
2010 Description: jbd statistics
2011 Details : Port older jbd statistics patch for sles10
2013 Severity : enhancement
2015 Description: Add printing inode info into message about error in writepage.
2020 Description: reply_lock_interpret crash due to race with it and lock cancel.
2021 Details : Do not replay locks that are being cancelled. Do not reference
2022 locks by their address during replay, just by their handle.
2026 Frequency : When flocks are used.
2027 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2028 Details : Flock locks might destroy just granted lock if it could be merged
2029 with another existing flock, this is done in completion handler,
2030 so teach ldlm_cli_enquque_fini that this is a valid case for
2036 Description: Oops in read and write path when failing to allocate lock.
2037 Details : Check if lock allocation failed and return error back.
2041 Description: lstripe command fails for valid OST index
2042 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2043 instead of lov->desc.ld_active_tgt_count.
2047 Description: sanity.sh failed test 103
2048 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2053 Description: df doesn't work properly if diskfs blocksize != 4K
2054 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2058 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2059 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2061 Severity : enhancement
2063 Description: Adaptive timeouts
2064 Details : RPC timeouts adapt to changing server load and network
2065 conditions to reduce resend attempts and improve recovery time.
2069 Description: llapi_file_create() does not allow some changes
2070 Details : add llapi_file_open() that allows specifying the mode and
2071 open flags, and also returns an open file handle.
2075 Description: merge and cleanup kernel patches.
2076 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2080 Description: Add journal checksum support.(Kernel part)
2081 Details : The journal checksum feature adds two new flags i.e
2082 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2083 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2084 indicates that the commit block contains the checksum for
2085 the blocks described by the descriptor blocks. Now commit
2086 record can be sent to disk without waiting for descriptor
2087 blocks to be written to disk. This behavior is controlled
2088 using JBD2_FEATURE_ASYNC_COMMIT flag.
2092 Description: OSS needs mutliple precreate threads
2093 Details : Add ability to start more than one create thread per OSS.
2097 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2098 Details : execute lfs setstripe on client
2102 Description: mds_obd_create error creating tmp object
2103 Details : When the user sets quota on root, llog will be affected and can't
2104 create files and write files.
2107 Frequency : Always on ia64 patchless client, and possibly others.
2109 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2110 Details : This allows the patchless client to be loaded on architectures
2111 without this export.
2115 Description: RedHat Update kernel for RHEL5
2116 Details : Add the kernel config file for RHEL5.
2120 Description: RedHat Update kernel for RHEL5
2121 Details : Modify the kernel config file more closer RHEL5.
2125 Description: Build failure against Centos5 (RHEL5)
2126 Details : Define PAGE_SIZE when it isn't present.
2130 Description: test_7 on sanity-quota.sh can't be used on separate machines
2131 Details : add facet to handle it
2135 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2136 Details : with NFS, the anon dentry's parent was set to itself in
2137 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2138 corresponding dentry other than use rec->ur_name.
2140 Severity : enhancement
2142 Description: lfs setstripe enhancement
2143 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2147 Description: enable data checksumming by default
2148 Details : enable checksum by default, allow --disable-checksum
2149 configure option and "-o nochecksum" mount option. Checksums
2150 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2154 Description: lustre support for RHEL5
2155 Details : Add support for RHEL5.
2159 Description: Client eviction due to failover config
2160 Details : after a connection loss, the lustre client should attempt to
2161 reconnect to the last active server first before trying the
2162 other potential connections.
2166 Description: when mds and osts use different quota unit(32bit and 64bit),
2167 quota will be released repeatly.
2168 Details : void sending multiple quota reqs to mds, which will keep the
2169 status between the reqs.
2173 Description: osts not allocated evenly to files
2174 Details : change the condition to increase offset_idx
2177 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2179 Description: Data corruption for OSTs that are formatted larger than 2TB
2181 Details : When generating the bio request for lustre file writes the
2182 sector number would overflow a temporary variable before being
2183 used for the IO. The data reads correctly from Lustre (which
2184 will overflow in a similar manner) but other file data or
2185 filesystem metadata may be corrupted in some cases.
2189 Description: Tune the kernel for good SCSI performance.
2190 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2191 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2196 Description: rewrite ext2-derived code in obdclass/uuid.c
2197 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2198 from scratch preserving functionality.
2202 Description: "lfs find -obd UUID" prints directories
2203 Details : "lfs find -obd UUID" will return all directory names instead
2204 of just file names. It is incorrect because the directories
2205 do not reside on the OSTs.
2209 Description: use i_size_read and i_size_write in 2.6 port
2210 Details : replace inode->i_size access with i_size_read/write()
2213 Frequency : when removing large files
2215 Description: scheduling issue during removal of large Lustre files
2216 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2217 It causes scheduling issues when removing large files (17TB in the
2221 Frequency : only with liblustre clients on XT3
2223 Description: evictions taking too long
2224 Details : allow llrd to evict clients directly on OSTs
2227 Frequency : only on ppc
2229 Description: /proc/fs/lustre/devices broken on ppc
2230 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2231 We should make sure the type of 'index' is loff_t and then cast
2232 explicitly as needed below. Do not assign an explicitly cast
2236 Frequency : only for rhel5
2238 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2239 Details : Modify the target file & which_kernel.
2241 Severity : enhancement
2243 Description: omit set fsid for export NFS
2244 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2245 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2249 Frequency : Always, for kernels after 2.6.16
2250 Description: Fix warning idr_remove called for id=.. which is not allocated.
2251 Details : Last kernels save old s_dev before kill super and not allow
2252 to restore from callback - restore it before call kill_anon_super.
2256 Description: buffer overruns could theoretically occur
2257 Details : llapi_semantic_traverse() modifies the "path" argument by
2258 appending values to the end of the origin string, and a
2259 overrun may occur. Adding buffer overrun check in liblustreapi.
2263 Description: Fix error on 'ls .' at the top of the Lustre mount.
2264 Details : Don't revalidate dentry if it is a root dentry.
2268 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2269 Details : Modify vm-tunables-rhel4.patch.
2273 Description: Kernel config for 2.6.18-vanilla.
2274 Details : Modify targets/2.6-vanilla.target.in.
2275 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2276 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2277 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2278 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2282 Description: improve handling recoverable errors
2283 Details : if request processig with error which can be recoverable on server
2284 request should be resend, otherwise page released from cache and
2289 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2290 Details : Modify target file & which_patch.
2291 A flaw was found in the IA32 system call emulation provided
2292 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2293 value could be stored in the %RAX register, which could trigger an
2294 out-of-bounds system call table access. An untrusted local user
2295 could exploit this flaw to run code in the kernel
2296 (ie a root privilege escalation). (CVE-2007-4573).
2300 Description: O_DIRECT bypasses client statistics.
2301 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2302 read_ahead_stats were updated. copied the stats section from
2303 osc_send_oap_rpc() into async_internal().
2307 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2311 Description: change order of libsysio includes
2312 Details : '#include sysio.h' should always come before '#include xtio.h'
2316 Description: handle "serious error: objid * already exists" more gracefully
2317 Details : If LAST_ID value on disk is smaller than the objects existing in
2318 the O/0/d* directories, it indicates disk corruption and causes an
2319 LBUG(). If the object is 0-length, then we should use the existing
2320 object. This will help to avoid a full fsck in most cases.
2322 Severity : enhancement
2324 Description: adapt the lustre_config script to support the upgrade case
2325 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2326 server targets to 1.6 in parallel.
2330 Description: To avoid grant space > avaible space when the disk is almost
2331 full. Without this patch you might see the error "grant XXXX >
2332 available" or some LBUG about grant, when the disk is almost
2334 Details : In filter_check_grant, for non_grant cache write, we should
2335 check the left space by if (*left > ungranted + bytes), instead
2336 of (*left > ungranted), because only we are sure the left space
2337 is enough for another "bytes", then the ungrant space should be
2338 increase. In client, we should update cl_avail_grant only there
2339 is OBD_MD_FLGRANT in the reply.
2343 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2344 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2345 described in CVE-2007-4573. This problem could allow untrusted
2346 local users to gain root access.
2349 Frequency : when using O_DIRECT and quotas
2351 Description: Incorrect file ownership on O_DIRECT output files
2352 Details : block usage reported by 'lfs quota' does not take into account
2353 files that have been written with O_DIRECT.
2358 Description: touch file failed when fs is not full
2359 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2360 otherwise we can get ENOSP while fs is not full.
2364 Description: parallel lock callbacks
2365 Details : Instead of sending blocking and completion callbacks as separated
2366 requests, adding them to a set and sending in parallel.
2369 Frequency : only for Cray XT3
2370 Bugzilla : 12829/13455
2371 Description: Changing primary group doesn't change the group lustre assigns to
2373 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2374 overridden with the primary group provided by the group upcall,
2375 whereas the supplied fsgid can be trusted if it is in the list of
2376 supplementary groups returned by the group upcall.
2378 Severity : enhancement
2380 Description: Allow masking D_WARNING, D_ERROR messages from console
2381 Details : Console messages can now be disabled via lnet.printk.
2385 Description: User code with malformed file open parameter crashes client node
2386 Details : Before packing join_file req, all the related reference should be
2387 checked carefully in case some malformed flags cause fake
2388 join_file req on client.
2392 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2393 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2394 mode is equal to the granted mode. However, it should also check
2395 that LDLM_FL_CP_REQD is not set.
2399 Description: Heavy nfs access might result in deadlocks
2400 Details : After ELC code landed, it is now improper to enqueue any mds
2401 locks under och_sem, because enqueue might want to decide to
2402 cancel open locks for same inode we are holding och_sem for.
2406 Description: Client eviction while running blogbench
2407 Details : A lot of unlink operations with concurrent I/O can lead to a
2408 deadlock causing evictions. To address the problem, the number of
2409 oustanding OST_DESTROY requests is now throttled to
2410 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2411 callbacks are processed in priority.
2415 Description: enable ACLs on MDS by default
2416 Details : ACLs must be enabled on MDS by default.
2419 Frequency : PPC/PPC64 only
2421 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2422 Details : fix duplicated definitions between asm-ppc64/types.h and
2423 lustre_types.h on PPC.
2426 Frequency : PPC/PPC64 only
2428 Description: asm-ppc/segment.h does not exist
2429 Details : fix compile issue on PPC.
2433 Description: better handle error messages in extents code
2436 Frequency : RHEL4 only
2438 Description: mkfs is very slow on IA64/RHEL4
2439 Details : A performance regression has been discovered in the MPT Fusion
2440 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2441 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2442 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2444 Severity : enhancement
2446 Description: SNMP support enhancement
2447 Details : Adding total number of sampled request for an MDS node in snmp
2450 Severity : enhancement
2452 Description: Optimize ldlm waiting list processing for PR extent locks
2453 Details : When processing waiting list for read extent lock and meeting read
2454 lock that is same or wider to it that is not contended, skip
2455 processing rest of the list and immediatelly return current
2456 status of conflictness, since we are guaranteed there are no
2457 conflicting locks in the rest of the list.
2461 Description: Time out and refuse to reconnect
2462 Details : When the failover node is the primary node, it is possible
2463 to have two identical connections in imp_conn_list. We must
2464 compare not conn's pointers but NIDs, otherwise we can defeat
2465 connection throttling.
2469 Description: port llog fixes from b1_6 into HEAD
2470 Details : Port llog reference couting and some llog cleanups from b1_6
2471 (bug 10800) into HEAD, for protect from panic and access to already
2472 free llog structures.
2476 Description: Detect stride IO mode in read-ahead
2477 Details : When a client does stride read, read-ahead should detect that and
2478 read-ahead pages according to the detected stride pattern.
2482 Description: data checksumming impacts single node performance
2483 Details : add support for several checksum algorithm. Currently, only CRC32
2484 and Adler-32 are supported. The checksum type can be changed on
2485 the fly via /proc/fs/lustre/osc/*/checksum_type.
2489 Description: use adler32 for page checksums
2490 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2495 Description: build for x2 fails
2496 Details : fix compile issue on Cray systems.
2500 Description: Properly match for duplicate locks
2501 Details : Due to different lock order from skiplists code, we need to
2502 traverse entire list for now
2505 Frequency : only on PPC/SLES10
2507 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2508 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2509 be defined. Add a hack in mkfs_lustre.c to work around this compile
2514 Description: LASSERT on MDS when client holding flock lock dies
2515 Details : ldlm pool logic depends on number of granted locks equal to
2516 number of released locks which is not true for flock locks, so
2517 just exclude such locks from consideration.
2521 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2522 Details : Use fsfilt_sync() for both the whole filesystem sync and
2523 individual file sync to eliminate dangerous inode locking
2524 with I_LOCK that can lead to a deadlock.
2528 Description: performance in 1.6.3
2529 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2530 sure that 1MB requests are not fragmented by the block layer.
2532 Severity : enhancement
2534 Description: organize the server-side client stats on per-nid basis
2535 Details : Change the structure of stats under obdfilter and mds to
2543 The "uuid"s file would list the uuids of _active_ exports.
2544 And the clear entry is to clear all stats and stale nids.
2546 Severity : enhancement
2548 Description: eliminate client locks in face of contention
2549 Details : file contention detection and lockless i/o implementation
2550 for contended files.
2554 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2557 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2559 Description: MDS may lose file striping (and hence file data) in some cases
2560 Details : If there are additional extended attributes stored on the MDS,
2561 in particular ACLs, SELinux, or user attributes (if user_xattr
2562 is specified for the client mount options) then there is a risk
2563 of attribute loss. Additionally, the Lustre file striping
2564 needs to be larger than default (e.g. striped over all OSTs),
2565 and an additional attribute must be stored initially in the
2566 inode and then increase in size enough to be moved to the
2567 external attribute block (e.g. ACL growing in size) for file
2572 Description: skiplist implementation simplification
2573 Details : skiplists are used to group compatible locks on granted list
2574 that was implemented as tracking first and last lock of each
2575 lock group the patch changes that to using doubly linked lists
2579 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2580 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2581 because LASSERT on both the data supplied by a client, and the
2582 data on disk is dangerous and incorrect.
2584 Severity : enhancement
2586 Description: Slow truncate/writes to huge files at high offsets.
2587 Details : Directly associate cached pages to lock that protect those pages,
2588 this allows us to quickly find what pages to write and remove
2589 once lock callback is received.
2593 Description: more ldlm soft lockups
2594 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2595 starve other threads from the resource lock for a long time in
2596 case of long waiting queue, so change the debug level from
2597 D_OTHER to the less frequently used D_INFO.
2599 Severity : enhancement
2601 Description: add -gid, -group, -uid, -user options to lfs find
2605 Description: Hung threads in invalidate_inode_pages2_range
2606 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2607 once one is completed. As a result, some RPCs are stuck in the
2608 queue and are never sent.
2612 Description: filter threads hungs on waiting journal commit
2613 Details : Cleanup filter group llog code, then only filter group llog will
2614 be only created in the MDS/OST syncing process.
2618 Description: Procfs and llog threads access destoryed import sometimes.
2619 Details : Sync the import destoryed process with procfs and llog threads by
2620 the import refcount and semaphore.
2622 Severity : enhancement
2624 Description: openlock cache of b1_6 port to HEAD
2629 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2630 Details : If insertion of an extent fails, then discard the inode
2631 preallocation and free data blocks else it can lead to duplicate
2636 Description: don't always update ctime in ext3_xattr_set_handle()
2637 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2638 In some cases the ctime should not be updated, for example for
2639 2.0->1.8 compatibility it is necessary to delete an xattr and it
2640 should not update the ctime.
2644 Bugzilla : 15713/16362
2645 Description: Assertion in iopen_connect_dentry in 1.6.3
2646 Details : looking up an inode via iopen with the wrong generation number can
2647 populate the dcache with a disconneced dentry while the inode
2648 number is in the process of being reallocated. This causes an
2649 assertion failure in iopen since the inode's dentry list contains
2650 both a connected and disconnected dentry.
2654 Description: assertion failure in ldlm_handle2lock()
2655 Details : fix a race between class_handle_unhash() and class_handle2object()
2656 introduced in lustre 1.6.5 by bug 13622.
2661 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2662 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2667 Description: oops in sd_iostats_seq_show()
2668 Details : unloading/reloading the scsi low level driver triggers a kernel
2669 bug when trying to access the sd iostat file.
2674 Description: Kernel panics during QLogic driver reload
2675 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2676 patch, causing memory corruption.
2681 Description: journal_dev option does not work in b1_6
2682 Details : pass mount option during pre-mount.
2684 Severity : enhancement
2686 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2687 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2688 extent information of a file. It can be used to map logical blocks
2689 in a file to physical blocks in the block device.
2693 Description: LDLM soft lockups - improvement
2694 Details : It is be possible to send the lock handle along with each read
2695 or write request because the client is already doing a lock match
2696 itself so there isn't any reason the OST should have to re-do that
2700 Frequency : only on Cray X2
2702 Description: X2 build failures
2703 Details : fix build failures on Cray X2.
2707 Description: xid & resent requests
2708 Details : Initialize RPC XID from clock at startup (randomly if clock is
2711 Severity : enhancement
2713 Description: Add lustre_start utility to start or stop multiple Lustre servers
2718 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2719 Details : In case of memory pressure, list_del() can be called twice on
2720 req->rq_history_list, causing a kernel oops.
2724 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2725 Details : fix stack overflow in the distributed lock manager by defering
2726 export eviction after a failed AST to the elt thread instead of
2727 handling it in the dlm interpret routine.
2731 Description: Convert some comments to new format.
2732 Details : Update documenting comments to match doxygen conventions.
2736 Description: Grammar fixes.
2737 Details : A couple of trivial spelling fixes.
2741 Description: OSD_COUNTERS-mandatory
2742 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2743 to simplify the code.
2747 Description: simplify cmm_device freeing logic.
2748 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2752 Description: Add lockdep support to dt_object_operations locking interface.
2753 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2754 parameter indicating lock ordering. Update mdd code to use new
2759 Description: Introduce failloc constants for lockless IO tests.
2760 Details : Add two new failloc constants to test lockless IO. Only one of
2761 them in implemented---another is checked in yet to be landed
2766 Description: Add lockdep support for inode mutex.
2767 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2768 in the situations where Lustre has to lock more than one inode
2773 Description: Add optional invariants checking support.
2774 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2775 configure switch. This macro is to be used for consistency and
2776 sanity checks that are too expensive to be left in `production'
2781 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2782 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2787 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2788 Details : Use spin_lock_nested() in (the only) situation where more than
2789 one ldlm_lock is locked simultaneously. Also, fix possible
2790 dead-lock in ldlm_lock_change_resource() by enforcing particular
2795 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2796 Details : Instead of specifying each ldlm_lock call-back through separate
2797 parameter, wrap them into struct ldlm_callback_suite.
