3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on patched kernels:
6 2.6.32-431.29.2.el6 (RHEL6.5)
7 2.6.32-504.30.3.el6 (RHEL6.6)
8 2.6.32-573.26.1.el6 (RHEL6.7)
9 2.6.32-642.15.1.el6 (RHEL6.8)
10 2.6.32-696.30.1.el6 (RHEL6.9)
11 3.10.0-862.14.4.el7 (RHEL7.5)
12 3.10.0-957.el7 (RHEL7.6)
13 3.0.101-0.47.71 (SLES11 SP3)
14 3.0.101-107 (SLES11 SP4)
15 3.12.74-60.64.40 (SLES12 SP1)
16 4.4.120-92.70 (SLES12 SP2)
17 4.4.162-94.69 (SLES12 SP3)
18 vanilla linux 4.6.7 (ZFS only)
19 * Client known to build on unpatched kernels:
20 2.6.32-431.29.2.el6 (RHEL6.5)
21 2.6.32-504.30.3.el6 (RHEL6.6)
22 2.6.32-573.26.1.el6 (RHEL6.7)
23 2.6.32-642.15.1.el6 (RHEL6.8)
24 2.6.32-696.30.1.el6 (RHEL6.9)
25 2.6.32-754.10.1.el6 (RHEL6.10)
26 3.10.0-862.14.4.el7 (RHEL7.5)
27 3.10.0-957.el7 (RHEL7.6)
28 3.0.101-0.47.71 (SLES11 SP3)
29 3.0.101-107 (SLES11 SP4)
30 3.12.74-60.64.40 (SLES12 SP1)
31 4.4.120-92.70 (SLES12 SP2)
32 4.4.162-94.69 (SLES12 SP3)
33 4.4.0-131 (Ubuntu 16.04)
35 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
36 * Recommended ZFS / SPL version: 0.7.9
37 * Tested with ZFS / SPL version: 0.7.9
38 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
39 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
40 stack overflow. For more information, please refer to bugzilla 17630.
41 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
42 the re-exporting nodes to be mounted with "32bitapi" mount option
46 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
47 for currently supported client and server kernel versions.
48 * Server known to build on patched kernels:
49 2.6.32-431.29.2.el6 (RHEL6.5)
50 2.6.32-504.30.3.el6 (RHEL6.6)
51 2.6.32-573.26.1.el6 (RHEL6.7)
52 2.6.32-642.15.1.el6 (RHEL6.8)
53 2.6.32-696.30.1.el6 (RHEL6.9)
54 3.10.0-862.14.4.el7 (RHEL7.5)
55 3.10.0-957.1.3.el7 (RHEL7.6)
56 3.0.101-0.47.71 (SLES11 SP3)
57 3.0.101-107 (SLES11 SP4)
58 3.12.74-60.64.40 (SLES12 SP1)
59 4.4.120-92.70 (SLES12 SP2)
60 4.4.155-94.50 (SLES12 SP3)
61 vanilla linux 4.6.7 (ZFS only)
62 * Client known to build on unpatched kernels:
63 2.6.32-431.29.2.el6 (RHEL6.5)
64 2.6.32-504.30.3.el6 (RHEL6.6)
65 2.6.32-573.26.1.el6 (RHEL6.7)
66 2.6.32-642.15.1.el6 (RHEL6.8)
67 2.6.32-696.30.1.el6 (RHEL6.9)
68 2.6.32-754.6.3.el6 (RHEL6.10)
69 3.10.0-862.14.4.el7 (RHEL7.5)
70 3.10.0-957.1.3.el7 (RHEL7.6)
71 3.0.101-0.47.71 (SLES11 SP3)
72 3.0.101-107 (SLES11 SP4)
73 3.12.74-60.64.40 (SLES12 SP1)
74 4.4.120-92.70 (SLES12 SP2)
75 4.4.155-94.50 (SLES12 SP3)
76 4.4.0-131 (Ubuntu 16.04)
78 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
79 * Recommended ZFS / SPL version: 0.7.9
80 * Tested with ZFS / SPL version: 0.7.9
81 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
82 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
83 stack overflow. For more information, please refer to bugzilla 17630.
84 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
85 the re-exporting nodes to be mounted with "32bitapi" mount option
89 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
90 for currently supported client and server kernel versions.
91 * Server known to build on patched kernels:
92 2.6.32-431.29.2.el6 (RHEL6.5)
93 2.6.32-504.30.3.el6 (RHEL6.6)
94 2.6.32-573.26.1.el6 (RHEL6.7)
95 2.6.32-642.15.1.el6 (RHEL6.8)
96 2.6.32-696.30.1.el6 (RHEL6.9)
97 3.10.0-862.3.2.el7 (RHEL7.5)
98 3.0.101-0.47.71 (SLES11 SP3)
99 3.0.101-107 (SLES11 SP4)
100 3.12.74-60.64.40 (SLES12 SP1)
101 4.4.120-92.70 (SLES12 SP2)
102 4.4.155-94.50 (SLES12 SP3)
103 vanilla linux 4.6.7 (ZFS only)
104 * Client known to build on unpatched kernels:
105 2.6.32-431.29.2.el6 (RHEL6.5)
106 2.6.32-504.30.3.el6 (RHEL6.6)
107 2.6.32-573.26.1.el6 (RHEL6.7)
108 2.6.32-642.15.1.el6 (RHEL6.8)
109 2.6.32-696.30.1.el6 (RHEL6.9)
110 2.6.32-754.6.3.el6 (RHEL6.10)
111 3.10.0-862.3.2.el7 (RHEL7.5)
112 3.0.101-0.47.71 (SLES11 SP3)
113 3.0.101-107 (SLES11 SP4)
114 3.12.74-60.64.40 (SLES12 SP1)
115 4.4.120-92.70 (SLES12 SP2)
116 4.4.155-94.50 (SLES12 SP3)
117 4.4.0-131 (Ubuntu 16.04)
119 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
120 * Recommended ZFS / SPL version: 0.7.9
121 * Tested with ZFS / SPL version: 0.7.9
122 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
123 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
124 stack overflow. For more information, please refer to bugzilla 17630.
125 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
126 the re-exporting nodes to be mounted with "32bitapi" mount option
128 05-22-2018 Intel Corporation
130 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
131 for currently supported client and server kernel versions.
132 * Server known to build on patched kernels:
133 2.6.32-431.29.2.el6 (RHEL6.5)
134 2.6.32-504.30.3.el6 (RHEL6.6)
135 2.6.32-573.26.1.el6 (RHEL6.7)
136 2.6.32-642.15.1.el6 (RHEL6.8)
137 2.6.32-696.30.1.el6 (RHEL6.9)
138 3.10.0-862.9.1.el7 (RHEL7.5)
139 3.0.101-0.47.71 (SLES11 SP3)
140 3.0.101-107 (SLES11 SP4)
141 3.12.74-60.64.40 (SLES12 SP1)
142 4.4.120-92.70 (SLES12 SP2)
143 4.4.132-94.33 (SLES12 SP3)
144 vanilla linux 4.6.7 (ZFS only)
145 * Client known to build on unpatched kernels:
146 2.6.32-431.29.2.el6 (RHEL6.5)
147 2.6.32-504.30.3.el6 (RHEL6.6)
148 2.6.32-573.26.1.el6 (RHEL6.7)
149 2.6.32-642.15.1.el6 (RHEL6.8)
150 2.6.32-696.30.1.el6 (RHEL6.9)
151 3.10.0-862.9.1.el7 (RHEL7.5)
152 3.0.101-0.47.71 (SLES11 SP3)
153 3.0.101-107 (SLES11 SP4)
154 3.12.74-60.64.40 (SLES12 SP1)
155 4.4.120-92.70 (SLES12 SP2)
156 4.4.133-94.33 (SLES12 SP3)
157 4.4.0-116 (Ubuntu 16.04)
159 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
160 * Recommended ZFS / SPL version: 0.7.9
161 * Tested with ZFS / SPL version: 0.7.9
162 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
163 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
164 stack overflow. For more information, please refer to bugzilla 17630.
165 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
166 the re-exporting nodes to be mounted with "32bitapi" mount option
168 01-31-2018 Intel Corporation
170 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
171 for currently supported client and server kernel versions.
172 * Server known to build on patched kernels:
173 2.6.32-431.29.2.el6 (RHEL6.5)
174 2.6.32-504.30.3.el6 (RHEL6.6)
175 2.6.32-573.26.1.el6 (RHEL6.7)
176 2.6.32-642.15.1.el6 (RHEL6.8)
177 2.6.32-696.23.1.el6 (RHEL6.9)
178 3.10.0-693.21.1.el7 (RHEL7.4)
179 3.0.101-0.47.71 (SLES11 SP3)
180 3.0.101-107 (SLES11 SP4)
181 3.12.74-60.64.40 (SLES12 SP1)
182 4.4.120-92.70 (SLES12 SP2)
183 4.4.120-94.17 (SLES12 SP3)
184 vanilla linux 4.6.7 (ZFS only)
185 * Client known to build on unpatched kernels:
186 2.6.32-431.29.2.el6 (RHEL6.5)
187 2.6.32-504.30.3.el6 (RHEL6.6)
188 2.6.32-573.26.1.el6 (RHEL6.7)
189 2.6.32-642.15.1.el6 (RHEL6.8)
190 2.6.32-696.23.1.el6 (RHEL6.9)
191 3.10.0-693.21.1.el7 (RHEL7.4)
192 3.0.101-0.47.71 (SLES11 SP3)
193 3.0.101-107 (SLES11 SP4)
194 3.12.74-60.64.40 (SLES12 SP1)
195 4.4.120-92.70 (SLES12 SP2)
196 4.4.120-94.17 (SLES12 SP3)
197 4.4.0-116 (Ubuntu 16.04)
199 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
200 * Recommended ZFS / SPL version: 0.7.5
201 * Tested with ZFS / SPL version: 0.7.5
202 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
203 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
204 stack overflow. For more information, please refer to bugzilla 17630.
205 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
206 the re-exporting nodes to be mounted with "32bitapi" mount option
208 --------------------------------------------------------------------------------
210 12-11-2017 Intel Corporation
212 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
213 for currently supported client and server kernel versions.