2801 Description: Kill join_lru obd method and its callers.
2802 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2803 in memory---a requirement implemented through ->o_join_lru() obd
2804 method in HEAD. Remove this method and its users.
2808 Description: Add asynchronous ldlm ENQUEUE completion handler.
2809 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2810 case that stock ldlm_completion_ast() cannot handle as it waits
2811 until lock is granted. Introduce new ldlm_completion_ast_async()
2812 for this. Also comment ldlm_completion_ast().
2816 Description: ldlm_error <-> errno conversion.
2817 Details : Add functions to map (rather arbitrary) between LDLM error codes
2818 and standard errno values. CLIO needs this to prevent LDLM specific
2819 constants from escaping ldlm and osc.
2823 Description: Kill unused ldlm_handle2lock_ns() function.
2824 Details : Kill unused ldlm_handle2lock_ns() function.
2828 Description: Add lu_ref support to ldlm_lock
2829 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2830 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2831 and ldlm_resource. LDLM interface has to be changed, because code that
2832 releases a reference on a lock, has to "know" what reference this is.
2833 In the most frequent case
2835 lock = ldlm_handle2lock(handle);
2837 LDLM_LOCK_PUT(lock);
2839 no changes are required. When any other reference (received _not_ from
2840 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2841 instead of LDLM_LOCK_PUT().
2843 Arguably, changes are pervasive, and interface requires some discipline
2844 for proper use. On the other hand, it was very instrumental in finding
2845 a few leaked lock references.
2849 Description: Add ldlm_lock_addref_try().
2850 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2851 attempts to addref a lock that might be being canceled
2856 Description: Add ldlm_weigh_callback().
2857 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2858 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2859 hard-coded `number of pages' logic.
2863 Description: Add lockdep annotations to llog code.
2864 Details : Use appropriately tagged _nested() locking calls in the places
2865 where llog takes more than one ->lgh_lock lock.
2869 Description: Add loi_kms_set().
2870 Details : Wrap kms updates into a helper function.
2874 Description: Constify instances of struct lsm_operations.
2875 Details : Constify instances of struct lsm_operations.
2879 Description: lu_conf support.
2880 Details : On a server, a file system object is uniquely identified
2881 by a fid, which is sufficient to locate and load all object
2882 state (inode). On a client, on the other hand, more data are
2883 necessary instantiate an object. Change lu_object_find() and
2884 friends to take additional `lu_conf' argument describing object.
2885 Typically this includes layout information.
2889 Description: lu_context fixes.
2890 Details : Introduce new lu_context functions that are needed on the client
2891 side, where some system threads (ptlrpcd) are shared by multiple
2892 modules, and so cannot be stopped during module shutdown.
2896 Description: Add start and stop methods to lu_device_type_operations.
2897 Details : Introduce two new methods in lu_device_type_operations, that are
2898 invoked when first instance of a given type is created and last one
2899 is destroyed respectively. This is need by CLIO.
2903 Description: Add lu_ref support to struct lu_device.
2904 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2905 track leaked references.
2909 Description: Introduce lu_kmem_descr.
2910 Details : lu_kmem_descr and its companion interface allow to create
2911 and destroy a number of kmem caches at once.
2915 Description: Fix lu_object finalization race.
2916 Details : Fix a race between lu_object_find() finding an object and its
2917 concurrent finalization. This race is (most likely) not possible
2918 on the server, but might happen on the client.
2922 Description: Introduce lu_ref interface.
2923 Details : lu_ref is a debugging module allowing to track references to
2924 a given object. It is quite cpu expensive, and has to be
2925 explicitly enabled with --enable-lu_ref. See usage description
2930 Description: Factor lu_site procfs stats into a separate function.
2931 Details : Separate lu_site stats printing code into a separate function
2932 that can be reused on a client.
2936 Description: Constify instances of struct {lu,dt,md}_device_operations.
2937 Details : Constify instances of struct {lu,dt,md}_device_operations.
2941 Description: Introduce struct md_site and move meta-data specific parts of
2942 struct lu_site here.
2943 Details : Move md-specific fields out of struct lu_site into special struct
2944 md_site, so that lu_site can be used on a client.
2948 Description: Kill mdd_lov_destroy().
2949 Details : Remove unused mdd code.
2953 Description: Add st_block checking to multistat.c.
2954 Details : Add st_block checking to multistat.c.
2958 Description: Add lu_ref support to struct obd_device.
2959 Details : Add lu_ref tracking to obd_device.
2963 Description: Kill obd_set_fail_loc().
2964 Details : Remove unused code.
2968 Description: Add special type for ptlrpc_request interpret functions.
2969 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2970 there. Actual usage will be in CLIO.
2974 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2975 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2976 doesn't work with lockdep.
2980 Description: Add rwv.c test program.
2981 Details : New testing program exercising readv(2) and writev(2) (Qian).
2985 Description: Add sendfile.c test program.
2986 Details : New testing program exercising sendfile(2) (Jay).
2990 Description: Ratelimit a message that can be very frequent.
2991 Details : Ratelimit a memory allocation failure message that can
2996 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2997 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
3001 Description: Liblustre build fixes.
3002 Details : Liblustre build fixes.
3006 Description: libcfs: add cfs_{need,cond}_resched() interface.
3007 Details : libcfs: add cfs_{need,cond}_resched() definition and
3008 implementations for Linux, NT, and liblustre.
3010 Severity : enhancement
3012 Description: More exported tunables for mballoc
3013 Details : Add support for tunable preallocation window and new tunables for
3014 large/small requests
3018 Description: Detect corruption of block bitmap and checking for preallocations
3019 Details : Checks validity of on-disk block bitmap. Also it does better
3020 checking of number of applied preallocations. When corruption is
3021 found, it turns filesystem readonly to prevent further corruptions.
3025 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
3026 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
3027 window to 0. If it does hit cache miss, and read-pattern is still
3028 stride-io mode, does not reset the stride window, but also does not
3029 increase the stride window length in this case.
3033 Frequency : only for big-endian servers
3034 Description: Check if system is big-endian while mounting fs with extents feature
3035 Details : Mounting a filesystem with extents feature will fail on big-endian
3036 systems since ext3-based ldiskfs is not supported on big-endian
3037 systems. This can be over-riden with "bigendian_extents" mount option.
3039 Severity : enhancement
3041 Description: The root squash functionality
3042 Details : A security feature, which is to prevent users from being able
3043 to mount lustre on their desktop, run as root, and delete
3044 all of the files in the filesystem. The goal is accomplished by
3045 remapping user id (UID) and group id (GID) of the root user to
3046 a UID and GID specified by the system administartor via Lustre
3047 configuration management server (MGS). The functionality also
3048 allows to specify sets of clients for which the remapping does
3053 Description: Excessive recovery window
3054 Details : With AT enabled, the recovery window can be excessively long (6000+
3055 seconds). To address this problem, we no longer use
3056 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3057 timeout no longer depends on the service time, it is set to
3058 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3059 time via pb_service_time.
3063 Description: Watchdog triggered on MDS failover
3064 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3065 the OSTs know that the MDS "UUID" can be reused for the same export
3066 from a different NID, so we do not need to wait for the export to be
3070 Frequency : rare, only if using MMP with Linux RAID
3072 Description: MMP doesn't work with Linux RAID
3073 Details : While using HA for Lustre servers with Linux RAID, it is possible
3074 that MMP will not detect multiple mounts. To make this work we
3075 need to unplug the device queue in RAID when the MMP block is being
3076 written. Also while reading the MMP block, we should read it from
3077 disk and not the cached one.
3079 Severity : enhancement
3081 Description: open file using fid
3082 Details : A file can be opened using just its fid, like
3083 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3086 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3089 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3090 Details : add proper locking for creation of jbd_4k slab cache
3094 Description: MMP check in ext3_remount() fails without displaying any error
3095 Details : When multiple mount protection fails during remount, proper error
3098 Severity : enhancement
3100 Description: Allow stripe size to be up to 4G-64k
3101 Details : Fix math logic to allow large stripe sizes.
3105 Description: add check for >8TB ldiskfs filesystems
3106 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3107 Don't allow >8TB ldiskfs filesystems to be mounted without
3108 force_over_8tb mount option
3113 Descriptoin: OSS complains "too many blocks" and crashes
3114 Details : In filter_commitrw_write():
3115 1. lquota_chkquota();
3116 2. obdo_from_inode();
3117 3. lquota_pending_commit();
3118 "2." has changed uid and gid in this case, then caused
3119 lquota_chkquota() and lquota_pending_commit() to use
3120 different uid and gid. That is the root of the bug.
3124 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3125 Details : Detect and print out the 'stripe count = -1' case properly.
3129 Description: Out or order replies might be lost on replay
3130 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3131 request with tid smaller than one currently being added, add it
3132 to the start, not end of the list.
3134 --------------------------------------------------------------------------------
3136 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3138 * Support for kernels:
3139 2.6.5-7.283 (SLES 9),
3140 2.6.9-55.EL (RHEL 4),
3141 2.6.16.46-0.14 (SLES 10),
3142 2.6.18.8 vanilla (kernel.org)
3143 * Client support for unpatched kernels:
3144 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3145 2.6.16 - 2.6.22 vanilla (kernel.org)
3146 * Due to recently discovered recovery problems, we do not recommend
3147 using patchless RHEL 4 clients with this or any earlier release.
3148 * Recommended e2fsprogs version: 1.39.cfs8
3149 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3150 * Starting with this release, the ldiskfs backing filesystem required
3151 by Lustre is now in its own package, lustre-ldiskfs. This package
3152 should be installed. It is versioned separately from Lustre and
3153 may be released separately in future.
3157 Description: block reactivating mgc import until all deactivates complete
3158 Details : Fix race when failing back MDT/MGS to itself (testing)
3161 Frequency : at statup only
3163 Description: mds_lov_synchronize race leads to various problems
3164 Details : simultaneous MDT->OST connections at startup can cause the
3165 sync to abort, leaving the OSC in a bad state.
3167 Severity : enhancement
3169 Description: add optional extra BUILD_VERSION info
3170 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3171 to override the lustre version.
3174 Frequency : 2.6.18 servers only
3176 Description: ll_kern_mount() doesn't release the module reference
3177 Details : The ldiskfs module reference count never drops down to 0
3178 because ll_kern_mount() doesn't release the module reference.
3183 Description: server LBUG when using old ost_num_threads parameter
3184 Details : Accept the old ost_num_threads parameter but warn that it
3185 is deprecated, and fix an off-by-one error that caused an LBUG.
3190 Description: Transient SCSI error results in persistent IO issue
3191 Details : iobuf->dr_error is not reinitialized to 0 between two
3195 Frequency : sometimes when underlying device returns I/O errors
3197 Description: OSTs not going read-only during write failures
3198 Details : OSTs are not remounted read-only when the journal commit threads
3199 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3200 instead of the ext3 wrappers.
3204 Description: poor connect scaling with increasing client count
3205 Details : Don't run filter_grant_sanity_check for more than 100 exports
3206 to improve scaling for large numbers of clients.
3209 Frequency : SLES10 only
3211 Description: sanity-quota.sh quotacheck failed: rc = -22
3212 Details : Quotas cannot be enabled on SLES10.
3215 Frequency : liblustre clients only
3217 Description: getdirentries does not give error when run on compute nodes
3218 Details : getdirentries does not fail when the size specified as an argument
3219 is too small to contain at least one entry
3221 Severity : enhancement
3223 Description: Add LNET router traceability for debug purposes
3224 Details : If a checksum failure occurs with a router as part of the
3225 IO path, the NID of the last router that forwarded the bulk data
3226 is printed so it can be identified.
3231 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3232 Details : Due to a race condition, liblustre clients were occasionally
3233 evicted incorrectly.
3235 Severity : enhancement
3237 Description: lfs setstripe use optional parameters instead of postional
3240 Severity : enhancement
3242 Description: Nanosecond timestamp support for ldiskfs
3243 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3244 resolution timestamps. There is not yet support for this at
3245 the Lustre filesystem level.
3248 Frequency : during server recovery
3250 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3251 Details : request with rq_no_resend flag not awake l_wait_event if they get
3255 Frequency : nfs export on patchless client
3257 Description: connectathon hang when test nfs export over patchless client
3258 Details : Disconnected dentry cannot be found with lookup, so we do not need
3259 to unhash it or make it invalid
3262 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3263 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3264 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3269 Description: open req refcounting wrong on reconnect
3270 Details : If reconnect happened between getting open reply from server and
3271 call to mdc_set_replay_data in ll_file_open, we will schedule
3272 replay for unreferenced request that we are about to free.
3273 Subsequent close will crash in variety of ways.
3274 Check that request is still eligible for replay in
3275 mdc_set_replay_data().
3280 Description: disable writes to filesystem when reading health_check file
3281 Details : the default for reading the health_check proc file has changed
3282 to NOT do a journal transaction and write to disk, because this
3283 can cause reads of the /proc file to hang and block HA state
3284 checking on a healthy but otherwise heavily loaded system. It
3285 is possible to return to the previous behaviour during configure
3286 with --enable-health-write.
3288 Severity : enhancement
3290 Description: 64-bit inode version
3291 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3292 to the inode. This will be required for version-based recovery.
3297 Description: MDS fails to start if a duplicate client export is detected
3298 Details : in some rare cases it was possible for a client to connect to
3299 an MDS multiple times. Upon recovery the MDS would detect this
3300 and fail during startup. Handle this more gracefully.
3302 Severity : enhancement
3304 Description: Add -o localflock option to simulate old noflock
3306 Details : This will achieve local-only flock/fcntl locks
3312 Description: log_commit_thread vs filter_destroy race leads to crash
3313 Details : Take import reference before releasing llog record semaphore
3318 Description: Wrong request locking in request set processing
3319 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3320 imp_delayed_list, in this place should be used imp_lock.
3323 Frequency : when reconnection
3325 Description: Grant Leak when osc reconnect to OST
3326 Details : When osc reconnect ost, OST(filter) should check whether it
3327 should grant more space to client by comparing fed_grant and
3328 cl_avail_grant, and return the granted space to client instead
3329 of "new granted" space, because client will call osc_init_grant
3330 to update the client grant space info.
3333 Frequency : when client reconnect to OST
3335 Description: Grant Leak when osc do resend and replay bulk write
3336 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3337 bulk write request, because the grant info while be sync between
3338 OSC and OST when reconnect, and we should ignore the grant info
3339 these of resend/replay write req.
3344 Description: Grant space more than available left space sometimes.
3345 Details : When then OST is about to be full, if two bulk writing from
3346 different clients came to OST. Accord the avaliable space of the
3347 OST, the first req should be permitted, and the second one
3348 should be denied by ENOSPC. But if the seconde arrived before
3349 the first one is committed. The OST might wrongly permit second
3350 writing, which will cause grant space > available space.
3353 Frequency : when client is evicted
3355 Description: Grant might be wrongly erased when osc is evicted by OST
3356 Details : when the import is evicted by server, it will fork another
3357 thread ptlrpc_invalidate_import_thread to invalidate the
3358 import, where the grant will be set to 0. While the original
3359 thread will update the grant it got when connecting. So if
3360 the former happened latter, the grant will be wrongly errased
3361 because of this race.
3366 Description: Checking Stale with correct fid
3367 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3368 to check whether it is stale, because sometimes, we want the
3369 enqueue happened anyway, and op_data.fid2 will not be initialized.
3371 Severity : enhancement
3373 Description: update patchless client
3374 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3377 Frequency : only with 2.4 kernel
3379 Description: random memory corruption
3380 Details : size of struct ll_inode_info is to big for union inode.u and this
3381 can be cause of random memory corruption.
3386 Description: Memory leak in recovery
3387 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3388 It should also check obd_fail before fsfilt_start, otherwise if
3389 fsfilt_start return -EROFS,(failover mds during mds recovery).
3390 then the req will return with repmsg->transno = 0 and rc = EROFS.
3391 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3392 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3393 be freed no matter whether fsfilt_commit success or not.
3396 Frequency : only with huge count clients
3398 Description: Prevents from taking the superblock lock in llap_from_page for
3400 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3401 for llap_from_page call in ll_removepage prevents from taking the
3402 superblock lock for a soon died page.
3407 Description: Not check open intent error before release open handle
3408 Details : in some rare cases, the open intent error is not checked before
3409 release open handle, which may cause
3410 ASSERTION(open_req->rq_transno != 0), because it tries to release
3411 the failed open handle.
3416 Description: Set cat log bitmap only after create log success.
3417 Details : in some rare cases, the cat log bitmap is set too early. and it
3418 should be set only after create log success.
3422 Description: Accessing a block bevice can re-enable I/O when Lustre is
3423 tearing down a device.
3424 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3429 Description: service threads may hog cpus when there are a lot of requests
3430 Details : Insert cond_resched to give other threads a chance to use some CPU
3435 Description: the cat log was not initialized in recovery
3436 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3437 unlink log on mds will not be initialized until mds post
3438 recovery. And also in mds post recovery, the unlink log will
3439 initialization will be done asynchronausly, so there will be race
3440 between add unlink log and unlink log initialization.
3444 Description: brw_stats were being printed incorrectly
3445 Details : brw_stats were being printed as log2 but all of them were not
3446 recorded as log2. Also remove some code duplication arising from
3447 filter_tally_{read,write}.
3451 Frequency : rare, only in recovery.
3452 Description: ASSERTION(req->rq_type != LI_POISON) failed
3453 Details : imp_lock should be held while iterating over imp_sending_list for
3454 prevent destroy request after get timeout in ptlrpc_queue_wait.
3458 Description: replay-single.sh test 52 fails
3459 Details : A lock's skiplist need to be cleanup when it being unlinked
3460 from its resource list.
3464 Description: Short directio read returns full requested size rather than
3466 Details : Direct I/O operations should return actual amount of bytes
3467 transferred rather than requested size.
3469 Severity : enhancement
3471 Description: metadata RPC reduction (e.g. for rm performance)
3472 Details : decrease the amount of synchronous RPC between clients and servers
3473 by canceling conflicing lock before the operation on the client
3474 and packing thier handles into the main operation RPC to server.
3476 Severity : enhancement
3478 Description: Async OSC create to avoid the blocking unnecessarily.
3479 Details : If an OST has no remaining object, system will block on the
3480 creation when it needs to create a new object on this OST. Now,
3481 ways use pre-created objects when available, instead of blocking on
3482 an empty osc while others are not empty. If we must block, we
3483 block for the shortest possible period of time.