214 * Server known to build on patched kernels:
215 2.6.32-431.29.2.el6 (RHEL6.5)
216 2.6.32-504.30.3.el6 (RHEL6.6)
217 2.6.32-573.26.1.el6 (RHEL6.7)
218 2.6.32-642.15.1.el6 (RHEL6.8)
219 2.6.32-696.20.1.el6 (RHEL6.9)
220 3.10.0-693.17.1.el7 (RHEL7.4)
221 3.0.101-0.47.71 (SLES11 SP3)
222 3.0.101-107 (SLES11 SP4)
223 3.12.74-60.64.40 (SLES12 SP1)
224 4.4.103-92.56 (SLES12 SP2)
225 4.4.103-6.38 (SLES12 SP3)
226 vanilla linux 4.6.7 (ZFS only)
227 * Client known to build on unpatched kernels:
228 2.6.32-431.29.2.el6 (RHEL6.5)
229 2.6.32-504.30.3.el6 (RHEL6.6)
230 2.6.32-573.26.1.el6 (RHEL6.7)
231 2.6.32-642.15.1.el6 (RHEL6.8)
232 2.6.32-696.20.1.el6 (RHEL6.9)
233 3.10.0-693.17.1.el7 (RHEL7.4)
234 3.0.101-0.47.71 (SLES11 SP3)
235 3.0.101-107 (SLES11 SP4)
236 3.12.74-60.64.40 (SLES12 SP1)
237 4.4.103-92.56 (SLES12 SP2)
238 4.4.103-6.38 (SLES12 SP3)
240 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
241 * Recommended ZFS / SPL version: 0.7.3
242 * Tested with ZFS / SPL version: 0.7.3
243 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
244 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
245 stack overflow. For more information, please refer to bugzilla 17630.
246 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
247 the re-exporting nodes to be mounted with "32bitapi" mount option
249 --------------------------------------------------------------------------------
251 10-02-2017 Intel Corporation
253 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
254 for currently supported client and server kernel versions.
255 * Server known to build on patched kernels:
256 2.6.32-431.29.2.el6 (RHEL6.5)
257 2.6.32-504.30.3.el6 (RHEL6.6)
258 2.6.32-573.26.1.el6 (RHEL6.7)
259 2.6.32-642.15.1.el6 (RHEL6.8)
260 2.6.32-696.10.3.el6 (RHEL6.9)
261 3.10.0-693.2.2.el7 (RHEL7.4)
262 3.0.101-0.47.71 (SLES11 SP3)
263 3.0.101-107 (SLES11 SP4)
264 3.12.74-60.64.40 (SLES12 SP1)
265 4.4.74-92.35 (SLES12 SP2)
266 vanilla linux 4.6.7 (ZFS only)
267 * Client known to build on unpatched kernels:
268 2.6.32-431.29.2.el6 (RHEL6.5)
269 2.6.32-504.30.3.el6 (RHEL6.6)
270 2.6.32-573.26.1.el6 (RHEL6.7)
271 2.6.32-642.15.1.el6 (RHEL6.8)
272 2.6.32-696.10.3.el6 (RHEL6.9)
273 3.10.0-693.2.2.el7 (RHEL7.4)
274 3.0.101-0.47.71 (SLES11 SP3)
275 3.0.101-107 (SLES11 SP4)
276 3.12.74-60.64.40 (SLES12 SP1)
277 4.4.74-92.35 (SLES12 SP2)
279 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
280 * Recommended ZFS / SPL version: 0.7.1
281 * Tested with ZFS / SPL version: 0.7.1
282 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
283 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
284 stack overflow. For more information, please refer to bugzilla 17630.
285 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
286 the re-exporting nodes to be mounted with "32bitapi" mount option
288 --------------------------------------------------------------------------------
290 2017-07-05 Intel Corporation
292 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
293 for currently supported client and server kernel versions.
294 * Server known to build on patched kernels:
295 2.6.32-431.29.2.el6 (RHEL6.5)
296 2.6.32-504.30.3.el6 (RHEL6.6)
297 2.6.32-573.26.1.el6 (RHEL6.7)
298 2.6.32-642.15.1.el6 (RHEL6.8)
299 2.6.32-696.3.1.el6 (RHEL6.9)
300 3.10.0-514.21.1.el7 (RHEL7.3)
301 3.0.101-0.47.71 (SLES11 SP3)
302 3.0.101-100 (SLES11 SP4)
303 3.12.74-60.64.40 (SLES12 SP1)
304 4.4.59-92.17 (SLES12 SP2)
305 vanilla linux 4.6.7 (ZFS only)
306 * Client known to build on unpatched kernels:
307 2.6.32-431.29.2.el6 (RHEL6.5)
308 2.6.32-504.30.3.el6 (RHEL6.6)
309 2.6.32-573.26.1.el6 (RHEL6.7)
310 2.6.32-642.15.1.el6 (RHEL6.8)
311 2.6.32-696.3.1.el6 (RHEL6.9)
312 3.10.0-514.21.1.el7 (RHEL7.3)
313 3.0.101-0.47.71 (SLES11 SP3)
314 3.0.101-100 (SLES11 SP4)
315 3.12.74-60.64.40 (SLES12 SP1)
316 4.4.59-92.17 (SLES12 SP2)
318 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
319 * Recommended ZFS / SPL version: 0.6.5.9
320 * Tested with ZFS / SPL version: 0.6.5.9
321 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
322 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
323 stack overflow. For more information, please refer to bugzilla 17630.