3487 Description: improve handling recoverable errors
3488 Details : if request processig with error which can be recoverable on server
3489 request should be resend, otherwise page released from cache and
3492 Severity : enhancement
3494 Description: refine locking for avoid write wrong info into lov_objid file
3495 Details : fix possible races with add new target and write/update data in
3498 --------------------------------------------------------------------------------
3500 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3505 Frequency : on some architectures
3507 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3508 Details : Uninitialized flags sometimes cause configuration commands to
3512 Frequency : patchless clients only
3514 Description: missing __iget() symbol export
3515 Details : The __iget() symbol export is missing. To avoid the need for
3516 this on patchless clients the deathrow inode reaper is turned
3517 off, and we depend on the VM to clean up old inodes. This
3518 dependency was during via the fix for bug 12181.
3520 --------------------------------------------------------------------------------
3522 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3524 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3525 INTEROPERATE with older versions automatically. In many cases a
3526 special upgrade step is needed. Please read the
3527 user documentation before upgrading any part of a live system.
3528 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3529 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3530 * WARNING: Lustre configuration and startup changes are required with
3531 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3535 Severity : enhancement
3537 Description: MountConf
3538 Details : Lustre configuration is now managed via mkfs and mount
3539 commands instead of lmc and lconf. New obd types (MGS, MGC)
3540 are added for dynamic configuration management. See
3541 https://mail.clusterfs.com/wikis/lustre/MountConf for
3544 Severity : enhancement
3546 Description: dynamic OST addition
3547 Details : OSTs can now be added to a live filesystem
3549 Severity : enhancement
3551 Description: startup order invariance
3552 Details : MDTs and OSTs can be started in any order. Clients only
3553 require the MDT to complete startup.
3555 Severity : enhancement
3557 Description: parallel, asynchronous orphan cleanup
3558 Details : orphan cleanup is now performed in separate threads for each
3559 OST, allowing parallel non-blocking operation.
3561 Severity : enhancement
3563 Description: optimized stripe assignment
3564 Details : stripe assignments are now made based on ost space available,
3565 ost previous usage, and OSS previous usage, in order to try
3566 to optimize storage space and networking resources.
3568 Severity : enhancement
3570 Description: Permanently set tunables
3571 Details : All writable /proc/fs/lustre tunables can now be permanently
3572 set on a per-server basis, at mkfs time or on a live system.
3574 Severity : enhancement
3576 Description: Lustre message v2
3577 Details : Add lustre message format v2.
3579 Severity : enhancement
3581 Description: client OST exclusion list
3582 Details : Clients can be started with a list of OSTs that should be
3583 declared "inactive" for known non-responsive OSTs.
3587 Description: SPEC SFS validation failure on NFS v2 over lustre.
3588 Details : Changes the blocksize for regular files to be 2x RPC size,
3589 and not depend on stripe size.
3591 Severity : enhancement
3593 Description: Multiple MD RPCs in flight.
3594 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3595 To avoid overly-overloading MDS, introduce a limit on number of
3596 MDS RPCs in flight for a single client and add /proc controls
3597 to adjust this limit.
3599 Severity : enhancement
3601 Description: client read/write statistics
3602 Details : Add client read/write call usage stats for performance
3603 analysis of user processes.
3604 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3605 file access. extents_stats shows chunk size distribution.
3606 extents_stats_per_process show chunk size distribution per
3609 Severity : enhancement
3611 Description: mds statistics
3612 Details : Add detailed mds operations statistics in
3613 /proc/fs/lustre/mds/*/stats.
3617 Description: Failure of copying files with lustre special EAs.
3618 Details : Client side always return success for setxattr call for lustre
3619 special xattr (currently only "trusted.lov").
3624 Description: Refcount LNET uuids
3625 Details : The global LNET uuid list grew linearly with every startup;
3626 refcount repeated list entries instead of always adding to
3629 Severity : enhancement
3631 Description: Dynamic service threads
3632 Details : Within a small range, start extra service threads
3633 automatically when the request queue builds up.
3636 Frequency : mixed-endian client/server environments
3638 Description: mixed-endian crashes
3639 Details : The new msg_v2 system had some failures in mixed-endian
3642 Severity : enhancement
3644 Description: Easy OST removal
3645 Details : OSTs can be permanently deactivated with e.g. 'lctl
3646 conf_param lustre-OST0001.osc.active=0'
3648 Severity : enhancement
3650 Description: MGS proc entries
3651 Details : Added basic proc entries for the MGS showing what filesystems
3654 Severity : enhancement
3656 Description: provide MGS failover
3657 Details : Added config lock reacquisition after MGS server failover.
3659 Severity : enhancement
3661 Description: add Linux 2.4 support
3662 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3666 Description: a large application tries to do I/O to the same resource and dies
3667 in the middle of it.
3668 Details : Check the req->rq_arrival time after the call to
3669 ost_brw_lock_get(), but before we do anything about
3670 processing it & sending the BULK transfer request. This
3671 should help move old stale pending locks off the queue as
3672 quickly as obd_timeout.
3675 Frequency : when an incorrect nid is specified during startup
3677 Description: ptlrpc connect to non-existant node causes kernel crash
3678 Details : LNET can't be re-entered from an event callback, which
3679 happened when we expire a message after the export has been
3680 cleaned up. Instead, hand the zombie cleanup off to another
3683 Severity : enhancement
3685 Description: plain/inodebits lock performance improvement
3686 Details : Grouping plain/inodebits in granted list by their request modes
3687 and bits policy, thus improving the performance of search through
3691 Frequency : only if OST filesystem is corrupted
3693 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3694 Details : for a short time RPCs with bulk IO are in the replay list,
3695 but replay of bulk IOs is unimplemented. If the OST filesystem
3696 is corrupted due to disk cache incoherency and then replay is
3697 started it is possible to trip an assertion. Avoid putting
3698 committed RPCs into the replay list at all to avoid this issue.
3701 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3703 Description: System hang on startup
3704 Details : This bug allowed the liblustre (e.g. catamount) client to
3705 return to the app before handling all startup RPCs. This
3706 could leave the node unresponsive to lustre network traffic
3707 and manifested as a server ptllnd timeout.
3709 Severity : enhancement
3711 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3712 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3713 boolean to control whether to print peer debug info when a
3714 client's RPC times out.
3717 Frequency : only for kernels with patches from Lustre below 1.4.3
3719 Description: Remove old rdonly API
3720 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3723 Frequency : only for devices with external journals
3725 Description: Set external device read-only also
3726 Details : During a commanded failover stop, we set the disk device
3727 read-only while the server shuts down. We now also set any
3728 external journal device read-only at the same time.
3731 Frequency : when upgrading from 1.4 while trying to change parameters
3733 Description: The wrong (new) MDC name was used when setting parameters for
3734 upgraded MDT's. Also allows changing of OSC (and MDC)
3735 parameters if --writeconf is specified at tunefs upgrade time.
3738 Frequency : when setting specific ost indicies
3740 Description: QOS code breaks on skipped indicies
3741 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3742 created with --index need not be sequential.
3744 Severity : enhancement
3746 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3747 Details : The uninit_groups feature works in conjunction with the kernel
3748 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3749 up the pass1 processing of e2fsck. This is a read-only feature
3750 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3751 mount filesystems that have had this feature enabled.
3753 Severity : enhancement
3755 Description: Improve multi-block allocation algorithm to avoid fragmentation
3756 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3757 allocation locality and avoid filesystem fragmentation.
3759 ------------------------------------------------------------------------------
3761 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3763 * Support for kernels:
3764 2.4.21-47.0.1.EL (RHEL 3)
3765 2.6.5-7.283 (SLES 9)
3766 2.6.9-42.0.10.EL (RHEL 4)
3767 2.6.12.6 vanilla (kernel.org)
3768 2.6.16.27-0.9 (SLES 10)
3769 * Recommended e2fsprogs version: 1.39.cfs5
3771 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3775 Frequency : occasional, depends on client load and configuration
3776 Bugzilla : 12181, 12203
3777 Description: data loss for recently-modified files
3779 Details : In some cases it is possible that recently written or created
3780 files may not be written to disk in a timely manner (this should
3781 normally be within 30s unless client IO load is very high).
3782 The problem appears as zero-length files or files that are a
3783 multiple of 1MB in size after a client crash or client eviction
3784 that are missing data at the end of the file.
3786 This problem is more likely to be hit on clients where files are
3787 repeatedly created and unlinked in the same directory, clients
3788 have a large amount of RAM, have many CPUs, the filesystem has
3789 many OSTs, the clients are rebooted frequently, and/or the files
3790 are not accessed by other nodes after being written.
3792 The presence of the problem can be detected by looking at
3793 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3794 smaller than the second (nr_unused) then dirty files will not
3795 be flushed automatically to disk. "sync; sleep 10" should be
3796 run several times on the node before unmounting it to update
3797 Lustre (this is also safe to run on nodes without this problem).
3799 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3800 that can cause this same problem, so customers using that kernel
3801 also need to update the kernel in addition to Lustre. In order
3802 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3804 It is normal that files written just before a client crash (less
3805 than 30s) may not yet have been flushed to disk, even for local
3809 Frequency : frequent on thin XT3 nodes
3811 Description: UUID collision on thin XT3 Linux nodes
3812 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3813 randomly, since we relied on an insufficiently-seeded PRNG.
3818 Description: OSS hangs after "All ost request buffers busy"
3819 Details : A deadlock between quota and journal operations caused OSS
3820 hangs after printing "All ost request buffers busy."
3823 Frequency : always on liblustre builds
3825 Description: Cleanup compiler warnings on liblustre
3828 Frequency : always on liblustre builds on XT3
3830 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3834 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3835 > 32000 subdirectories
3836 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3837 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3838 subdirectory count crosses 32000. This will aid e2fsck to
3839 correctly handle more than 32000 subdirectories.
3842 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3844 Description: System hang on startup
3845 Details : This bug allowed the liblustre (e.g. catamount) client to
3846 return to the app before handling all startup RPCs. This
3847 could leave the node unresponsive to lustre network traffic
3848 and manifested as a server ptllnd timeout.
3850 Severity : enhancement
3852 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3853 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3854 boolean to control whether to print peer debug info when a
3855 client's RPC times out.
3860 Description: make O_SYNC working on 2.6 kernels
3861 Details : 2.6 kernels use different method for mark pages for write,
3862 so need add a code to lustre for O_SYNC work.
3867 Description: Failure to close file and release space on NFS
3868 Details : Put inode details into lock acquired in ll_intent_file_open.
3869 Use mdc_intent_lock in ll_intent_open to properly
3870 detect all kind of errors unhandled by mdc_enqueue.
3875 Description: proc file read during shutdown sometimes raced obd removal,
3877 Details : Add lock to prevent obd access after proc file removal.
3880 Frequency : Only for files larger than 4GB on 32-bit clients.
3882 Description: improperly doing page alignment of locks
3883 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3884 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3889 Description: RPCs being resent when they shouldn't be
3890 Details : Some RPCs that should not be resent are being resent. This
3891 can cause inconsistencies in the RPC state machine. Do not
3892 resend such requests.
3895 Frequency : rare, only with NFS export
3897 Description: Crash on NFS re-export node
3898 Details : under very unusual load conditions an assertion is hit in
3899 ll_intent_file_open()
3902 Frequency : only if OST filesystem is corrupted
3904 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3905 Details : for a short time RPCs with bulk IO are in the replay list,
3906 but replay of bulk IOs is unimplemented. If the OST filesystem
3907 is corrupted due to disk cache incoherency and then replay is
3908 started it is possible to trip an assertion. Avoid putting
3909 committed RPCs into the replay list at all to avoid this issue.
3914 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3915 Details : Large single O_DIRECT read and write calls can fail to allocate
3916 a sufficiently large buffer to process the request. In case of
3917 allocation failure the allocation is retried with a smaller
3918 buffer and broken into smaller requests.
3920 Severity : enhancement
3922 Description: Add -o localflock option to simulate old noflock behaviour.
3923 Details : This will achieve local-only flock/fcntl locks coherentness.
3928 Description: versioning check is incomplete
3929 Details : Checking the version difference of client vs. server, report
3930 error if the gap is too big.
3935 Description: add support PG_writeback bit
3936 Details : add support for PG_writeback bit for Lustre, for more carefull
3937 work with page cache in 2.6 kernel. This also fix some deadlocks
3938 and remove hack for work O_SYNC with 2.6 kernel.
3940 Severity : enhancement
3942 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3943 Details : The uninit_groups feature works in conjunction with the kernel
3944 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3945 up the pass1 processing of e2fsck. This is a read-only feature
3946 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3947 mount filesystems that have had this feature enabled.
3949 Severity : enhancement
3951 Description: Improve multi-block allocation algorithm to avoid fragmentation
3952 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3953 allocation locality and avoid filesystem fragmentation.
3955 ------------------------------------------------------------------------------
3957 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3959 * Support for kernels:
3960 2.6.9-42.0.3.EL (RHEL 4)
3961 2.6.5-7.276 (SLES 9)
3962 2.4.21-47.0.1.EL (RHEL 3)
3963 2.6.12.6 vanilla (kernel.org)
3964 2.6.16.21-0.8 (SLES10)
3965 * Recommended e2fsprogs version: 1.39.cfs2-0
3967 * The backwards-compatible /proc/sys/portals symlink has been removed
3968 in this release. Before upgrading, please ensure that you change
3969 any configuration scripts or /etc/sysctl.conf files that access
3970 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3971 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3972 in advance of the upgrade on any system running Lustre 1.4.6 or
3973 newer, since /proc/sys/lnet was added in that version.
3974 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3978 Frequency : only when quota is used
3980 Description: avoid scanning export list for quota master
3981 Details : Change the algorithms to avoid scanning export list in order
3982 to improve the efficiency.
3985 Frequency : MDS failover only, very rarely
3987 Description: "went back in time" messages on mds failover
3988 Details : The greatest transno may be lost when the current operation
3989 finishes with an error (transno==0) and the client's last_rcvd
3990 record is over-written. Save the greatest transno in the
3991 mds_last_transno for this case.
3994 Frequency : always for specific kernels and striping counts
3996 Description: client may get "Matching packet too big" without ACL support
3997 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3998 when trying to access files in certain configurations. The
3999 clients should in fact be denied when mounting because they do
4000 not understand ACLs.
4003 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
4005 Description: many clients connecting with IO in progress causes connect timeouts
4006 Details : Avoid synchronous journal commits to avoid delays caused by many
4007 clients connecting/disconnecting when bulk IO is in progress.
4008 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
4009 OST_IO_PORTAL to avoid delays behind potentially many pending
4013 Frequency : occasionally with multiple writers to a single file
4015 Description: shared writes to file may result in wrong size reported by stat()
4016 Details : Allow growing of kms when extent lock is cancelled
4019 Frequency : always with random mmap IO to multi-striped file
4021 Description: mmap write might be lost if we are writing to a 'hole' in stripe
4022 Details : Only if the hole is at the end of OST object so that kms is too
4023 small. Fix is to increase kms accordingly in ll_nopage.
4026 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
4028 Description: writes to a missing object would leak memory on the OST
4029 Details : If there is an inconsistency between the MDS and OST filesystems,
4030 such that the MDS references an object that doesn't exist, writes
4031 to that object will leak memory due to incorrect cleanup in the
4032 error handling path, eventually running out of memory on the OST.
4037 Description: Creating too long symlink causes lustre errors
4038 Details : Check symlink and name lengths before sending requests to MDS.
4041 Frequency : only if flock is enabled (not on by default)
4043 Description: posix locks not released on fd closure on 2.6.9+
4044 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4045 kernels, this caused such locks not to be released on fd close and
4046 then assertions on fs unmount about still used locks.
4049 Frequency : MDS failover only, very rarely
4051 Description: clients may get ASSERTION(granted_lock != NULL)
4052 Details : When request was taking a long time, and a client was resending
4053 a getattr by name lock request. The were multiple lock requests
4054 with the same client lock handle and
4055 mds_getattr_name->fixup_handle_for_resent_request found one of the
4056 lock handles but later failed with ASSERTION(granted_lock != NULL).
4061 Description: handle->h_buffer_credits > 0, assertion failure
4062 Details : h_buffer_credits is zero after truncate, causing assertion
4063 failure. This patch extends the transaction or creates a new
4067 Frequency : NFS re-export or patchless client
4068 Bugzilla : 11179, 10796
4069 Description: Crash on NFS re-export node (__d_move)
4070 Details : We do not want to hash the dentry if we don't have a lock.
4071 But if this dentry is later used in d_move, we'd hit uninitialised
4072 list head d_hash, so we just do this to init d_hash field but
4073 leave dentry unhashed.
4076 Frequency : NFS re-export or patchless client
4078 Description: NFS exports has problem with symbolic link
4079 Details : lustre client didn't properly install dentry when re-exported
4080 to NFS or running patchless client.
4083 Frequency : NFS re-export or patchless client
4085 Description: Various nfs/patchless fixes.
4086 Details : fixes reuse disconected alias for lookup process - this fixes
4087 warning "find_exported_dentry: npd != pd",
4088 fix permission error with open files at nfs.
4089 fix apply umask when do revalidate.
4092 Frequency : occasional
4094 Description: Crash on NFS re-export node
4095 Details : calling clear_page() on the wrong pointer triggered oops in
4096 generic_mapping_read().
4099 Frequency : rarely, using O_DIRECT IO
4101 Description: unaligned directio crashes client with LASSERT
4102 Details : check for unaligned buffers before trying any requests.
4105 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4107 Description: stale data returned from RAID cache
4108 Details : If only a small amount of IO is done to the RAID device before
4109 reading it again it is possible to get stale data from the RAID
4110 cache instead of reading it from disk.
4113 Frequency : always for sles10 kernel
4115 Description: sles10 support
4116 Details : ll_follow_link: compile fixes and using of nd_set_link
4117 under newer kernels.
4120 Frequency : depends on arch, kernel and compiler version, always on sles10
4123 Description: recursive or deep enough symlinks cause stack overflow
4124 Details : getting rid of large stack-allocated variable in
4128 Frequency : depends on hardware
4130 Description: lustre write performance loss in the SLES10 kernel
4131 Details : the performance loss is caused by using of write barriers in the
4132 ext3 code. The SLES10 kernel turns barrier support on by
4133 default. The fix is to undo that change for ldiskfs.
4135 ------------------------------------------------------------------------------
4137 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4139 * Support for kernels:
4140 2.6.9-42.0.3EL (RHEL 4)
4141 2.6.5-7.276 (SLES 9)
4142 2.4.21-47.0.1.EL (RHEL 3)
4143 2.6.12.6 vanilla (kernel.org)
4147 Frequency : quota enabled and large files being deleted
4149 Description: releasing more than 4GB of quota at once hangs OST
4150 Details : If a user deletes more than 4GB of files on a single OST it
4151 will cause the OST to spin in an infinite loop. Release
4152 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4157 Description: statfs data retrieved from /proc may be stale or zero
4158 Details : When reading per-device statfs data from /proc, in the
4159 {kbytes,files}_{total,free,avail} files, it may appear
4160 as zero or be out of date.