324 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
325 the re-exporting nodes to be mounted with "32bitapi" mount option
327 --------------------------------------------------------------------------------
329 12-07-2016 Intel Corporation
331 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
332 for currently supported client and server kernel versions.
333 * Server known to build on patched kernels:
334 2.6.32-431.29.2.el6 (RHEL6.5)
335 2.6.32-504.30.3.el6 (RHEL6.6)
336 2.6.32-573.26.1.el6 (RHEL6.7)
337 2.6.32-642.11.1.el6 (RHEL6.8)
338 3.10.0-514.el7 (RHEL7.3)
339 3.0.101-0.47.71 (SLES11 SP3)
340 3.0.101-80 (SLES11 SP4)
341 3.12.62-60.64.8 (SLES12 SP1)
342 vanilla linux 4.5.7 (ZFS only)
343 * Client known to build on unpatched kernels:
344 2.6.32-431.29.2.el6 (RHEL6.5)
345 2.6.32-504.30.3.el6 (RHEL6.6)
346 2.6.32-573.26.1.el6 (RHEL6.7)
347 2.6.32-642.11.1.el6 (RHEL6.8)
348 3.10.0-514.el7 (RHEL7.3)
349 3.0.101-0.47.71 (SLES11 SP3)
350 3.0.101-80 (SLES11 SP4)
351 3.12.62-60.64.8 (SLES12 SP1)
353 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
354 * Recommended ZFS / SPL version: 0.6.5.8
355 * Tested with ZFS / SPL version: 0.6.5.8
356 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
357 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
358 stack overflow. For more information, please refer to bugzilla 17630.
359 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
360 the re-exporting nodes to be mounted with "32bitapi" mount option
362 --------------------------------------------------------------------------------
364 02-29-2016 Intel Corporation
366 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
367 for currently supported client and server kernel versions.
368 * Server known to build on patched kernels:
369 2.6.32-431.29.2.el6 (RHEL6.5)
370 2.6.32-504.30.3.el6 (RHEL6.6)
371 2.6.32-573.12.1.el6 (RHEL6.7)
372 3.10.0-327.3.1.el7 (RHEL7.2)
373 3.0.101-0.47.71 (SLES11 SP3)
374 3.0.101-68 (SLES11 SP4)
375 vanilla linux 4.2.1 (ZFS only)
376 * Client known to build on unpatched kernels:
377 2.6.32-431.29.2.el6 (RHEL6.5)
378 2.6.32-504.30.3.el6 (RHEL6.6)
379 2.6.32-573.12.1.el6 (RHEL6.7)
380 3.10.0-327.3.1.el7 (RHEL7.2)
381 3.0.101-0.47.71 (SLES11 SP3)
382 3.0.101-68 (SLES11 SP4)
385 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
386 * Recommended ZFS / SPL version: 0.6.4.2
387 * Tested with ZFS / SPL version: 0.6.4.2
388 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
389 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
390 stack overflow. For more information, please refer to bugzilla 17630.
391 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
392 the re-exporting nodes to be mounted with "32bitapi" mount option
394 --------------------------------------------------------------------------------
396 03-10-2015 Intel Corporation
398 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
399 for currently supported client and server kernel versions.
400 * Server known to build on patched kernels:
401 2.6.32-431.29.2.el6 (RHEL6.5)
402 2.6.32-504.8.1.el6 (RHEL6.6)
403 3.0.101-0.46 (SLES11 SP3)
404 * Client known to build on unpatched kernels:
405 2.6.32-431.29.2.el6 (RHEL6.5)
406 2.6.32-504.8.1.el6 (RHEL6.6)
407 3.10.0-123.20.1.el7 (RHEL7)
408 3.0.101-0.46 (SLES11 SP3)
409 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
410 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
411 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
412 stack overflow. For more information, please refer to bugzilla 17630.
413 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
414 the re-exporting nodes to be mounted with "32bitapi" mount option
416 Severity : enhancement
418 Description: control OST-index in IDIF via ROCOMPAT flag.
419 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
420 last_rcvd file. For new formatted OST device, it will be auto set;
421 for the case of upgrading from old OST device, you can enable it
422 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
423 enabled, for new created OST-object, its IDIF-in-LMA will contain
424 the OST-index; for the existing OST-object, the OSD will convert
425 old format IDIF as new format IDIF with OST-index stored in the
426 LMA EA when accessing such OST-object or via OI scrub. Once such
427 flag is enabled, it cannot be reverted back, so the system cannot
428 be downgraded to the orignal incompatible version.
430 --------------------------------------------------------------------------------
432 07-30-2014 Intel Corporation
434 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
435 for currently supported client and server kernel versions.
436 * Server known to build on patched kernels:
437 2.6.32-431.20.3.el6 (RHEL6.5)
438 3.0.101-0.31 (SLES11 SP3)
439 * Client known to build on unpatched kernels:
440 2.6.32-431.20.3.el6 (RHEL6.5)
441 3.0.101-0.31 (SLES11 SP3)
444 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
445 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
446 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
447 stack overflow. For more information, please refer to bugzilla 17630.