4163 Frequency : systems with MD RAID1 external journal devices
4165 Description: lconf's call to blkid is confused by RAID1 journal devices
4166 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4167 of returning all block devices that match the journal UUID.
4170 Frequency : always, for aggregate stripe size over 4GB
4172 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4173 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4174 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4177 Frequency : always if "lfs find" used on a local file/directory
4179 Description: "lfs find" segfaults if used on a local file/directory
4180 Details : The case where a directory component was not specified wasn't
4181 handled correctly. Handle this properly.
4184 Frequency : always on ppc64
4186 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4187 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4188 perform endianness conversion.
4191 Frequency : rarely (truncate to non-zero file size after write under load)
4192 Bugzilla : 10730, 10687
4193 Description: Files padded with zeros to next 4K multiple
4194 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4195 it is possible that files that are truncated to a non-zero size
4196 immediately after being written are filled with zero bytes beyond
4197 the truncated size. No file data is lost.
4199 Severity : enhancement
4201 Description: Allow recovery/failover for liblustre clients.
4202 Details : liblustre clients were unaware of failover configurations until
4205 Severity : enhancement
4207 Description: user file locks should fail when not mounting with flock option
4208 Details : Set up an error-returning stub in ll_file_operations.lock field
4209 to prevent incorrect behaviour when client is mounted without
4210 flock option. Also, set up properly f_op->flock field for
4214 Frequency : always on ia64
4216 Description: "lfs df" loops on printing out MDS statfs information
4217 Details : The obd_ioctl_data was not initialized and in some systems
4218 this caused a failure during the ioctl that did not return
4219 an error. Initialize the struct and return an error on failure.
4222 Frequency : SLES 9 only
4224 Description: Error of copying files with lustre special EAs as root
4225 Details : Client side always return success for setxattr call for lustre
4226 special xattr (currently only "trusted.lov").
4229 Frequency : rarely on clusters with both ia64+i386 clients
4231 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4232 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4233 clients are doing concurrent writes to the same file it is
4234 possible that the ia64 clients may LASSERT because the OST
4235 extent locks are not PAGE_SIZE aligned. Ensure that grown
4236 locks are always aligned on the request boundary.
4239 Frequency : specific use, occasional
4241 Description: Overwriting in use executable truncates on-disk binary image
4242 Details : If one node attempts to overwrite an executable in use by
4243 another node, we now correctly return ETXTBSY instead of
4244 truncating the file.
4246 Severity : enhancement
4248 Description: Async OSC create to avoid the blocking unnecessarily.
4249 Details : If an OST has no remaining object, system will block on the creating
4250 when need to create a new object on this OST. Now, ways use
4251 pre-created objects when available, instead of blocking on an
4252 empty osc while others are not empty. If we must block, we block
4253 for the shortest possible period of time.
4258 Description: chmod on Lustre root is propagated to other clients
4259 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4260 invalid by the follow_mount time.
4263 Frequency : liblustre clients only
4265 Description: Race in 'instant cancel' lock handling could lead to such locks
4266 never to be granted in case of SMP MDS
4267 Details : Do not destroy not yet granted but cbpending locks in
4271 Frequency : replay/resend of open
4273 Description: non null lock assetion failure in mds_intent_policy
4274 Details : Trying to replay/resend lockless open requests resulted in
4275 mds_open() returning 0 with no lock. Now it sets a flag if
4276 a lock is going to be returned.
4278 Severity : enhancement
4280 Description: Checksum enhancements
4281 Details : New checksum enhancements allow for resending RPCs that failed
4284 Severity : enhancement
4286 Description: Tunables on number of dirty pages in cacche
4287 Details : Allow to set limit on number of dirty pages cached.
4292 Description: client crash on unmount - lock still has references
4293 Details : In some error handling cases it was possible to leak a lock
4294 reference on a client while accessing a file. This was not
4295 harmful to the client during operation, but would cause the
4296 client to crash when the filesystem is unmounted.
4299 Frequency : specific case, rare
4301 Description: ETXTBSY on mds though file not in use by client
4302 Details : ETXTBSY is no longer incorrectly returned when attempting to
4303 chmod or chown a directory that the user previously tried to
4304 execute or a currently-executing binary.
4307 Frequency : extremely rare except on liblustre-based clients
4309 Description: Lustre space not freed when files are deleted
4310 Details : Clean up open-unlinked files after client eviction. Previously
4311 the unlink was skipped and the files remained as orphans.
4316 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4317 Details : In some cases it was possible to send a blocking callback to
4318 a client doing a glimpse, even though that client didn't get
4319 a lock granted. When the glimpse lock is cancelled on the OST
4320 the freed lock is left on the waiting list and corrupted the list.
4323 Frequency : all core dumps
4325 Description: Broke core dumps to lustre
4326 Details : Negative dentry may be unhashed if parent does not have UPDATE
4327 lock, but some callers, e.g. do_coredump, expect dentry to be
4328 hashed after successful create, hash it in ll_create_it.
4330 ------------------------------------------------------------------------------
4332 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4334 * Support for kernels:
4335 2.6.9-42.0.2.EL (RHEL 4)
4336 2.6.5-7.276 (SLES 9)
4337 2.4.21-40.EL (RHEL 3)
4338 2.6.12.6 vanilla (kernel.org)
4342 Frequency : always on RHEL 3
4344 Description: Number of open files grows over time
4345 Details : The number of open files grows over time, whether or not
4346 Lustre is started. This was due to a filp leak introduced
4347 by one of our kernel patches.
4349 ------------------------------------------------------------------------------
4351 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4353 * Support for kernels:
4354 2.6.9-42.EL (RHEL 4)
4355 2.6.5-7.276 (SLES 9)
4356 2.4.21-40.EL (RHEL 3)
4357 2.6.12.6 vanilla (kernel.org)
4362 Bugzilla : 5719, 9635, 9792, 9684
4363 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4364 Details : If a server is under heavy load and cannot reply to new
4365 connection requests before the client resends the (re)connect,
4366 the connection handling code can behave badly if two service
4367 threads are concurrently handing separate (re)connections from
4368 the same client. Add better locking to the connection handling
4369 code, and ensure that only a single connection will be processed
4370 for a given client UUID, even if the lock is dropped.
4372 Severity : enhancement
4374 Description: add TCP zero-copy support to kernel
4375 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4376 sends if the hardware supports scatter-gather and checksumming.
4377 This allows socklnd to do client-write and server-read more
4378 efficiently and reduce CPU utilization from skbuf copying.
4381 Frequency : only if NFS exporting from client
4383 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4384 Details : If setting attributes on a file created under NFS that had
4385 never been opened it would be possible to oops the client
4386 if the file had no objects.
4389 Frequency : always for liblustre
4391 Description: liblustre client does MDS+OSTs setattr RPC for each write
4392 Details : When doing a write from a liblustre client, the client
4393 incorrectly issued an RPC to the MDS and each OST the file was
4394 striped over in order to update the timestamps. When writing
4395 with small chunks and many clients this could overwhelm the MDS
4396 with RPCs. In all cases it would slow down the write because
4397 these RPCs are unnecessary.
4399 Severity : enhancement
4401 Description: allow number of MDS service threads to be changed at module load
4402 Details : It is now possible to change the number of MDS service threads
4403 running. Adding "options mds mds_num_threads={N}" to the MDS's
4404 /etc/modprobe.conf will set the number of threads for the next
4405 time Lustre is restarted (assuming the "mds" module is also
4406 reloaded at that time). The default number of threads will
4407 stay the same, 32 for most systems.
4412 Description: OST crash if filesystem is unformatted or corrupt
4413 Details : If an OST is started on a device that has never been formatted
4414 or if the filesystem is corrupt and cannot even mount then the
4415 error handling cleanup routines would dereference a NULL pointer.
4420 Description: NULL pointer deref in llap_from_page.
4421 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4422 incorrect) mapping if the page was truncated/reclaimed while it was
4423 searched for. Check for this condition and skip such pages when
4424 doing readahead. Introduce extra check to llap_from_page() to
4425 verify page->mapping->host is non-NULL (so page is not anonymous).
4428 Frequency : Sometimes when using sys_sendfile
4430 Description: "page not covered by a lock" warnings from ll_readpage
4431 Details : sendfile called ll_readpage without right page locks present.
4432 Now we introduced ll_file_sendfile that does necessary locking
4433 around call to generic_file_sendfile() much like we do in
4437 Frequency : with certain MDS communication failures at client mount time
4439 Description: NULL pointer deref after failed client mount
4440 Details : a client connection request may delayed by the network layer
4441 and not be sent until after the PTLRPC layer has timed out the
4442 request. If the client fails the mount immediately it will try
4443 to clean up before the network times out the request. Add a
4444 reference from the request import to the obd device and delay
4445 the cleanup until the network drops the request.
4448 Frequency : occasionally during client (re)connect
4450 Description: assertion failure during client (re)connect
4451 Details : processing a client connection request may be delayed by the
4452 client or server longer than the client connect timeout. This
4453 causes the client to resend the connection request. If the
4454 original connection request is replied in this interval, the
4455 client may trip an assertion failure in ptlrpc_connect_interpret()
4456 which thought it would be the only running connect process.
4459 Frequency : only with obd_echo servers and clients that are rebooted
4461 Description: kernel BUG accessing uninitialized data structure
4462 Details : When running an obd_echo server it did not start the ping_evictor
4463 thread, and when a client was evicted an uninitialized data
4464 structure was accessed. Start the ping_evictor in the RPC
4465 service startup instead of the OBD startup.
4467 Severity : enhancement
4468 Bugzilla : 10193 (patchless)
4469 Description: Remove dependency on various unexported kernel interfaces.
4470 Details : No longer need reparent_to_init, exit_mm, exit_files,
4471 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4474 Frequency : rare (only users of deprecated and unsupported LDAP config)
4476 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4477 Details : LDAP apparently contains 'lustreName' attributes instead of
4478 'name'. A simple remapping of the name is sufficient.
4481 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4483 Description: waiting_locks_callback trips kernel BUG if client is evicted
4484 Details : Running with the dump_on_timeout debug flag turned on makes
4485 it possible that the waiting_locks_callback() can try to dump
4486 the Lustre kernel debug logs from an interrupt handler. Defer
4487 this log dumping to the expired_lock_main() thread.
4489 Severity : enhancement
4491 Description: Support NFS exporting on 2.6 kernels.
4492 Details : Implement non-rawops metadata methods for NFS server to use without
4493 changing NFS server code.
4496 Frequency : very rare (synthetic metadata workload only)
4498 Description: two racing renames might cause an MDS thread to deadlock
4499 Details : Running the "racer" program may cause one MDS thread to rename
4500 a file from being the source of a rename to being the target of
4501 a rename at exactly the same time that another thread is doing
4502 so, and the second thread has already enqueued these locks after
4503 doing a lookup of the target and is trying to relock them in
4504 order. Ensure that we don't try to re-lock the same resource.
4507 Frequency : only very large systems with liblustre clients
4509 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4510 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4511 all OSTs in parallel. This allows the eviction of stale liblustre
4512 clients to proceed much faster than if they were done in series,
4513 and also offers similar improvements for other set_info RPCs.
4518 Description: excessive CPU usage during initial read phase on client
4519 Details : During the initial read phase on a client, it would agressively
4520 retry readahead on the file, consuming too much CPU and impacting
4521 performance (since 1.4.5.8). Improve the readahead algorithm
4522 to avoid this, and also improve some other common cases (read
4523 of small files in particular, where "small" is files smaller than
4524 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4529 Description: MDS crash when receiving packet with unknown intent.
4530 Details : Do not LBUG in unknown intent case, just return -EFAULT
4532 Severity : enhancement
4533 Bugzilla : 9293, 9385
4534 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4535 Details : Do not serialize getattr (non-intent version) and statfs.
4538 Frequency : occasional, when OST network is overloaded/intermittent
4540 Description: client evicted by OST after bulk IO timeout
4541 Details : If a client sends a bulk IO request (read or write) the OST
4542 may evict the client if it is unresposive to its data GET/PUT
4543 request. This is incorrect if the network is overloaded (takes
4544 too long to transfer the RPC data) or dropped the OST GET/PUT
4545 request. There is no need to evict the client at all, since
4546 the pinger and/or lock callbacks will handle this, and the
4547 client can restart the bulk request.
4550 Frequency : Always when mmapping file with no objects
4552 Description: client crashes when mmapping file with no objects
4553 Details : Check that we actually have objects in a file before doing any
4554 operations on objects in ll_vm_open, ll_vm_close and
4560 Description: Request leak when working with deleted CWD
4561 Details : Introduce advanced request refcount tracking for requests
4562 referenced from lustre intent.
4564 Severity : Enhancement
4566 Description: Cache open file handles on client.
4567 Details : MDS now will return special lock along with openhandle, if
4568 requested and client is allowed to hold openhandle, even if unused,
4569 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4570 closing files for every read/write openration.
4572 Severity : Enhancement
4574 Description: Cache open negative dentries on client when possible.
4575 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4576 negative dentries on lock revocation.
4581 Description: Remounting a client read-only wasn't possible with a zconf mount
4582 Details : It wasn't possible to remount a client read-only with llmount.
4584 Severity : enhancement
4585 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4586 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4587 MPICH is included to add Lustre-specific ADIO interfaces.
4588 This is based closely on the UFS ADIO layer and only differs
4589 in file creation, in order to allow the OST striping to be set.
4590 This is user-contributed code and not supported by CFS.
4595 Description: extended inode attributes (immutable, append-only) work improperly
4596 when 2.4 and 2.6 kernels are used on client/server or vice versa
4597 Details : Introduce kernel-independent values for these flags.
4599 Severity : enhancement
4602 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4603 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4604 granularity. Now, Fractional MB granularity is be supported,
4605 this is very useful for low memory system.
4607 Severity : enhancement
4609 Description: Getattr by fid
4610 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4611 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4612 have locks and avoids getattr rpc requests for it on every stat.
4615 Frequency : Always, for filesystems larger than 2TB
4617 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4618 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4619 filesystems larger than 2TB to be mounted with mballoc enabled.
4622 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4624 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4625 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4626 filesystems larger than 2TB is mounted then the kernel will
4627 silently corrupt the start of the filesystem. CONFIG_LBD is
4628 enabled for all CFS-supported kernels, but the possibility of
4629 this happening with a modified kernel config exists.
4631 Severity : enhancement
4633 Description: add client O_DIRECT support for 2.6 kernels
4634 Details : It is now possible to do O_DIRECT reads and writes to files
4635 in the Lustre client mountpoint on 2.6 kernel clients.
4637 Severity : enhancement
4639 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4640 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4641 parallel, not waiting for response from every OST before sending
4642 a rpc to the next OST.
4647 Description: setattr vs write race when updating file timestamps
4648 Details : Client processes that update a file timestamp into the past
4649 right after writing to the file (e.g. tar) it is possible that
4650 the updated file modification time can be reset to the current
4651 time due to a race between processing the setattr and write RPC.
4653 Severity : enhancement
4655 Description: Bring 'lfs find' closer in line with regular Linux find.
4656 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4657 -print0 options and obtains all the needed info through the lustre
4660 Severity : enhancement
4662 Description: support up to 1024 configured devices on one node
4663 Details : change obd_dev array from statically allocated to dynamically
4664 allocated structs as they are first used to reduce memory usage
4669 Description: Flush dirty partially truncated pages during truncate
4670 Details : Immediatelly flush partially truncated pages in filter_setattr,
4671 this way we completely avoid having any pages in page cache on OST
4672 and can retire ugly workarounds during writes to flush such pages.
4677 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4678 Details : protect inode from truncation within vfs_unlink() context
4679 just take a reference before calling vfs_unlink() and release it
4680 when parent's i_sem is free.
4685 Description: last_id value checked outside lock on OST caused LASSERT failure
4686 Details : If there were multiple MDS->OST object precreate requests in
4687 flight, it was possible that the OST's last object id was checked
4688 outside a lock and incorrectly tripped an assertion. Move checks
4689 inside locks, and discard old precreate requests.
4692 Frequency : always, if extents are used on OSTs
4694 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4695 Details : OSTs using the extents format would not zero the high 16 bits of
4696 the index physical block number. This is not a problem for any
4697 OST filesystems smaller than 16TB, and no kernels support ext3
4698 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4699 new/modified files) and can be fixed for existing filesystems
4700 with e2fsprogs-1.39-cfs1.
4705 Description: import connection selection may be incorrect if timer wraps
4706 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4707 import connections to be ignored if the 32-bit jiffies counter
4708 wraps. Use a 64-bit jiffies counter.
4711 Frequency : very large clusters immediately after boot
4713 Description: LNET request buffers exhausted under heavy short-term load
4714 Details : If a large number of client requests are generated on a service
4715 that has previously never seen so many requests it is possible
4716 that the request buffer growth cannot keep up with the spike in
4717 demand. Instead of dropping incoming requests, they are held in
4718 the LND until the RPC service can accept more requests.
4721 Frequency : Sometimes during replay
4723 Description: Assertion failure in ll_local_open after replay.
4724 Details : If replay happened on an open request reply before we were able
4725 to set replay handler, reply will become not swabbed tripping the
4726 assertion in ll_local_open. Now we set the handler right after
4727 recognising of open request
4730 Frequency : very rare
4732 Description: kernel reports "badness in vsnprintf"
4733 Details : Reading from the "recovery_status" /proc file in small chunks
4734 may cause a negative length in lprocfs_obd_rd_recovery_status()
4735 call to vsnprintf() (which is otherwise harmless). Exit early
4736 if there is no more space in the output buffer.
4738 Severity : enhancement
4740 Description: clear OBD RPC statistics by writing to them
4741 Details : It is now possible to clear the OBD RPC statistics by writing
4742 to the "stats" file.
4747 Description: Client mtime is not the same on different clients after utimes
4748 Details : In some cases, the client was using the utimes() syscall on
4749 a file cached on another node. The clients now validate the
4750 ctime from the MDS + OSTs to determine which one is right.
4755 Description: Inability to activate failout mode
4756 Details : lconf script incorrectly assumed that in pythong string's numeric
4757 value is used in comparisons.
4760 Frequency : always with multiple stripes per file
4762 Description: Inefficient object allocation for mutli-stripe files
4763 Details : When selecting which OSTs to stripe files over, for files with
4764 a stripe count that divides evenly into the number of OSTs,
4765 the MDS is always picking the same starting OST for each file.
4766 Return the OST selection heuristic to the original design.
4771 Description: mount failures may take full timeout to return an error
4772 Details : Under some heavy load conditions it is possible that a
4773 failed mount can wait for the full obd_timeout interval,
4774 possibly several minutes, before reporting an error.