448 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
449 the re-exporting nodes to be mounted with "32bitapi" mount option
451 --------------------------------------------------------------------------------
453 10-31-2013 Intel Corporation
455 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
456 * Server builds for kernels:
457 2.6.32-358.18.1.el6 (RHEL6.4)
458 * Client builds for unpatched kernels:
459 2.6.32-358.18.1.el6 (RHEL6.4)
460 3.0.92-0.8 (SLES11 SP3)
462 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
463 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
464 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
465 stack overflow. For more information, please refer to bugzilla 17630.
466 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
467 the re-exporting nodes to be mounted with "32bitapi" mount option
469 --------------------------------------------------------------------------------
471 05-30-2013 Intel Corporation
473 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
474 * Server builds for kernels:
475 2.6.32-358.6.2.el6 (RHEL6.4)
476 * Client builds for unpatched kernels:
477 2.6.32-358.6.2.el6 (RHEL6.4)
478 3.0.74-0.6.6 (SLES11 SP2)
480 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
481 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
482 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
483 stack overflow. For more information, please refer to bugzilla 17630.
484 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
485 the re-exporting nodes to be mounted with "32bitapi" mount option
487 --------------------------------------------------------------------------------
489 9-30-2012 Whamcloud, Inc.
491 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
492 * Server support for kernels:
493 2.6.32-279.5.1.el6 (RHEL6)
494 * Client support for unpatched kernels:
495 2.6.32-279.5.1.el6 (RHEL6)
497 2.6.18-238.19.1.el5 (RHEL5)
498 2.6.16.60-0.69.1 (SLES10)
499 * Recommended e2fsprogs version: 1.41.90.wc4
500 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
501 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
502 stack overflow. For more information, please refer to bugzilla 17630.
503 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
504 the re-exporting nodes to be mounted with "32bitapi" mount option
506 Severity : enhancement
508 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
510 Severity : enhancement
512 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
514 Severity : enhancement
516 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
518 Severity : enhancement
520 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
522 Severity : enhancement
524 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
526 --------------------------------------------------------------------------------
528 04-30-2012 Whamcloud, Inc.
530 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
531 * Server support for kernels:
532 2.6.32-220.4.2.el6 (RHEL6)
533 * Client support for unpatched kernels:
534 2.6.32-220.4.2.el6 (RHEL6)
536 2.6.18-238.19.1.el5 (RHEL5)
537 2.6.16.60-0.69.1 (SLES10)
538 * Recommended e2fsprogs version: 1.41.90.wc4
539 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
540 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
541 stack overflow. For more information, please refer to bugzilla 17630.
542 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
543 the re-exporting nodes to be mounted with "32bitapi" mount option
545 Severity : enhancement
547 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
549 Severity : enhancement
551 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
553 Severity : enhancement
555 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
557 Severity : enhancement
559 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
561 --------------------------------------------------------------------------------
563 09-30-2011 Whamcloud, Inc.
565 * Server support for kernels:
566 2.6.18-238.19.1.el5 (RHEL5)
567 2.6.32-131.6.1.el6 (RHEL6)
568 * Client support for unpatched kernels:
569 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
570 2.6.18-238.19.1.el5 (RHEL5)
571 2.6.32-131.6.1.el6 (RHEL6)
572 2.6.29.4-167.fc11 (FC11)
573 2.6.16.60-0.69.1 (SLES10)
575 * Recommended e2fsprogs version: 1.41.90.wc3
576 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
577 Lustre filesystem with 4K stack may cause a stack overflow. For more
578 information, please refer to bugzilla 17630.
579 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
580 mounted with "32bitapi" mount option
582 Severity : enhancement
583 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
585 Severity : enhancement
586 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
588 Severity : enhancement
589 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
592 Frequency : only with SLES10
594 Description: Use OFED "KMP" provided by Novell
595 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
596 Lustre is now built against this package. That means you need to
597 install the ofed-kmp package from Novell for the patchless client.
598 Note that the ofed-kmp that Novell ships may not exactly match the
599 kernel version but should still be compatible.
601 Severity : enhancement
603 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
604 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
606 --------------------------------------------------------------------------------
608 2010-07-15 Oracle, Inc.
610 * Server support for kernels:
611 2.6.18-164.11.1.el5 (RHEL 5)
612 2.6.18-164.11.1.0.1.el5 (OEL 5)
613 * Client support for unpatched kernels:
614 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
615 2.6.18-164.11.1.el5 (RHEL 5),
616 2.6.18-164.11.1.0.1.el5 (OEL 5)
617 2.6.16.60-0.42.8 (SLES 10),
619 2.6.29.4-167.fc11 (FC11)
620 * Recommended e2fsprogs version: 1.41.10-sun2
622 Severity : enhancement
624 Description: Distribute lfs_migrate script with lustre
625 Details : The lfs_migrate mentioned in the manual is now included in the
626 lustre tree in lustre/scripts and will distributed in lustre RPMs.
628 Severity : enhancement
630 Description: Failover nid list added to import proc file.
632 Severity : enhancement
634 Description: "som_preview" mount option required to use SOM
635 Details : The clients will need to use som_preview mount option
639 Description: Lustre does not do 1MB IOs to HW RAID
640 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
642 Severity : enhancement
644 Description: lfs quota output is not very convenient for awk/sed-parsing
645 Details : Some positions in lfs quota output table could be empty or
646 non-empty which made it hard to parse it with scripts, now
647 a dash is put instead of space where there is not supposed
652 Description: Update to OFED 1.5.1
653 Details : OFED 1.5.1 will be used in environments that don't use
654 vendor supplied OFED.
658 Description: Remove lnet. prefix from lctl params display
659 Details : Also renamed lnet.memused to lnet_memused to
664 Description: Use conf_param -d to remove permanant settings.
668 Description: Add list_param -R
669 Details : lctl list_param now supports recursive mode.
672 Bugzilla : 21829, 22312
673 Description: Fix llobdstat.
674 Details : 1. Now updating the write_bytes counter
675 2. Port script fixes from 1.8.
679 Description: decrease the usage of memory on clients.
680 Details : 1. On clients, recycle dentries and inodes unused.
681 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
686 Description: llobdstat fix and enhancement
687 Details : add a counter to set a limit to how many samples will be returned
688 fix a wildcard in the path to limit to obdfilter stats only
692 Description: lfs check only allowed for root
693 Details : Code cleanups and added a regresion test of non-root use of
698 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
699 Details : Fix hr->hr_index overload.
703 Description: Fix FIEMAP functionality.
707 Description: FC11 patchless client support added
711 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
712 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
716 Description: New acc-sm script and various test-framework enhancments.
720 Description: Previously landed patch was reverted.
724 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
728 Description: Add ldiskfs maxdirsize mount option.
732 Description: Fix mkfs.lustre support for 16TB LUNs.
736 Description: Print mmp_check_interval
737 Details : print mmp_check_interval and make it possible to abort
738 mount operation in case it takes too long.
742 Description: Add cascading_rw to lustre mpi tests.
746 Description: Store SOM data in EA.
747 Details : Also several SOM related cleanups.
751 Description: Add refcount to md_open_data
755 Description: Use refcount on lov_request_set to prevent memory corruption.
759 Description: Statahead: use parent dir's i_mutex for synchronization.
763 Description: loadgen improvements.
767 Description: Enable peer health detection in ksocklnd and ko2iblnd.
771 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
775 Description: Fix race on obd_devcie:md_stats
779 Description: Fix SLES 11 client builds.
780 Details : Resolved uninitalized pointer gcc warnings.
784 Description: Allow non-root access for "lfs check".
785 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
789 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
791 Severity : enhancement
793 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
794 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
798 Description: Sanity test_27p fails.
799 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
800 Sanity tests 27 were cleaned up and improved.
804 Description: Unify req format on client/servers
805 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
806 buffers and swabbers approach..
811 Description: drop join file stuff.
813 Severity : enhancement
815 Description: Add 2.6.27 and SLES11 (patchless client) support
819 Description: Fix in ptlrpc_expire_one_request() to print the signed time
822 Severity : enhancement
824 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
827 Severity : enhancement
829 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
831 Severity : enhancement
833 Description: correctly handle big reply message.
834 Details : send LNet event if reply is bigger than buffer and adjust this buffer
839 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
841 Severity : enhancement
843 Description: Add support for OEL5.
847 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
849 Severity : enhancement
851 Description: "weak-modules" support
852 Details : Implement "weak-modules" support which enables kernel modules
853 to be used with any kernel that implements the same kABI. In
854 order to achieve this modules are now installed in
855 /lib/modules/$(uname -r)/updates/kernel on all distributions.
859 Description: client could not reconnect to OST because of an active request.
860 Details : The client now retries to reconnect to the same server, if a
861 connect request failed with EBUSY or -EAGAIN. Test case 35b
862 conf-sanity.sh has been created.
866 Description: Don't allow make backward step on assignin osc next id.
867 Details : race between allocation next id and ll_sync thread can be cause
868 of set wrong osc next id and can be kill valid ost objects.
872 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
876 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
880 Description: Error handling in osc_statfs_interpret() has been improved.
881 Details : Check in osc_statfs_interpret() for EBADR.
885 Description: Do not send statfs() requests to OSTs disabled by administrator.
886 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
888 Severity : enhancement
890 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
892 Severity : enhancement
894 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
896 Severity : enhancement
898 Description: Enhancement for lfs(1) command to use numeric uid/gid.
902 Description: Improved cleanup procedure for conf-sanity test_46a.
906 Description: Do not start bulk transfer, if client tried to reconnect.
907 Details : Do not start bulk transfer, if client tried to reconnect,
908 but failed due to active rpcs. Check for exp_abort_active_req
909 has been implemented.
913 Description: Prevent inconsistences between linux and lustre mount structures.
914 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
915 Make the sleep interruptible.
917 Severity : enhancement
919 Description: provide server to client comms path
920 Details : server to client communications path via new LDLM_SET_INFO rpc
924 Description: truncate starts GFP_FS allocation under transaction causing deadlock
925 Details : ldiskfs_truncate calls grab_cache_page which may start page
926 allocation under an open transaction. This may lead to
927 calling prune_icache with consequent lustre reentrance.
931 Description: send correctly lsm on open replay
932 Details : MDS is trust to LSM size on replay open, but client can set wrong size
937 Description: enable client interface failover
938 Details : When a child reconnects from another NID, properly update export
939 nid hash position and ldlm reverse import.
941 Severity : enhancement
943 Description: Communicate OST degraded/readonly state via statfs to MDS
944 Details : Flags in the statfs returned from OSTs indicate whether the
945 OST is in a degraded RAID state, or if the filesystem has
946 turned read-only after a filesystem error is detected.