4775 Instead return an error as soon as the status is known.
4777 Frequency : quota enabled and large files being deleted
4779 Description: releasing more than 4GB of quota at once hangs OST
4780 Details : If a user deletes more than 4GB of files on a single OST it
4781 will cause the OST to spin in an infinite loop. Release
4782 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4787 Description: statfs data retrieved from /proc may be stale or zero
4788 Details : When reading per-device statfs data from /proc, in the
4789 {kbytes,files}_{total,free,avail} files, it may appear
4790 as zero or be out of date.
4793 Frequency : systems with MD RAID1 external journal devices
4795 Description: lconf's call to blkid is confused by RAID1 journal devices
4796 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4797 of returning all block devices that match the journal UUID.
4800 Frequency : always, for aggregate stripe size over 4GB
4802 Description: assertion fails when trying to use 4GB stripe size
4803 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4804 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4807 Frequency : always on ppc64
4809 Description: the first write on an ext3 filesystem with mballoc got stuck
4810 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4811 perform endianness conversion.
4813 ------------------------------------------------------------------------------
4815 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4817 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4818 INTEROPERATE with older versions automatically. Please read the
4819 user documentation before upgrading any part of a live system.
4820 * WARNING: Lustre networking configuration changes are required with
4821 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4824 * Support for kernels:
4825 2.6.9-22.0.2.EL (RHEL 4)
4826 2.6.5-7.244 (SLES 9)
4827 2.6.12.6 vanilla (kernel.org)
4830 Severity : enhancement
4831 Bugzilla : 7981/8208
4832 Description: Introduced Lustre Networking (LNET)
4833 Details : LNET is new networking infrastructure for Lustre, it includes
4834 a reorganized network configuration mode (see the user
4835 documentation for full details) as well as support for routing
4836 between different network fabrics. Lustre Networking Devices
4837 (LNDs) for the supported network fabrics have also been
4838 created for this new infrastructure.
4840 Severity : enhancement
4841 Description: Introduced Access control lists
4842 Details : clients can set ACLs on files and directories in order to have
4843 more fine-grained permissions than the standard Unix UGO+RWX.
4844 The MDS must be started with the "-o acl" mount option.
4846 Severity : enhancement
4847 Description: Introduced filesystem quotas
4848 Details : Administrators may now establish per-user quotas on the
4851 Severity : enhancement
4853 Description: Configuration change for the XT3
4854 The PTLLND is now used to run Lustre over Portals on the XT3
4855 The configure option(s) --with-cray-portals are no longer used.
4856 Rather --with-portals=<path-to-portals-includes> is used to
4857 enable building on the XT3. In addition to enable XT3 specific
4858 features the option --enable-cray-xt3 must be used.
4863 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4864 Details : A race between allocating a new llcd and re-getting the llcd_lock
4865 allowed another thread to grab newly-allocated llcd.
4867 Severity : enhancement
4869 Description: 2.6 OST async journal commit and locking fix to improve performance
4870 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4871 2.6 kernels are now async as they already were in 2.4 kernels so
4872 that they can commit concurrently with the network bulk transfer.
4873 For block-allocated files the filter allocation semaphore is held
4874 to avoid filesystem fragmentation during allocation. BKL lock
4875 removed for 2.6 xattr operations where it is no longer needed.
4880 Description: lconf incorrectly determined whether two IP networks could talk
4881 Details : In some more complicated routing and multiple-network
4882 configurations, lconf will avoid trying to make a network
4883 connection to a disjoint part of the IP space. It was doing the
4884 math incorrectly for one set of cases.
4889 Description: Fix for potential infinite loop processing records in an llog.
4890 Details : If an llog record is corrupted/zeroed, it is possible to loop
4891 forever in llog_process(). Validate the llog record length
4892 and skip the remainder of the block on error.
4895 Frequency : occasional (liblustre only)
4897 Description: liblustre could not open files whose last component is a symlink
4898 Details : sysio_path_walk() would incorrectly pass the open intent to
4899 intermediate path components.
4902 Frequency : rare (liblustre only with non-standard tuning)
4903 Bugzilla : 7201 (7350)
4904 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4905 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4906 after it has been released, only do this for referenced locks
4908 Severity : enhancement
4910 Description: specifying an (invalid) directory default stripe_size of -1
4911 would reset the directory default striping
4912 Details : stripe_size -1 was used internally to signal directory stripe
4913 removal, now use "all default" to signal dir stripe removal
4914 as a directory striping of "all default" is not useful
4917 Frequency : common for large clusters running liblustre clients
4919 Description: doing an ls when liblustre clients are running is slow
4920 Details : sending a glimpse AST to a liblustre client waits for every AST
4921 to time out, as liblustre clients will not respond. Since they
4922 cannot cache data we refresh the OST lock LVB from disk instead.
4924 Severity : enhancement
4926 Description: doing an ls at the same time as file IO can be slow
4927 Details : enqueue and other "small" requests can be blocked behind many
4928 large IO requests. Create a new OST IO portal for non-IO
4929 requests so they can be processed faster.
4932 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4934 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4935 created file with mode 000
4936 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4937 allow an ftruncate() as a non-root user to a file with mode 000.
4938 Lustre now respects this flag to disable mode checks when
4939 truncating a file owned by the user
4942 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4945 Description: Revoking locks from clients that went dead or catatonic might take
4947 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4948 cancellation of such locks instant on servers without waiting for
4949 any reply from clients. Clients drops these locks when cancel
4950 notification from server is received without replying.
4953 Frequency : liblustre-only, when liblustre client dies or becomes busy
4955 Description: Doing ls on Linux clients can take a long time with active
4957 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4958 granting such locks to it in the first place if possible. Locks
4959 are taken by proxy on the OST during the read or write and
4960 dropped immediately afterward. Add connect flags handling, do
4961 not grant locks to liblustre clients for glimpse ASTs.
4963 Severity : enhancement
4965 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4966 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4967 sequential reads and behaves badly with applications doing random
4968 reads. Improve it by reading ahead at least read region, and
4969 avoiding excessive large RPC for small reads.
4971 Severity : enhancement
4973 Description: Creating more than 1000 files for a single job may cause a load
4974 imbalance on the OSTs if there are also a large number of OSTs.
4975 Details : qos_prep_create() uses an OST index reseed value that is an
4976 even multiple of the number of available OSTs so that if the
4977 reseed happens in the middle of the object allocation it will
4978 still utilize the OSTs as uniformly as possible.
4983 Description: OST or MDS may oops in ping_evictor_main()
4984 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4985 but doesn't restart at beginning of obd_exports_timed list
4988 Severity : enhancement
4990 Description: improve by-nid export eviction on the MDS and OST
4991 Details : allow multiple exports with the same NID to be evicted at one
4992 time without re-searching the exports list.
4995 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4997 Description: MDS may oops in groups_free()
4998 Details : in rare race conditions a newly allocated group_info struct is
4999 freed again, and this can be NULL. The 2.4 compatibility code
5000 for groups_free() checked for a NULL pointer, but 2.6 did not.
5003 Frequency : common for liblustre clients doing little filesystem IO
5004 Bugzilla : 9352, 7313
5005 Description: server may evict liblustre clients accessing contended locks
5006 Details : if a client is granted a lock or receives a completion AST
5007 with a blocking AST already set it would not reply to the AST
5008 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
5011 Frequency : lfs setstripe, only systems with more than 160 OSTs
5013 Description: unable to set striping with a starting offset beyond OST 160
5014 Details : llapi_create_file() incorrectly limited the starting stripe
5015 index to the maximum single-file stripe count.
5018 Frequency : LDAP users only
5020 Description: lconf did not handle in-kernel recovery with LDAP properly
5021 Details : lconf/LustreDB get_refs() is searching the wrong namespace
5023 Severity : enhancement
5025 Description: bind OST threads to NUMA nodes to improve performance
5026 Details : all OST threads are uniformly bound to CPUs on a single NUMA
5027 node and do their allocations there to localize memory access
5029 Severity : enhancement
5031 Description: llmount can determine client NID directly from Myrinet (GM)
5032 Details : the client NID code from gmnalnid was moved directly into
5033 llmount, removing the need to use this or specifying the
5034 client NID explicitly when mounting GM clients with zeroconf
5037 Frequency : if client is started with down MDS
5039 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5040 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5041 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5042 ptlrpc_import_delayed_req() because we didn't check if the
5043 request was interrupted, and we also didn't break out of the
5044 event loop if there was no timeout
5049 Description: data loss during non-page-aligned writes to a single file from
5050 both multiple nodes and multiple threads on one node at same time
5051 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5052 inconsistency led to false short-reads, that were cached and later
5053 used by ->prepare_write() to fill in partially written page,
5054 leading to data loss.
5057 Frequency : always, if lconf --abort_recovery used
5059 Description: lconf --abort_recovery fails with 'Operation not supported'
5060 Details : lconf was attempting to abort recovery on the MDT device and not
5063 Severity : enhancement
5065 Description: remove cleanup logs
5066 Details : replace lconf-generated cleanup logs with lustre internal
5067 cleanup routines. Eliminates the need for client-cleanup and
5070 Severity : enhancement
5072 Description: add support for EAs (user and system) on lustre filesystems
5073 Details : it is now possible to store extended attributes in the Lustre
5074 client filesystem, and with the user_xattr mount option it
5075 is possible to allow users to store EAs on their files also
5077 Severity : enhancement
5079 Description: Add possibility (config option) to show minimal available OST free
5081 Details : When compiled with --enable-mindf configure option, statfs(2)
5082 (and so, df) will return least minimal free space available from
5083 all OSTs as amount of free space on FS, instead of summary of
5084 free spaces of all OSTs.
5086 Severity : enhancement
5088 Description: do not expand extent locks acquired on OST-side
5089 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5090 by server: they are not cached anyway.
5093 Frequency : when mmap is used/binaries executed from Lustre
5095 Description: Unmmap pages before throwing them away from read cache.
5096 Details : llap_shrink cache now attempts to unmap pages before discarding
5097 them (if unmapping failed - do not discard). SLES9 kernel has
5098 extra checks that trigger if this unmapping is not done first.
5103 Description: lconf didn't resolve symlinks before checking to see whether a
5104 given mountpoint was already in use
5107 Frequency : when migrating failover services
5108 Bugzilla : 6395, 9514
5109 Description: When migrating a subset of services from a node (e.g. failback
5110 from a failover service node) the remaining services would
5111 time out and evict clients.
5112 Details : lconf --force (implied by --failover) sets the global obd_timeout
5113 to 5 seconds in order to quickly disconnect, but this caused
5114 other RPCs to time out too quickly. Do not change the global
5115 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5117 Severity : enhancement
5118 Frequency : if MDS is started with down OST
5119 Bugzilla : 9439,5706
5120 Description: Allow startup/shutdown of an MDS without depending on the
5121 availability of the OSTs.
5122 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5123 Add appropriate locking and lov-osc refcounts for safe
5124 cleaning. Add osc abort_inflight calls in case the
5125 synchronize never started.
5128 Frequency : occasional (Cray XT3 only)
5130 Description: root not authorized to access files in CRAY_PORTALS environment
5131 Details : The client process capabilities were not honoured on the MDS in
5132 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5133 been accessed by an authorized user then root was able to access
5134 the file on the local client also. The root user capabilities
5135 are now allowed on the MDS, as this environment has secure UID.
5138 Frequency : occasional
5140 Description: ldiskfs "too long searching" message happens too often
5141 Details : A debugging message (otherwise harmless) prints too often on
5142 the OST console. This has been reduced to only happen when
5143 there are fragmentation problems on the filesystem.
5148 Description: Division by zero in statfs when all OSCs are inactive
5149 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5150 lov_statfs divides by value returned by lov_get_stripecnt().
5154 Bugzilla : 9489, 3273
5155 Description: First write from each client to each OST was only 4kB in size,
5156 to initialize client writeback cache, which caused sub-optimal
5157 RPCs and poor layout on disk for the first written file.
5158 Details : Clients now request an initial cache grant at (re)connect time
5159 and so that they can start streaming writes to the cache right
5160 away and always do full-sized RPCs if there is enough data.
5161 If the OST is rebooted the client also re-establishes its grant
5162 so that client cached writes will be honoured under the grant.
5167 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5168 Details : Now I/O RPCs go to different portal number and (presumably) fast
5169 lock requests (and glimses) and other RPCs get their own service
5170 threads pool that should be able to service those RPCs
5173 Severity : enhancement
5175 Description: Ability to exchange lustre version between client and servers and
5176 issue warnings at client side if client is too old. Also for
5177 liblustre clients there is ability to refuse connection of too old
5179 Details : New 'version' field is added to connect data structure that is
5180 filled with version info. That info is later checked by server and
5184 Frequency : rare, liblustre only.
5185 Bugzilla : 9296, 9581
5186 Description: Two simultaneous writes from liblustre at offset within same page
5187 might proceed at the same time overwriting eachother with stale
5189 Details : I/O lock withing llu_file_prwv was released too early, before data
5190 actually was hitting the wire. Extended lock-holding time until
5191 server acknowledges receiving data.
5194 Frequency : extremely rare. Never observed in practice.
5196 Description: avoid generating lustre_handle cookie of 0.
5197 Details : class_handle_hash() generates handle cookies by incrementing
5198 global counter, and can hit 0 occasionaly (this is unlikely, but
5199 not impossible, because initial value of cookie counter is
5200 selected randonly). Value of 0 is used as a sentinel meaning
5201 "unassigned handle" --- avoid it. Also coalesce two critical
5202 sections in this function into one.
5204 Severity : enhancement
5206 Description: allow liblustre clients to delegate truncate locking to OST
5207 Details : To avoid overhead of locking, liblustre client instructs OST to
5208 take extent lock in ost_punch() on client's behalf. New connection
5209 flag is added to handle backward compatibility.
5211 Severity : enhancement
5212 Bugzilla : 4928, 7341, 9758
5213 Description: allow number of OST service threads to be specified
5214 Details : a module parameter allows the number of OST service threads
5215 to be specified via "options ost ost_num_threads={N}" in the
5216 OSS's /etc/modules.conf or /etc/modprobe.conf.
5220 Bugzilla : 6146, 9635, 9895
5221 Description: servers crash with bad pointer in target_handle_connect()
5222 Details : In rare cases when a client is reconnecting it was possible that
5223 the connection request was the last reference for that export.
5224 We would temporarily drop the export reference and get a new
5225 one, but this may have been the last reference and the export
5226 was just destroyed. Get new reference before dropping old one.
5228 Severity : enhancement
5229 Frequency : if client is started with failover MDS
5231 Description: Allow multiple MDS hostnames in the mount command
5232 Details : Try to read the configuration from all specified MDS
5233 hostnames during a client mount in case the "primary"
5236 Severity : enhancement
5238 Description: Stop sending data to evicted clients as soon as possible.
5239 Details : Check if the client we are about to send or are sending data to
5240 was evicted already. (Check is done every second of waiting,
5241 for which l_wait_event interface was extended to allow checking
5242 of exit condition at specified intervals).
5245 Frequency : rare, normally only when NFS exporting is done from client
5247 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5249 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5252 Frequency : always, if srand() is called before liblustre initialization
5254 Description: Liblustre uses system PRNG disturbing its usage by user application
5255 Details : Introduce internal to lustre fast and high-quality PRNG for
5256 lustre usage and make liblustre and some other places in generic
5257 lustre code to use it.
5259 Severity : enhancement
5260 Bugzilla : 9477, 9557, 9870
5261 Description: Verify that the MDS configuration logs are updated when xml is
5262 Details : Check if the .xml configuration logs are newer than the config
5263 logs stored on the MDS and report an error if this is the case.
5264 Request --write-conf, or allow starting with --old_conf.
5266 Severity : enhancement
5268 Description: Handle symlinks in the path when checking if Lustre is mounted.
5269 Details : Resolve intermediate symlinks when checking if a client has
5270 mounted a filesystem to avoid duplicate client mounts.
5275 Description: lconf can hit an error exception but still return success.
5276 Details : The lconf command catches the Command error exception at the top
5277 level script context and will exit with the associated exit
5278 status, but doesn't ensure that this exit status is non-zero.
5283 Description: failure of ptlrpc thread startup can cause oops
5284 Details : Starting a ptlrpc service thread can fail if there are a large
5285 number of threads or the server memory is very fragmented.
5286 Handle this without oopsing.
5289 Frequency : always, only if liblustre and non-default acceptor port was used
5291 Description: liblustre cannot connect to servers with non-default acceptor port
5292 Details : tcpnal_set_default_params() was not called and was therefore
5293 ignoring the environment varaible TCPNAL_PORT, as well as other
5294 TCPNAL_ environment variables
5299 Description: two objects could be created on the same OST for a single file
5300 Details : If an OST is down, in some cases it was possible to create two
5301 objects on a single OST for a single file. No problems other
5302 than potential performance impact and spurious error messages.
5306 Bugzilla : 5681, 9562
5307 Description: Client may oops in ll_unhash_aliases
5308 Details : Client dcache may become inconsistent in race condition.
5309 In some cases "getcwd" can fail if the current directory is
5315 Description: Inode refcounting problems in NFS export code
5316 Details : link_raw functions used to call d_instantiate without obtaining
5317 extra inode reference first.
5321 Bugzilla : 9942, 9903
5322 Description: Referencing freed requests leading to crash, memleaks with NFS.
5323 Details : We used to require that call to ll_revalidate_it was always
5324 followed by ll_lookup_it. Also with revalidate_special() it is
5325 possible to call ll_revalidate_it() twice for the same dentry
5326 even if first occurence returned success. This fix changes semantic
5327 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5328 reference on a request referred from the intent.
5329 ll_intent_release() then releases such a request.
5332 Frequency : rare, normally benchmark loads only
5334 Description: unlinked inodes were kept in memory on the client
5335 Details : If a client is repeatedly creating and unlinking files it
5336 can accumulate a lot of stale inodes in the inode slab cache.
5337 If there is no other client load running this can cause the
5338 client node to run out of memory. Instead flush old inodes
5339 from client cache that have the same inode number as a new inode.
5342 Frequency : SLES9 2.6.5 kernel and long filenames only
5343 Bugzilla : 9969, 10379
5344 Description: utime reports stale NFS file handle
5345 Details : SLES9 uses out-of-dentry names in some cases, which confused
5346 the lustre dentry revalidation. Change it to always use the
5350 Frequency : rare, unless heavy write-truncate concurrency is continuous
5351 Bugzilla : 4180, 6984, 7171, 9963, 9331
5352 Description: OST becomes very slow and/or deadlocked during object unlink
5353 Details : filter_destroy() was holding onto the parent directory lock
5354 while truncating+unlinking objects. For very large objects this
5355 may block other threads for a long time and slow overall OST
5356 responsiveness. It may also be possible to get a lock ordering
5357 deadlock in this case, or run out of journal credits because of
5358 the combined truncate+unlink. Solution is to do object truncate
5359 first in one transaction without parent lock, and then do the
5360 final unlink in a new transaction with the parent lock. This
5361 reduces the lock hold time dramatically.
5364 Frequency : rare, 2.4 kernels only
5366 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5367 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5368 before unmounting their filesystem in order to avoid deadlock.
5369 The kernel_locked() function in 2.4 kernels only checks whether
5370 the kernel lock is held, not whether it is this process that is
5371 holding it as 2.6 kernels do.
5376 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5377 Details : The client ptlrpc code may be trying to reconnect to a down
5378 server before a previous connection attempt has timed out.
5379 Increase the reconnect interval to be longer than the connection
5380 timeout interval to avoid sending duplicate connections to
5384 Frequency : echo_client brw_test command
5386 Description: fix echo_client to work with OST preallocated code
5387 Details : OST preallocation code (5137) didn't take echo_client IO path
5388 into account: echo_client calls filter methods outside of any
5389 OST thread and, hence, there is no per-thread preallocated
5390 pages and buffers to use. Solution: hijack pga pages for IO. As
5391 a byproduct, this avoids unnecessary data copying.
5395 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5396 Description: Client can oops in mdc_commit_close() after open replay
5397 Details : It was possible for the MDS to return an open request with no
5398 transaction number in mds_finish_transno() if the client was
5399 evicted, but without actually returning an error. Clients
5400 would later try to replay that open and may trip an assertion
5401 Simplify the client close codepath, and always return an error
5402 from the MDS in case the open is not successful.
5405 Frequency : rare, 2.6 OSTs only
5407 Description: OST may deadlock under high load on fragmented files
5408 Details : If there was a heavy load and highly-fragmented OST filesystems
5409 it was possible to have all the OST threads deadlock waiting on
5410 allocation of biovecs, because the biovecs were not released
5411 until the entire RPC IO was completed. Instead, release biovecs
5412 as soon as they are complete to ensure forward IO progress.
5414 Severity : enhancement
5416 Description: Support for specifying external journal device at mount
5417 Details : If an OST or MDS device is formatted with an external journal
5418 device, this device major/minor is stored in the ext3 superblock
5419 and may not be valid for failover. Allow detecting and
5420 specifying the external journal at mount time.
5425 Description: Mounting an MDS with pending unlinked files may cause oops
5426 Details : target_finish_recovery() calls mds_postrecov() which returned
5427 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5428 considers this an error and immediately begins cleaning up the
5429 lov, just after starting the mds_lov process
5431 Severity : enhancement
5433 Description: Implement 'lfs df' to report actual free space on per-OST basis
5434 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5435 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5436 usage of objects; '-h' to report in human readable format.
5438 ------------------------------------------------------------------------------
5440 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5447 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5448 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5449 at startup during orphan unlinking. Instead, skip buddy update
5450 before setup, as it will be regenerated after recovery is complete.
5455 Description: If an OST is inactive, its locks might reference stale inodes.
5456 Details : lov_change_cbdata() must iterate over all namespaces, even if
5457 they are inactive to clear inode references from the lock.
5459 Severity : enhancement
5460 Frequency : occasional, if non-standard max_dirty_mb used
5462 Description: Client will block write RPCs if not enough grant
5463 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5464 then the client will block writes while waiting for another RPC
5465 to complete instead of consuming its dirty limit. With change
5466 we get improved performance when max_dirty_mb is small.
5468 Severity : enhancement
5469 Bugzilla : 3389, 6253
5470 Description: Add support for supplementary groups on the MDS.
5471 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5472 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5473 lookups for user supplementary groups into a cache.
5477 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5478 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5479 for liblustre O_EXCL is enforced by MDS.
5483 Description: readdir never returns NULL in liblustre.
5484 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5485 dentry in struct dirent.
5489 Frequency : liblustre only, depends on application IO pattern
5490 Description: liblustre clients evicted if not contacting servers
5491 Details : Don't put liblustre clients into the ping_evictor list, so
5492 they will not be evicted by the pinger ever.
5494 Severity : enhancement
5496 Description: Add ability to evict clients by NID from MDS.
5497 Details : By echoing "nid:$NID" string into
5498 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5499 $NID would be instantly evicted from this MDS and from all active
5500 OSTs connected to it.
5504 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5505 Details : lookup_it_finish() used to query file size from OSTs that was not
5510 Description: service threads change working directory to that of init
5511 Details : Starting lustre service threads may pin the working directory
5512 of the parent thread, making that filesystem busy. Threads
5513 now change to the working directory of init to avoid this.
5517 Frequency : during shutdown only
5518 Description: shutdown with a failed MDS or OST can cause unmount to hang
5519 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5524 Frequency : 2.6 only
5525 Description: chmod/chown may include an extra supplementary group
5526 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5527 supplementary group and if none is specified this is used.
5530 Bugzilla : 5479 (6816)
5532 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5533 Details : If lookup is in progress on a file that is unlinked we might try
5534 to revalidate the inode and fail in revalidate after lookup is
5535 complete and ll_file_open() enqueues the open again but
5536 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5539 Frequency : always, if lconf --abort_recovery used
5541 Description: lconf --abort_recovery fails with 'Operation not supported'
5542 Details : lconf was attempting to abort recovery on the MDT device and not
5545 ------------------------------------------------------------------------------
5547 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5552 Frequency : rare (only unsupported configurations with a node running as an
5554 Bugzilla : 6514, 5137
5555 Description: Mounting a Lustre file system on a node running as an OST could
5557 Details : OSTs now preallocates memory needed to write out data at
5558 startup, instead of when needed, to avoid having to
5559 allocate memory in possibly low memory situations.
5560 Specifically, if the file system is mounted on on OST,
5561 memory pressure could force it to try to write out data,
5562 which it needed to allocate memory to do. Due to the low
5563 memory, it would be unable to do so and the node would
5564 become unresponsive.
5566 Severity : enhancement
5568 Description: Addition of lconf --service command line option
5569 Details : lconf now accepts a '--service <arg>' option, which is
5570 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5572 Severity : enhancement
5574 Description: Failover mode is now the default for OSTs.
5575 Details : By default, OSTs will now run in failover mode. To return to
5576 the old behaviour, add '--failout' to the lmc line for OSTs.
5578 Severity : enhancement
5580 Description: Health checks are now provided for MDS and OSTs
5581 Details : Additional detailed health check information on MSD and OSTs
5582 is now provided through the procfs health_check value.
5585 Frequency : occasional, depends on IO load
5587 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5588 after numerous create/delete cycles
5589 Details : The ext3 inode allocation policy would not allocate new inodes
5590 very well on the OSTs because there are no new directories
5591 being created. Instead we look for groups with free space if
5592 the parent directories are nearly full.
5597 Description: Network or server problems during mount may cause partially
5598 mounted clients instead of returning an error.
5599 Details : The config llog parsing code may overwrite the error return
5600 code during mount error handling, returning success instead
5606 Description: MDS can fail to allocate large reply buffers
5607 Details : After long uptimes the MDS can fail to allocate large reply
5608 buffers (e.g. zconf client mount config records) due to memory
5609 fragmentation or consumption by the buffer cache. Preallocate
5610 some large reply buffers so that these replies can be sent even
5611 under memory pressure.
5615 Frequency : rare (liblustre)
5616 Description: fsx running with liblustre complained that using truncate() to
5617 extend the file doesn't work. This patch corrects that issue.
5618 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5619 ATTR_SIZE and lsm use in llu_setattr_raw.
5623 Frequency : rare, only 2.6 kernels
5624 Description: Unusual file access patterns on the MDS may result in inode
5625 data being lost in very rare circumstances.
5626 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5627 code in the 2.6 kernel caused the inode and/or EA data not to
5628 be read from disk, causing single-file corruption.
5632 Frequency : rare, only 2.6 filesystems using extents
5633 Description: Heavy concurrent write and delete load may cause data corruption.
5634 Details : It was possible under high-load situations to have an extent
5635 metadata block in the block device cache from a just-unlinked
5636 file overwrite a newly-allocated data block. We now unmap any
5637 metadata buffers that alias just-allocated data blocks.
5641 Frequency : filesystems with default stripe_count larger than 77
5642 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5643 Details : lconf specifies an inode size of 4096 bytes when the default
5644 stripe_count is larger than 77. This conflicts with the default
5645 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5646 this case to avoid pinning too much memory for large EAs.
5648 ------------------------------------------------------------------------------
5650 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5655 Frequency : rare (extremely heavy IO load with hundreds of clients)
5657 Description: Client is evicted, gets IO error writing to file
5658 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5659 caused clients to be evicted for AST timeouts. The fixes in
5660 bug 5192 mean we no longer need to have such short AST timeouts
5661 so ldlm_timeout has been increased.
5664 Frequency : occasional during --force or --failover shutdown under load
5665 Bugzilla : 5949, 4834
5666 Description: Server oops/LBUG if stopped with --force or --failover under load
5667 Details : a collection of import/export refcount and cleanup ordering
5668 issues fixed for safer force cleanup
5671 Frequency : only filesystems larger than 120 OSTs
5672 Bugzilla : 5990, 6223
5673 Description: lfs getstripe would oops on a very large filesystem
5674 Details : lov_getconfig used kfree on vmalloc'd memory
5677 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5678 Bugzilla : 6242, 6243
5679 Description: reading from files that had been truncated to a non-zero size
5680 but never opened returned no data
5681 Details : ll_file_read() reads zeros from no-object files to EOF
5686 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5687 Details : There's little harm in aborting MDS/OSS recovery and letting it
5688 try again, so I removed the LASSERT and return an error instead.
5690 Severity : enhancement
5692 Description: New debugging infrastructure for tracking down data corruption
5693 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5694 (b) cover more of the client-side code path, and (c) try to narrow
5695 down where problems occurred
5699 Bugzilla : 3819, 4364, 4397, 6313
5700 Description: Racing close and eviction MDS could cause assertion in mds_close
5701 Details : It was possible to get multiple mfd references during close and
5702 client eviction, leading to one thread referencing a freed mfd.
5704 Severity: : enhancement
5705 Bugzilla : 3262, 6359
5706 Description: Attempts to reconnect to servers are now more aggressive.
5707 Details : This builds on the enhanced upcall-less recovery that was added
5708 in 1.4.2. When trying to reconnect to servers, clients will
5709 now try each server in the failover group every 10 seconds. By
5710 default, clients would previously try one server every 25 seconds.
5715 Description: After recovery, certain operations trigger a failed
5716 assertion on a client.
5717 Details : Failing over an mds, using lconf -d --failover, while a
5718 client was doing a readdir() call would cause the client to
5719 LBUG after recovery completed and the readdir() was resent.
5721 Severity : enhancement
5723 Description: Default groups are now added by lconf
5724 Details : You can now run lconf --group <servicename> without having to
5725 manually add groups with lmc.
5728 Frequency : occasional
5730 Description: Nodes with an elan id of 0 trigger a failed assertion
5733 Frequency : always when accessing e.g. tty/console device nodes
5735 Description: tty and some other devices nodes cannot be used on lustre
5736 Details : file's private_data field is used by device data and lustre
5737 values in there got lost. New field was added to struct file to
5738 store fs-specific private data.
5741 Frequency : when exporting Lustre via NFS
5743 Description: NFSD failed occasionally when looking up a path component
5744 Details : NFSD is looking up ".." which was broken in ext3 directories
5745 that had grown large enough to become hashed.
5748 Frequency : Clusters with multiple interfaces not on the same subnet
5750 Description: Nodes will repeatedly try to reconnect to an interface which it
5751 cannot reach and report an error to the log.
5752 Details : Extra peer list entries will be created by lconf with some peers
5753 unreachable. lconf now validates the peer before adding it.
5756 Frequency : Only if a default stripe is set on the filesystem root.
5758 Description: Setting a default stripe on the filesystem root prevented the
5759 filesystem from being remounted.
5760 Details : The client was sending extra request flags in the root getattr
5761 request and did not allocate a reply buffer for the dir EA.
5764 Frequency : occasional, higher if lots of files are accessed by one client
5765 Bugzilla : 6159, 6097
5766 Description: Client trips assertion regarding lsm mismatch/magic
5767 Details : While revalidating inodes the VFS looks up inodes with ifind()
5768 and in rare cases can find an inode that is being freed.
5769 The ll_test_inode() code will free the lsm during ifind()
5770 when it finds an existing inode and then the VFS later attaches
5771 this free lsm to a new inode.
5775 Bugzilla : 6422, 7030
5776 Description: MDS deadlock between mkdir and client eviction
5777 Details : Creating a new file via mkdir or mknod (starting a transaction
5778 and getting the ns lock) can deadlock with client eviction
5779 (gets ns lock and trying to finish a synchronous transaction).
5782 Frequency : occasional
5783 Description: While starting a server, the fsfilt_ext3 module could not be
5785 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5786 kernels. Previously, lconf would still use the ext3 name
5787 when trying to load modules. Now, it will correctly use
5788 ext3 on 2.4 and ldiskfs on 2.6.
5790 Severity : enhancement
5791 Description: The default stripe count has been changed to 1
5792 Details : The interpretation of the default stripe count (0, to lfs
5793 or lmc) has been changed to mean striping across a single
5794 OST, rather than all available. For general usage we have
5795 found a stripe count of 1 or 2 works best.
5797 Severity : enhancement
5798 Description: Add support for compiling against Cray portals.
5799 Details : Conditional compiling for some areas that are different
5803 Frequency : occasional
5804 Bugzilla : 6409, 6834
5805 Description: Creating files with an explicit stripe count may lead to
5806 a failed assertion on the MDS
5807 Details : If some OSTs are full or unavailable, creating files may
5808 trigger a failed assertion on the MDS. Now, Lustre will
5809 try to use other servers or return an error to the
5813 Frequency : occasional
5815 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5816 caused lock timeout during readahead (since 1.4.2).
5817 Details : Processes doing readahead might match a lock that hasn't been
5818 granted yet if there are overlapping and conflicting lock
5819 requests. The readahead process waits on ungranted lock
5820 (original lock is CBPENDING), while OST waits for that process
5821 to cancel CBPENDING read lock and eventually evicts client.
5823 Severity : enhancement
5825 Description: Initial enabling of flock support for clients
5826 Details : Implements fcntl advisory locking and file status functions.
5827 This feature is provided as an optional mount flag (default
5828 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5829 locking are implemented yet, and those that are are not guaranteed
5830 to be completely correct in production environments.
5831 mount -t lustre -o [flock|noflock] ...
5834 Frequency : occasional
5836 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5837 assertion in the ext3 JBD (journaling) layer.
5838 Details : The b_committed_data struct is protected by the big kernel lock
5839 in 2.4 kernels, serializing journal_commit_transaction() and
5840 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5841 access to this struct. In 2.6 kernels there is finer grained
5842 locking to improve SMP performance of the JBD layer.
5846 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5851 Description: merge and cleanup kernel patches.
5854 -----------------------------------------------------------------------------
5856 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5858 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5859 versions of Lustre. Please update all servers and clients to
5860 version 1.4.2 or later at the same time. You must also run
5861 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5862 to update the configuration logs.
5864 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5865 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5866 - swab llog records of type '0' so we get proper header size/idx (5861)
5867 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5868 - fix rename of one directory over another leaking an inode (5953)
5869 - avoid SetPageDirty on 2.6 (5981)
5870 - don't re-add just-being-destroyed locks to the waiting list (5653)
5871 - when creating new directories, inherit the parent's custom
5872 striping settings if present parent (3048)
5873 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5874 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5875 - don't hold client locks on temporary worklist from l_lru (5666)
5876 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5877 - automatically evict dead clients (5921)
5878 - Update file size properly in create+truncate+fstat case (6196)
5879 - Do not unhash mountpoint dentries, do not allow removal of
5881 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5882 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5883 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5884 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5885 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5886 - Correctly handle reads of files with no objects (6243)
5887 - lctl recover will also mark a device active if deactivate used (5933)
5889 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5890 - allow --write-conf on an MDS with different nettype than client (5619)
5891 - don't write config llogs to MDS for mounts not from that MDS (5617)
5892 - lconf should create multiple TCP connections from a client (5201)
5893 - init scripts are now turned off by default; run chkconfig --on
5894 lustre and chkconfig --on lustrefs to use them
5895 - upcalls are no longer needed for clients to recover to failover
5897 - add --abort-recovery option to lconf to abort recovery on device
5899 - add support for an arbitrary number of OSTs (3026)
5900 - Quota support protocol changes.