951 Description: don't panic if EPROTO was hit when reading symlink
952 Details : correctly handling request reference in error cases.
954 Severity : enhancement
956 Description: Add LustreNetLink, a kernel-userspace communcation path.
958 Severity : enhancement
960 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
965 Description: access to llog context before init.
966 Details : move handling CATALOGS file at osc layer and forbid access to llog
970 Frequency : rare, only when client is writing to OST on local node
972 Description: Avoid deadlock for local client writes
973 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
974 memory freeing context. This allows OST threads to set the
975 PF_MEMALLOC flag on task structures in order to allocate memory
976 from reserved pools and complete IO.
977 Use GFP_HIGHUSER for OST allocations for non-local client writes,
978 so that the OST threads generate memory pressure and allow
979 inactive pages to be reclaimed.
981 Severity : enhancement
983 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
988 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
989 Details : move ldlm namespace creation in setup phase to avoid grab
990 _lprocfs_lock with cli_sem held.
994 Description: Temporarily disable grant shrink.
995 Details : Disable the feature for debugging.
998 Frequency : only when formatting test filesystems
1000 Description: Unable to run several mkfs.lustre on loop devices at the same time
1001 Details : mkfs.lustre returns error 256 on the concurrent loop devices
1002 formatting. The solution is to proper handle the error.
1004 Severity : enhancement
1006 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
1008 Severity : enhancement
1010 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
1015 Description: resolve race between obd_disconnect and class_disconnect_exports
1016 Details : if obd_disconnect will be called to already disconnected export he
1017 forget release one reference and osc module can't unloaded.
1019 Severity : enhancement
1021 Description: Allow tuning service thread via /proc
1022 Details : For each service a new
1023 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
1024 created that can be used to set min/max thread counts, and get the
1025 current number of running threads.
1030 Descriptoin: don't return error if have partially created objects for file.
1031 Details : lov_update_create_set uses set->set_success as index for created
1032 objects, so if some requests will be failed, they will have hole
1033 at end of array and qos_shrink_lsm for allocate correct lsm.
1035 Severity : enhancement
1037 Description: Update OFED support to 1.4.1
1042 Description: Reduce small size read RPC
1043 Details : Set read-ahead limit for every file and only do read-ahead when
1044 available read-ahead pages are bigger than 1M to avoid small size
1047 Severity : enhancement
1049 Description: move AT tunable parameters for more consistent usage
1050 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1052 Severity : enhancement
1054 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1055 Details : allow skip disconnected ost for send statfs request and hide error
1061 Description: fix racy locking of mballoc block bitmaps causing BUG
1062 Details : The locking of the mballoc buddy bitmap and the in-memory
1063 block bitmap was using two different spin locks in some
1064 cases. This made it possible to incorrectly access the
1065 mballoc bitmap while another process was modifying it,
1066 causing a sanity assertion to fail. While no on-disk corruption
1067 was reported, there was some risk of this happening.
1070 Frequency : rare, on llog test 6
1072 Descriptoin: don't allow connect to already connected import
1073 Details : allowing connect to already connected import is hide connecting problem.
1078 Descriptoin: allow kill process which wait statahead result
1079 Details : for some reasons 'ls' can stick in waiting result from statahead,
1080 in this case need way for kill this process.
1082 Severity : enhancement
1084 Description: Add state history info file, enhance import info file
1085 Details : Track import connection state changes in a new osc/mdc proc file;
1086 add overview-type data to the osc/mdc import proc file.
1088 Severity : enhancement
1090 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1092 Severity : enhancement
1094 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1099 Descriptoin: OST grant too much space to client even there are not enough space.
1100 Details : Client will shrink its grant cache to OST if there are no write
1101 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1102 this grant cache if there are already not enough avaible space
1103 (left_space < total_clients * 32M).
1106 Frequency : start MDS on uncleanly shutdowned MDS device
1108 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1109 Details : stay in waiting mds<>ost recovery finished produce random bugs
1110 due race between two ll_sync thread for one lov target. send
1111 ACTIVATE event only if connect realy finished and import have
1115 Frequency : rare, connect and disconnect target at same time
1117 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1118 Details : don't call obd_disconnect under lov_lock. this long time
1119 operation and can block ptlrpcd which answer to connect request.
1124 Descriptoin: don't lose wakeup for imp_recovery_waitq
1125 Details : recover_import_no_retry or invalidate_import and import_close can
1126 both sleep on imp_recovery_waitq, but we was send only one wakeup
1130 Frequency : always with long access acl
1132 Descriptoin: mds can't pack reply with long acl.
1133 Details : mds don't control size of acl but they limited by reint/getattr
1136 Severity : enhancement
1138 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1140 Severity : enhancement
1142 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1145 Frequency : start MDS on uncleanly shutdowned MDS device
1147 Descriptoin: aborting recovery hang on MDS
1148 Details : don't throttle destroy RPCs for the MDT.
1151 Frequency : on remount
1153 Description: external journal device not working after the remount
1154 Details : clear dev_rdonly flag for external journal devices in
1160 Description: shutdown vs evict race
1161 Details : client_disconnect_export vs connect request race.
1162 if client will evicted at this time - we start invalidate
1163 thread without referece to import and import can be freed
1167 Frequency : rare, need acl's on inode.
1169 Description: client can't handle ost additional correctly
1170 Details : if ost was added after client connected to mds client can have
1171 hit lnet_try_match_md ... to big messages to wide striped files.