5901 - forward compatibility changes to wire structs (6007)
5902 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5903 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5904 - improved "lustre status" script
5905 - initialize blocksize for non-regular files (6062)
5906 - added --disable-server and --disable-client configure options (5782)
5907 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5908 - Vanilla 2.4.29 support
5909 - increase maximum number of obd devices to 520 (6242)
5910 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5911 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5912 SLES 9 2.6.5 kernels
5913 - limit stripes per file to 160 (the maximum EA size) (6093)
5915 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5918 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5919 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5920 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5921 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5922 - small ext3 extents cleanups and fixes (5733)
5923 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5924 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5925 - avoid needless client->OST connect, fix handle mismatch (5317)
5926 - fix DLM error path that led to out-of-sync client, long delays (5779)
5927 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5928 - fix several locking issues related to i_size (5492,5624,5654,5672)
5929 - don't move pending lock onto export if it is already evicted (5683)
5930 - fix kernel oops when creating .foo in unlinked directory (5548)
5931 - fix deadlock in obdfilter statistics vs. object create (5811)
5932 - use time_{before,after} to avoid timer jiffies wrap (5882)
5933 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5934 - Do not leak request if server was not able to process it (5154)
5935 - If mds_open unable to find parent dir, make that negative lookup(5154)
5936 - don't create new directories with extent-mapping (5909, 5936)
5938 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5939 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5940 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5941 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5942 - support for RHEL4 kernel on i686 (5773)
5943 - provide error messages when incompatible logs are encountered (5898)
5945 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5946 * version 1.4.0.10 (1.4.1 release candidate 1)
5948 - don't keep a lock reference when lock is not granted (4238)
5949 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5950 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5951 - drop import inflight refcount on signal_completed_replay error (5255)
5952 - unlock page after async write error during send (3677)
5953 - handle missing objects in filter_preprw_read properly (5265)
5954 - no transno return for symlink open, don't save no-trasno open (3440)
5955 - don't try to complete elan receive that already failed (4012)
5956 - free RPC server reply state on error (5406)
5957 - clean up thread from ptlrpc_start_thread() on error (5160)
5958 - readahead could read extra page into cache that wasn't ejected (5388)
5959 - prevent races in class_attach/setup/cleanup/detach (5260)
5960 - don't dereference de->d_inode after l_dput of de (5458)
5961 - use "int" for stripe value returned from lock_to_stripe (5544)
5962 - mballoc allocation and error-checking fixes in 2.6 (5504)
5963 - block device patches to fix I/O request sizes in 2.6 (5482)
5964 - look up hostnames for IB nals (5602)
5965 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5966 - don't start multiple acceptors for the same port (5277)
5967 - fix incorrect LASSERT in mds_getattr_name (5635)
5968 - export a proc file for general "ping" checking (5628)
5969 - fix "lfs check" to not block when the MDS is down (5628)
5971 - service request history (4965)
5972 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5973 - create an "evict_client" /proc entry on OSTs, like the MDS has
5974 - fix mount usage message, return errors per mount(8) (5168)
5975 - change grep [] to grep "[]" in tests so they work in more UMLs
5976 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5977 - remove ancient LOV_MAGIC_V0 EA support (5047)
5978 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5979 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5980 - allow readahead allocations to fail when low on memory (5383)
5981 - mmap locking landed again, after considerable improvement (2828)
5982 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5984 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5987 - send OST transaction number in read/write reply to free req (4966)
5988 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5989 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5990 - the watchdog thread now runs as interruptible (5246)
5991 - flock/lockf fixes (but it's still disabled, pending 5135)
5992 - don't use EXT3 constants in llite code (5094)
5993 - memory shortage at startup could cause assertion (5176)
5995 - reorganization of lov code
5996 - single portals codebase
5998 - add extents/mballoc support (5025)
5999 - direct I/O reads in the obdfilter (4048)
6000 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
6002 tbd Cluster File Systems, Inc. <info@clusterfs.com>
6005 - send OST transaction number in read/write reply to free req (4966)
6006 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
6007 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
6008 - the watchdog thread now runs as interruptible (5246)
6009 - handle missing objects in filter_preprw_read properly (5265)
6010 - unsafe list practices (rarely) led to infinite eviction loop (4908)
6011 - drop import inflight refcount on signal_completed_replay error (5255)
6012 - unlock page after async write error during send (3677)
6013 - return original error code on reconstructed replies (3761)
6014 - no transno return for symlink open, don't save no-trasno open (3440)
6016 - add pid to ldlm debugging output (4922)
6017 - bump the watchdog timeouts -- we can't handle 30sec yet
6018 - extra debugging for orphan dentry/inode bug (5259)
6020 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
6023 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
6024 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
6025 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
6026 - fix NULL dereference in /proc/sys/portals/routes (4827)
6027 - allow failed mdc_close() operations to be interrupted (4561)
6028 - stop precreate on OST before MDS would time out on it (4778)
6029 - don't send partial-page writes before EOF from client (4410)
6030 - discard client grant for sub-page writes on large-page clients (4520)
6031 - don't free dentries not owned by NFS code, check generation (4806)
6032 - fix lsm leak if mds_create_objects() fails (4801)
6033 - limit debug_daemon file size, always print CERROR messages (4789)
6034 - use transno after validating reply (3892)
6035 - process timed out requests if import state changes (3754)
6036 - update mtime on OST during writes, return in glimpse (4829)
6037 - add mkfsoptions to LDAP (4679)
6038 - use ->max_readahead method instead of zapping global ra (5039)
6039 - don't interrupt __l_wait_event() during strace
6041 - add software watchdogs to catch hung threads quickly (4941)
6042 - make lustrefs init script start after nfs is mounted
6043 - fix CWARN/ERROR duplication (4930)
6044 - return async write errors to application if possible (2248)
6045 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6046 - print NAL number in %x format (4645)
6047 - update barely-supported suse-2.4.21-171 series (4842)
6048 - support for sles 9 %post scripts
6049 - support for building 2.6 kernel-source packages
6050 - support for sles km_* packages
6052 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6055 - ignore -ENOENT errors in osc_destroy (3639)
6056 - notify osc create thread that OSC is being cleaned up (4600)
6057 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6058 - reconstruct ost_handle() like mds_handle() (4657)
6059 - create a new thread to do import eviction to avoid deadlock (3969)
6060 - let lconf resolve symlinked-to devices (4629)
6061 - don't unlink "objects" from directory with default EA (4554)
6062 - hold socknal file ref over connect in case target is down (4394)
6063 - allow more than 32000 subdirectories in a single directory (3244)
6064 - fix blocks count for O_DIRECT writes (3751)
6065 - OST returns ENOSPC from object create when no space left (4539)
6066 - don't send truncate RPC if file size isn't changing (4410)
6067 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6068 - bind to privileged port in socknal and tcpnal (3689)
6070 - rate limit CERROR/CWARN console message to avoid overload (4519)
6071 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6072 - basic mmap support (3918)
6073 - kernel patch series update from b1_4 (4711)
6075 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6078 - avoid crash during MDS cleanup with OST shut down (2775)
6079 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6080 - don't use bad inodes on the MDS (3744)
6081 - dynamic object preallocation to improve recovery speed (4236)
6082 - don't hold spinlock over lock dumping or change debug flags (4401)
6083 - don't zero obd_dev when it is force cleaned (3651)
6084 - print grants to console if they go negative (4431)
6085 - "lctl deactivate" will stop automatic recovery attempts (3406)
6086 - look for existing locks in ldlm_handle_enqueue() (3764)
6087 - don't resolve lock handle twice in recovery avoiding race (4401)
6088 - revalidate should check working dir is a directory (4134)
6090 - don't always mark "slow" obdfilter messages as errors (4418)
6092 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6095 - don't close LustreDB during write_conf until it is done (3860)
6096 - fix typo in lconf for_each_profile (3821)
6097 - allow dumping logs from multiple threads at one time (3820)
6098 - don't allow multiple threads in OSC recovery (3812)
6099 - fix debug_size parameters (3864)
6100 - fix mds_postrecov to initialize import for llog ctxt (3121)
6101 - replace config semaphore with spinlock (3306)
6102 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6103 - don't allow enqueue to complete on a destroyed export (3822)
6104 - down write_lock before checking llog header bitmap (3825)
6105 - recover from lock replay timeout (3764)
6106 - up llog sem before sending rpc (3652)
6107 - reduce ns lock hold times when setting kms (3267)
6108 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6109 - fix NULL deref and obd_dev leak on setup error (3312)
6110 - replace some LBUG about llog ops with error handling (3841)
6111 - don't match INVALID dentries from d_lookup and spin (3784)
6112 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6113 - fix invalid assertion in ptlrpc_set_wait (3880)
6115 - add libwrap support for the TCP acceptor (3996)
6116 - add /proc/sys/portals/routes for non-root route listing (3994)
6117 - allow setting MDS UUID in .xml (2580)
6118 - print the stack of a process that LBUGs (4228)
6120 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6123 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6124 - make sure to unset replay flag from failed open requests (3440)
6125 - if default stripe count is 0, use OST count for inode size (3636)
6126 - update parent mtime/ctime on client for create/unlink (2611)
6127 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6128 - free recovery state on server during a forced cleanup (3571)
6129 - unregister_reply for resent reqs (3063)
6130 - loop back devices mounting and status check on 2.6 (3563)
6131 - fix resource-creation race that can provoke i_size == 0 (3513)
6132 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6133 - more debugging for page-accounting assertion (3746)
6134 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6135 - avoid deadlock after precreation failure (3758)
6136 - fix race and lock order deadlock in orphan handling (3450, 3750)
6137 - add validity checks when grabbing inodes from l_ast_data (3599)
6139 - add /proc/.../recovery_status to obdfilter (3428)
6140 - lightweight CDEBUG infrastructure, debug daemon (3668)
6141 - change default OSC RPC parameters to be better on small clusters
6142 - turn off OST read cache for files smaller than 32MB
6143 - install man pages and include them in rpms (3100)
6144 - add new init script for (un)mounting lustre filesystems (2593)
6145 - run chkconfig in %post for init scripts (3701)
6146 - drop scimac NAL (unmaintained)
6148 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6151 - clean kiobufs before and after use (3485)
6152 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6153 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6154 - change init script to fail more clearly if not run as root (1528)
6155 - allow clients to reconnect during replay (1742)
6156 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6157 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6158 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6159 - fix cond_resched (3554)
6160 - only evict unfinished clients after recovery (3515)
6161 - allow bulk resend, prevent data loss (3570)
6162 - dynamic ptlrpc request buffer allocation (2102)
6163 - don't allow unlinking open directory if it isn't empty (2904)
6164 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6165 - remove extraneous obd dereference causing LASSERT failure (3334)
6166 - don't use get_cycles() when creating temp. files on the mds (3156)
6167 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6168 - handle EEXIST for set-stripe, set proper directory name (3336)
6170 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6171 - fix ksocknal_fmb_callback() error messages (2918)
6173 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6176 - don't copy lvb into (possibly NULL) reply on error (2983)
6177 - don't deref dentry after dput, don't free lvb on error (2922)
6178 - use the kms to determine writeback rpc length (2947)
6179 - increment oti_logcookies when osc is inactive (2948)
6180 - update client's i_blocks count via lvb messages (2543)
6181 - handle intent open/close of special files properly (1557)
6182 - mount MDS with errors=remount-ro, like obdfilter (2009)
6183 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6184 - don't use cancelling-locks' kms values (2947)
6185 - use highest lock extent for kms, not last one (2925)
6186 - don't dereference ERR_PTR() dentry in error handling path (3107)
6187 - fix thread race in portals_debug_dumplog() (3122)
6188 - create lprocfs device entries at setup instead of at attach (1519)
6189 - common AST error handler, don't evict client on completion race (3145)
6190 - zero nameidata in detach_mnt in 2.6 (3118)
6191 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6192 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6193 - initialize RPC timeout timer earlier for 2.6 (3219)
6194 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6195 - print nal/nid for unknown nid (3258)
6196 - additional checks for oscc recovery before doing precreate (3284)
6197 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6198 - don't crash in mdc_close for bad permissions on open (3285)
6199 - zero i_rdev for non-device files (3147)
6200 - clear page->private before handing to FS, better assertion (3119)
6201 - tune the read pipeline (3236)
6202 - fix incorrect decref of invalidated dentry (2350)
6203 - provide read-ahead stats and refine rpc in flight stats (3328)
6204 - don't hold journal transaction open across create RPC (3313)
6205 - update atime on MDS at close time (3265)
6206 - close LDAP connection when recovering to avoid server load (3315)
6207 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6208 - don't leak open file on MDS after open resend (3325)
6209 - serialize filter_precreate and filter_destroy_precreated (3329)
6210 - loop device shouldn't call sync_dev() for nul device (3092)
6211 - clear page cache after eviction (2766)
6212 - resynchronize MDS->OST in background (2824)
6213 - refuse to mount the same filesystem twice on same mountpoint (3394)
6214 - allow llmount to create routes for mounting behind routers (3320)
6215 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6216 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6217 - fix rare problem with rename on htree directories (3417)
6219 - allow default OST striping configuration per directory (1414)
6220 - fix compilation for qswnal for 2.6 kernels (3125)
6221 - increase maximum number of MDS request buffers for large systems
6222 - change liblustreapi to be useful for external progs like lfsck (3098)
6223 - increase local configuration timeout for slow disks (3353)
6224 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6226 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6229 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6230 - don't overwrite extent policy data in reply if lock was blocked (2901)
6231 - drop filter export grants atomically with removal from device (2663)
6232 - del obd_self_export from work_list in class_disconnect_exports (2908)
6233 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6234 - swab reply message in mdc_close, other PPC fixes (2464)
6235 - fix destroying of named logs (2325)
6236 - overwrite old logs when running lconf --write_conf (2264)
6237 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6238 - fix race in target_handle_connect (2898)
6239 - mds_reint_create() should take same inode create lock (2926)
6240 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6241 - don't close files for self_export to avoid uninitialized obd (2936)
6242 - allow MDS with the same name as client node (2939)
6243 - hold dentry reference for closed log files for unlink (2325)
6244 - reserve space for all logs during transactions (2059)
6245 - don't evict page beyond end of stripe extent (2925)
6246 - don't oops on a deleted current working directory (2399)
6247 - handle hard links to targets without a parent properly (2517)
6248 - don't dereference NULL lock when racing during eviction (2867)
6249 - don't grow lock extents when lots of conflicting locks (2919)
6251 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6254 - account for cache space usage on clients to avoid data loss (974)
6255 - lfsck support in lustre kernel code (2349)
6256 - reduce journal credits needed for BRW writes (2370)
6257 - orphan handling to avoid losing space on client/server crashes
6258 - ptlrpcd can be blocked, stopping ALL progress (2477)
6259 - use lock value blocks to assist in proper KMS, faster stat (1021)
6260 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6261 - recovery for initial connections (2355)
6262 - fixes for mds_cleanup_orphans (1934)
6263 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6264 - block all file creations until orphan recovery completes (1901)
6265 - client remove rq_connection from request struct (2423)
6266 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6267 - recovery timer race (2670)
6268 - mdc_close recovey bug (2532)
6269 - ptlrpc cleanup bug (2710)
6270 - mds timeout on local locks (2588)
6271 - namespace lock held during RPCs (2431)
6272 - handle interrupted sync write properly (2503)
6273 - don't try to handle a message that hasn't been replied to (2699)
6274 - client assert failure during cleanup after abort recovery (2701)
6275 - leak mdc device after failed mount (2712)
6276 - ptlrpc_check_set allows timedout requests to complete (2714)
6277 - wait for inflight reqs when ptlrpcd finishes (2710)
6278 - make sure unregistered services are removed from the srv_list
6279 - reset bulk XID's when resending them (caught by 1138 test)
6280 - unregister_bulk after timeout
6281 - fix lconf error (2694)
6282 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6283 - readahead locks pages, leaves pending causing memory pressure (2673)
6284 - increase OST request buffers to 4096 on large machines (2729)
6285 - fix up permission of existing directories in simple_mkdir (2661)
6286 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6287 - don't assign transno to errored transactions (2742)
6288 - don't delete objects on OST if given a bogus objid from MDS (2751)
6289 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6290 - if rq_no_resend, then timeout request after recovery (2432)
6291 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6292 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6293 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6294 - MDS recovery shouldn't delete objects that it has given out (2730)
6295 - if enqueue arrives after completion, don't clobber LVB (2819)
6296 - don't unlock pages twice when trigger_group_io returns error (2814)
6297 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6298 - don't write pages to disk if there was an error (1450)
6299 - don't ping imports that have recovery disabled (2676)
6300 - take buffered bytes into account when balancing socknal conn (2817)
6301 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6302 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6303 - remove little-endian swabbing of llog records (1987)
6304 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6305 - retry reposting request buffers if they fail (1191)
6306 - grow extent at grant time to avoid granting a revoked lock (2809)
6307 - lock revoke doesn't evict page if covered by a second lock (2765)
6308 - disable VM readahead to avoid reading outside lock extents (2805)
6310 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6311 - updated kernel patches for hp-2.4.20 kernel (2681)
6313 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6316 - fix truncated write corruption (2366)
6317 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6319 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6320 - ensure len(uuid) < 37 in lmc (1171)
6321 - fix ia64 OOPS in llog_test (2255)
6322 - zero end of page at obdfilter for partial page writes (2648)
6323 - don't leave stale dentries around after renames (bug 2428)
6324 - fix timeouts when evicting a client with a single lock held (2642)
6325 - set deadline for the initial HELLO message to drain (2634)
6326 - print out dotted-quad IP addresses in the socknal (2302)
6328 - additional debugging for MDS client eviction problem (2443)
6329 - fix mkfsoptions support for osts (2603, 2604)
6331 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6334 - add series for the vanilla 2.6.0 kernel
6335 - add series for the vanilla 2.4.24 kernel
6336 - add series for a cray x86/64 UL kernel drop
6337 - fix xattr patches for the vanilla 2.4.19 series
6339 - generate true UUIDs in lmc (1171)
6340 - have portals stack dumping break in UML (2466)
6341 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6342 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6343 - wake callback waiting threads on client eviction (2460)
6344 - Add --ptldebug and --subsystem to lmc (1719)
6345 - update assertion to allow safe interrupt allocation
6346 - set rq_no_resend for cancel requests (2432)
6347 - recalculate ptlrpcd timeout after resend (2494)
6348 - call vfs_rmdir when removing pending directories (2368)
6349 - fix renaming a file to itself (2429)
6350 - lmc creates a default one-stripe lov (2454)
6351 - expand procfs space to handle large clusters (2326)
6352 - increase UML stack to avoid overflow
6353 - update lconf's list of debug and subsystem masks
6354 - fix lfs find --obd (2510)
6355 - /proc tunable for disabling filter read caching (2591)
6356 - stop rpm packages from altering slapd.conf (2301)
6357 - disable nagle in the socknal under 0conf (2578)
6358 - choose mds inode size based on stripe count (2572)
6359 - fix kernel-source rpm problems (2516)
6361 - add --disable-doc to avoid pdf generation (2421)
6362 - update documentation, tests, type-os, comments
6363 - avoid format warnings on ia64
6364 - remove the TOE NAL
6365 - tiny code cleanups by removing unused fields
6367 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6370 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6371 - remove the most blatant lies from BUILDING file (2371)
6372 - change default debug level to reasonable production setting
6373 - reduce client side cache size to reduce cache flush time
6374 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6375 - make TCP zerocopy and pinger support enabled by default (2476)
6376 - sync writes completed after process exits caused crashes (2319)
6377 - maintain correct mount count on the MDS (2356)
6378 - backout 1557, because 2316 wasn't really fixed
6379 - better file I/O statistics gathering in /proc
6380 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6381 - another kernel patch to fix zero-copy TCP function export
6382 - don't take duplicate lock when processing re-sent getattr (2420)
6383 - lctl uses obd_self_export instead of creating new conn (2353)
6384 - MDS/OST recovery case which requires object creation asserted (2425)
6385 - move lfs from /usr/sbin to /usr/bin in packages
6386 - fix race between mds_client_add and mds_client_free (2417)
6387 - use kmalloc instead of slabs in portals (2430)
6388 - don't create duplicate records when a failover MDS is present (2442)
6389 - remove unnecessary mount age check (2332)
6390 - don't remove directory inodes from locks prematurely (2451)
6391 - don't break if MDS service name is the same as hostname (2103)
6392 - fix races in client write RPC generation when cache full (2482)
6394 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6397 - remove now-unused request->rq_obd (278)
6398 - if an allocation fails, print out how much memory we've used (1933)
6399 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6400 - add the "configurable stack size" patch to most series files (1256)
6401 - ability to write large log records, for 100+ OST configs (2306)
6402 - fix NULL deref when filter_prep fails (2314)
6403 - fix operator precedence error in filter_sync
6404 - dynamic allocation of socknal TX descriptors (2315)
6405 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6406 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6407 - ensure that CWARN messages in Portals always get to the syslog
6408 - __init/__exit are not for prototype decls (ldlm_init/exit)
6409 - x86-64 compile warning fixes
6410 - fix gateway LMC keyword conflict (2318)
6411 - fix MDS lock inversions in getattr/reint paths (1844)
6412 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6413 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6414 - DLM race condition prevented some lock evictions (2328)
6415 - ENOMEM detection and retry on socknal sends (2230)
6416 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6417 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6418 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6419 - better error messages when a client is rejected during recovery (1505)
6420 - avoid cancelling locks which were never granted, after failure (2330)
6421 - fix i_sem/journal inversion in mds_client_add (2333)
6422 - fix truncate/getattr lock cycle deadlock (2334)
6423 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6424 - fix two rare exit paths which could leak an l_lock() ref (2321)
6425 - fencepost error in MDS/OST orphan recovery (2226)
6426 - make log record alignment 8 bytes (1988)
6427 - lstripe now fails when requested offset > ost_count (2237)
6428 - ensure that all kernel series have a complete list.h (1607)
6429 - fix crashes in special-file operations (2316)
6430 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6432 - allow configurable automake binary, for testing new versions
6433 - small update to the lfs documentation
6435 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6437 * fix negative export reference count in fsfilt_sync (2312)
6439 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6440 * release candidate 0.9.1
6442 - orphans are moved into the PENDING directory for possible recovery
6443 - replayed opens now open by fid for orphan/rename safety (1042)
6444 - last close of an orphan inode generates a transno (683)
6445 - chdir() and mount() now pin the directory entry (1020)
6446 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6447 - discard very old requests without processing them (1502)
6448 - remove some common, well-understood CERRORs (1505)
6449 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6450 - clear "grant" flags in OST replies until OST grant code lands (1644)
6451 - fix read performance by not clobbering i_blksize on client (1598)
6452 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6453 - make LRU size a /proc tunable, clears locks when reduced (707)
6454 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6455 - server locks take references on exports now (1558)
6456 - build fixes for 2.4.20-rh trees (1663)
6457 - return an error from lov_create if all OSCs are inactive (1751)
6458 - fix import levels when a reconnect happens without a timeout (1597)
6459 - exit early from mds_open if we get a lookup error (1749)
6460 - partial page read at EOF wouldn't wait for disk before sending (1642)
6461 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6462 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6463 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6464 - write/truncate lock inversion (1639)
6465 - Don't auto-load obdclass, portals modules during cleanup (1495)
6466 - fix timestamps from jumping to "now" (1763)
6467 - extra journal assertions (1648)
6468 - add an extra multiunlink test (1771)
6469 - fix read_record/write_record API (1776)
6470 - fix leak of offset_extent, possible incorrect i_size later (1772)
6471 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6472 - Debugging for the kqswnal_get_idle_tx problems (1820)
6473 - Allow recovery to be attempted multiple times (1536)
6474 - Write out MDS last_rcvd file after it is first created (1600)
6475 - Fix tx_descriptor leak in failed transmit situations (1827)
6476 - ext3 journaling fixes for assertion failure after IO error (1871)
6477 - class_export_put() on freed export after completion AST error (1896)
6478 - Fix revalidate looping in VFS (1322)
6479 - Don't access a freed export during MDS_REINT timeout (1521)
6480 - Add open-unlink recovery support on the MDS (1673,1764)
6481 - Return an error if no MDS data was read from last_rcvd (1946)
6482 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6483 - Don't setup a disk device that doesn't match exported UUID (317)
6484 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6485 - avoid committing NULL handle in force close
6486 - local.sh is now a one-stripe LOV configuration
6487 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6488 - fix ext3 htree duplicate directory entry corruption (1516)
6489 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6490 - update to new LOV EA format (2097)
6491 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6493 2003-06-15 Phil Schwan <phil@clusterfs.com>
6496 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6497 - per-import/export recovery handling (958, 931, 959)
6498 - multiple last-rcvd slots, for serving multiple FSes (949)
6499 - connections are again shared between multiple imp/exports (963, 964)
6500 - "umount -f" would hang if any requests needed to be sent (393, 978)
6501 - avoid pinning large req buffer by copying for queued messages (989)
6502 - add "uuid" to "lctl device" command to help upcalls (991)
6503 - "open" RPCs with transnos would confuse recovery counters (1037)
6504 - do proper endian conversion of all wire messages (288, 340, 891)
6505 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6506 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6507 - fix ptlrpc_connection leak in target_handle_connect (1174)
6508 - fix import refcounting bug in OST and MDS cleanup (1134)
6509 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6510 - fix very unlikely obd_types race condition (501)
6511 - remove osc_open hack for echo_client (1187)
6512 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6513 - a failure in read_inode2 leads to deadlock (1139)
6514 - cancel ack-locks as soon as transaction is committed (1072)
6515 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6516 - make sure to commitrw after any preprw to avoid deadlock (1162)
6517 - failing to execute a file in a lustre FS would lock inode (1203)
6518 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6519 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6520 - obd_uuid2tgt would walk off the end of the list (1255)
6521 - on IA64 the osc would give portals incorrect bulk size (1258)
6522 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6523 - fix lock inversion caused by new llite matching code (1282)
6524 - limit the number of dirty pages on a client to 10MB (1286)
6525 - timed out locks were not being corrected cancelled (1289)
6526 - fix O_DIRECT above 4GB on IA-32 (1292)
6527 * major user-visible changes
6528 - fail out/fail over policy now controlled by the upcall (993)
6530 - add OBD_PING to check server availability and failure (954)
6531 - lustre messages are now sent in sending host order (288, 340, 891)
6532 - add eadatalen to MDS getattr reply (340)
6533 - OST read replies may contain second buffer, with per-page status (593)
6535 2003-03-11 Phil Schwan <phil@clusterfs.com>
6538 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6539 - failed open()s could cause deadlock; fixed (867, 869)
6540 - stop cancelling OST locks when files are closed (481)
6541 - overlapping XID spaces caused network corruption (851, 853)
6542 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6543 - setattr_raw added, to do single-RPC, server-side setattrs
6544 - lmc/lconf syntax change for OST UUIDs
6545 - fix crashy race condition between ptlrpc_free_req and osc_close
6546 - don't use request in mdc_enqueue if we hit a timeout (889)
6547 - don't set the inode i_size for regular files from the MDS (896)
6548 - handle out of order completion AST (842)
6549 - don't LBUG if a lock request times out after receiving AST (913)
6550 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6551 - if a bad lock AST arrives, send an error instead of dropping entirely
6552 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6553 - fix leak in bulk IO when only partially completed (899, 900, 926)
6554 - fix O_DIRECT for ia64 (55)
6555 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6556 - C-z after timeout could hang a process forever; fixed (977)
6558 - client-side I/O cache (678, 924, 929, 941, 970)
6560 - READPAGE and SETATTRs which don't take server-side locks get
6563 2003-02-11 Phil Schwan <phil@clusterfs.com>
6566 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6567 - Fix fsfilt_extN_readpage() to read a full page of directory
6568 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6569 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6570 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6571 - Most intent operations were removed, in favour of a new RPC mode
6572 that does a single RPC to the server and bypasses most of the VFS
6573 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6574 - Aggressively cancel local locks on DLM servers
6575 - mds_reint_unlink sends EA to the client if it's the last nlink.
6576 client uses that EA to unlink OST objects.
6577 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6578 - recursive symlinks were fixed (439)
6579 - fixed NULL deref in DEBUG_REQ
6580 - filter_update_lastobjid no longer calls sync, which annoyed extN
6581 - fixed multi-client small-writes to a single file problem (445)
6582 - fixed mtime updates during file writes (607)
6583 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6584 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6585 - fixed Linux OST size reporting problem (444, 656)
6586 - OST now updates object mtime with writes or setattr (607, 619)
6587 - client verifies file size before zeroing page past EOF (445)
6588 - OST now writes last allocated objid to disk with allocation (108)
6589 - LOV on echo now works (409)
6591 - mds_reint_unlink sends a new buffer, with the EA included. this
6592 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6593 set if a regular file was being unlinked, and it was the last link
6594 - use PtlGet from the target for bulk writes (315)
6595 - OST now updates object mtime with writes or setattr (607, 619)
6596 - LDLM now has a grant-time callback to revalidate locked items, if
6598 - Many MDS operations were reorganized to combat race conditions
6600 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6601 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6603 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6606 - Fully reactivate OST imports after reconnection (512, others)
6607 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6608 - More graceful error handling for truncating on dead OST (515)
6609 - Don't error out unless we're actually accessing dead stripes (474)
6610 - Fix garbage sizes when stripes are missing (410)
6611 - LRU counters were broken, causing constant lock purge (433, 432)
6612 - garbage on read from stripes with failed OSTs (441)
6613 - mark OSCs as active before reconnecting during recovery (438)
6614 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6615 - lfind did not preserve OST order in output (443)
6616 - symlinks cause hung clients, incorrect data (439)
6617 - stop dereferencing request after dropping refcount (457)
6618 - don't LASSERT(spin_is_locked) on non-SMP (455)
6619 - fixes for many rename() bugs
6620 - fstat didn't correctly synchronize attributes (399)
6621 - server must handle lock cancellation during blocking AST prep (487)
6622 - bulk descriptors were free()d too soon (511)
6623 - fix paths in lconf, which would load incorrect modules (451, 507)
6624 - fix confusing lconf 'host not found' error message (386)
6625 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6626 - fix race condition in mdc_blocking_ast() for inode access (526)
6627 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6628 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6629 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6630 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6631 - fix lstripe in several ways (kernel side) (527)
6632 - fix request leak in ldlm_cli_enqueue (262)
6633 - incorrect OSC was marked inactive after OST failure
6634 - call mds_fs_cleanup before unmounting filesystem (524)
6635 - fix races between taking ns_lock and ldlm_lock_change_resource
6636 - fix races updating LOV export open file list
6637 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6638 - fix recovery NULL deref in ldlm_cli_cancel_unused
6639 - fix some DLM races by using new hash table for lock handles (419)
6640 - permit the client to specify desired inodes, at replay
6641 - duplicate requests when we queue them for replay reintegration
6642 - fix last_rcvd offset calculation
6643 - sync after each recovered transaction, so we always make progress
6644 - never, not always, ERESTART requests without transnos
6645 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6647 - skip replay if the MDS says that the client is already connected
6648 - don't check for a recovery-enabled export to match lctl's UUID
6649 - don't INC_USE_COUNT for phantom exports
6650 - don't crash when cleaning up phantom exports (567)
6651 - don't double-finish or set replay data for errored mdc_open requests
6652 - abort requests when they time out, so we don't get old replies
6653 - send/receive replies for AST messages again
6654 - if the client says that it doesn't have the lock, cancel it on the
6656 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6657 mark it as destroyed, it will all work out when decref is called
6658 - fix module use counts (22, 581)
6660 - ASTs now expect a reply (server cancels lock on error reply)
6662 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6665 - fix many simultaneous client startup (392)
6666 - fix dentry->d_it clobbering
6667 - credentials weren't being shipped for readdir/getattr operations
6668 - remove invalid assertions triggered during some concurrent MD
6670 - proper Lustre versions added (336, 389)
6671 - fix memory leak for create error case (398)
6672 - fix LOV locking bug that would get cli/srv out of sync
6673 - fix echo client over LOV (409)
6674 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6675 - fix double-O_EXCL intent crash (424)
6676 - avoid sending multiple lock CANCELs (352)
6678 - MDS can do multi-client recovery (modulo bugs in new code)
6680 - many updates, edits, cleanups
6682 2002-11-18 Phil Schwan <phil@clusterfs.com>
6685 - fix null d_it dereference (346)
6686 - fix full OST/dbench hang (333)
6687 - fix permission problem with file removal (286)
6688 - fix removal of OSCs from LOV when they fail
6689 - fix NULL deref during bulk timeout (214)
6690 - fix problems related to multiple filesystems on one MDS (241)
6691 - fixed serious subtle metadata locking bugs
6692 - free locks on clients when inodes are removed due to memory
6694 - fix inode pointer in lock data (285)
6695 - partial support for multiple MDS on a single host (241)
6696 - data locks weren't cancelled at clear_inode time (290, 311)
6697 - intent locks could lead to unbounded lock growth (205)
6698 - added a maximum lock count, an LRU list, and a flusher
6699 - fix multiple rename (365)
6700 - properly abstracted the echo client
6701 - OSC locked 1 byte too many; fixed
6702 - rewrote brw callback code:
6703 - fixed recovery bugs related to LOVs (306)
6704 - fixed too-many-pages-in-one-write crash (191)
6705 - fixed (again) crash in sync_io_timeout (214)
6706 - probably fixed callback-related race (385)
6708 - Add capability to MDS protocol
6709 - LDLM cancellations and callbacks on different portals
6711 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6714 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6715 - defer open object destruction to close time (601981, 138)
6716 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6717 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6718 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6719 - don't call obd_setattr in ll_file_release for destroyed objects
6720 * protocol change to lustre_msg: move |version| and add |flags|
6721 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6722 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6723 * added replay of create, unlink, link and rename operations during
6724 MDS failover; recovery should be much more robust now
6725 * remove failed OSCs from LOVs (only lov_create uses this so far)
6726 * the lustre-HOWTO was brought (more) up to date (582544)
6728 2002-10-23 Phil Schwan <phil@clusterfs.com>
6731 - in-use dentries weren't being reused properly (617851)
6732 - prevent multiple LDLM setup (599178)
6733 - fix LOV size calculations for truncate (617853)
6734 - fix client handling of MDS intent errors (POSIX)
6735 - fix permission bug in lovstripe.c test (624321)
6736 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6737 - truncate past end of file could corrupt data
6738 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6739 - a race in recovery could return ETIMEDOUT to apps (623947)
6740 - building outside the source directory was fixed
6741 * the lustre-HOWTO was brought (more) up to date (582544)
6742 * major progress was made on recovery functionality
6744 2002-10-10 Phil Schwan <phil@clusterfs.com>
6747 - recovery deadlock fix
6748 - rm -rf causes LBUG fix (617817)
6749 - file open by multiple tasks fix (618962)
6750 - directory permissions bugs (602707 and 620007)
6751 - journal_stop fixed with locking (611313)
6752 - O_APPEND failures resolved (618273, perhaps 614459)
6753 - lconf PATH fix (619770)
6754 - IA64 build fix (621450)
6755 - RPC buffer sizes scale with amount of memory
6757 2002-10-01 Phil Schwan <phil@clusterfs.com>
6760 - locks would be cancelled without throwing away data pages,
6761 resulting in inconsistent data (605627)
6762 - inode attributes were not always being refreshed (605627, 612449)
6763 - lconf now continues to cleanup after lctl reports an error
6764 - MDS now enforces user permissions (602707)
6765 - lprocfs cleanup fixed, but not yet enabled (614157)
6766 - fixed infinite server hang, should a client not respond to an AST
6767 - avoid going into recovery if user calls readlink() with a buffer
6768 that's too small (613941)
6769 - AST RPCs no longer require replies (614867) -- this may be changed
6770 - don't crash server if client sends an IOV that's too big (611336)
6771 - fixed lock conversion deadlock (611892)
6772 - fixed the following of symlinks (614622)
6773 * recovery: the server can remove locks from a client that dies, other
6774 clients can make progress
6775 * more extN patch fixes
6776 * compile-time configurable ptlrpc buffer allocations
6778 - collaborative read cache document
6779 - Lustre Lite Performance CDR document-in-progress
6781 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6784 - fix typo in patch-2.4.18
6786 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6789 - clear ptlrpc request each time in handle_incoming_request()
6790 - unlink of files now destroys the object on the OST
6792 2002-09-19 Peter Braam <braam@clusterfs.com>
6794 * add hard link support
6795 * change obdfile creation method
6796 * kernel patch changed
6798 2002-09-19 Peter Braam <braam@clusterfs.com>
6801 - stack overflow bug in extN fixed
6803 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6805 * documentation updates
6806 - add man pages for config tools
6807 - update tests/README to describe testing with new config tools
6808 - finish metadata API descriptions
6809 * bug fixes and cleanups
6810 - statfs workaround for 16TB limit
6811 - LOV stripe allocation improved, can stripe on subset of OSTs
6812 - LOV file size/IO offset was wrong for files > 4GB in size
6813 - object EA data was being dropped, caused files to be unreadable
6814 - memory overflow with non-LOV OST caused memory corruption
6815 - fixed regression tests to work with new config tools, obdfilter
6816 - fixed bug when directory size became larger than 1 block
6817 - fixed bug (for single client case) when PWD was deleted
6818 - invalidate local directory pages when doing intent-based ops
6819 - avoid LDLM oops when lock callback contained bad data
6821 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6823 * documentation updates
6824 * bug fixes and cleanups
6825 - configuration tools
6828 - 64-bit compile warnings
6829 - 64-bit internal statfs data
6831 * test_brw on persistent OST devices
6833 * lprocfs (disabled)
6835 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6837 * documentation updates
6838 * bug fixes and cleanups
6839 * configuration tools
6841 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6844 * many small fixes to 0_5_4
6845 * io/network handling
6846 * thinkos in MDS operations
6848 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6851 * crucial basic fixes to 0.5.3
6852 * IOR, Iozone work over Elan
6855 2002-08-07 Phil Schwan <phil@clusterfs.com>
6856 * version 0_5_3, our first alpha
6857 * we use the new Portals iovs
6858 * documentation updates
6859 * bug fixes and cleanups
6860 * small changes in the DLM wire protocol
6862 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6863 * version 0_5_1 with some initial stability,
6864 * locking on MD and file I/O.
6865 * documentation updates
6866 * several bug fixes since 0.5.0
6867 * small changes in wire protocol
6869 2002-07-18 Phil Schwan <phil@clusterfs.com>
6871 * delivered as Lustre Light Alpha
6872 * fixed a crash after handling invalid MDS requests
6873 * fixed directory pages for architectures with non-4k pages sizes
6875 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6876 * release version v0_4_4
6877 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6879 * Fixes a number of interruption problems with OST operations.
6880 * Update documentation for portals header changes
6881 * Move all wire protocol structs/defines to lustre_idl.h
6882 * Fixes symlink length bug.
6883 * Add tcpdump to repository.
6885 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6886 * release version v0_4_3
6887 * Fixes statfs for inodes on extN.
6888 * Fixes bug in runtests which would delete /etc/hosts.
6889 * Use 64-bit object IDs wherever possible (not into VFS though)
6890 Remove ost_get_info, which is unused by lustre, and out of date.
6892 2002-07-03 Peter Braam <braam@clusterfs.com>
6893 * release version v0_4_2 Fixes a lookup error (type not passed)
6894 * move forward to head of Portals
6895 * move forward to latest Lustre kernel
6897 2002-06-25 Peter Braam <braam@clusterfs.com>
6898 * release version v0_4_1. Hopefully stable on single node use.