1172 in this case need teach client to handle config events about add
1173 lov target and update client max ea size at that event.
1175 Severity : enhancement
1177 Description: Changelogs
1178 Details : Changelogs are a lightweight mechanism to track filesystem
1179 metadata and namespace changes. The changelog is recorded
1180 permanently on the MDTs, and is periodically "consumed" / purged
1181 when records are no longer needed.
1183 Severity : enhancement
1185 Description: compact fld format with extents
1186 Details : Store range of seq rather than every seq in FLD. Seq
1187 controller update FLD rather than clients. In Case of CMD, mdt0
1188 has FLD, all other metadata server act as non persistent proxy
1189 for FLD queries and cache fld entries in fld cache.
1194 Description: don't skip ost target if they assigned to file
1195 Details : Drop slow OSCs if we can, but not for requested start idx.
1196 This means "if OSC is slow and it is not the requested
1197 start OST, then it can be skipped, otherwise skip it only
1198 if it is inactive/recovering/out-of-space.
1202 Description: more cleanup in mds_lov
1203 Details : not send LOV EA under replay, we can't know about they size at this
1204 time. Don't allow client connect to mds before any ost connected,
1205 for avoid problems with LOV EA size and returning EIO to client.
1207 Severity : enhancement
1209 Description: Interoperability at server side (Disk interoperability)
1211 Severity : enhancement
1213 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1215 Severity : enhancement
1217 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1219 Severity : enhancement
1221 Description: New client IO stack (CLIO).
1223 Severity : enhancement
1225 Description: Commit on sharing. Eliminate inter-client dependencies between
1226 uncommitted transactions by doing transaction commits.
1227 Thereby clients may recovery independently.
1230 Frequency : Create a symlink file with a very long name
1232 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1233 Details : If there is no extra space in the request for early cancels,
1234 ldlm_req_handles_avail() returns 0 instead of a negative value.
1236 Severity : enhancement
1238 Description: Add /proc entry for import status
1239 Details : The mdc, osc, and mgc import directories now have
1240 an import directory that contains useful import data for debugging
1241 connection problems.
1243 Severity : enhancement
1245 Description: Re-disable certain /proc logging
1246 Details : Enable and disable client's offset_stats, extents_stats and
1247 extents_stats_per_process stats logging on the fly.
1250 Frequency : Only on FC kernels 2.6.22+
1252 Description: oops in statahead
1253 Details : Do not drop reference count for the dentry from VFS when lookup,
1254 VFS will do that by itself.
1256 Severity : enhancement
1258 Description: Generic /proc file permissions
1259 Details : Set /Proc file permissions in a more generic way to enable non-
1260 root users operate on some /proc files.
1264 Description: Hitting mdc_commit_close() ASSERTION
1265 Details : Properly handle request reference release in
1266 ll_release_openhandle().
1270 Description: quota recovery deadlock during mds failover
1271 Details : This patch includes att18982, att18236, att18237 in bz14840.
1273 1. fix osts hang when mds does failover with quotaon
1274 2. prevent watchdog storm when osts threads wait for the
1279 Frequency : only patchless client
1280 Description: add workaround for race between add/remove dentry from hash
1282 Severity : enhancement
1284 Description: Allow OST glimpses to return PW locks
1288 Description: LBUG when llog conf file is full
1289 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1293 Description: Prevent import from entering FULL state when server in recovery
1297 Description: service mount cannot take device name with ":"
1298 Details : Only when device name contains ":/" will mount treat it as
1304 Description: replace ptlrpcd with the statahead thread to interpret the async
1305 statahead RPC callback
1309 Frequency : on recovery
1310 Description: I/O failures after umount during fail back
1311 Details : if client reconnected to restarted server we need join to recovery
1312 instead of find server handler is changed and process self
1313 eviction with cancel all locks.
1315 Severity : enhancement
1317 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1319 Severity : enhancement
1321 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1323 Severity : enhancement
1325 Description: Upcall on Lustre log has been dumped
1326 Details : Allow for a user mode script to be called once a Lustre log has
1327 been dumped. It passes the filename of the dumped log to the
1328 script, the location of the script can be specified via
1329 /proc/sys/lnet/debug_log_upcall.
1334 Description: avoid idr_remove called for id which is not allocated.
1335 Details : Move assigment s_dev for clustered nfs to end of initialization,
1336 to avoid problem with error handling.
1341 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1342 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1343 into CONN_USED_HASH and this prodice warning when put connection
1344 again in unused hash.
1347 Severity : enhancement
1349 Description: File striping can now be set to use an arbitrary pool of OSTs.
1351 Severity : enhancement
1353 Description: Export bytes_read/bytes_write count on OSC/OST.
1357 Description: Early reply size mismatch, MGC loses connection
1358 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1359 the connect flags are properly negotiated.
1364 Description: quotas are not honored with O_DIRECT
1365 Details : all writes with the flag O_DIRECT will use grants which leads to
1366 this problem. Now using OBD_BRW_SYNC to guard this.
1370 Description: add quota statistics
1371 Details : 1. sort out quota proc entries and proc code.
1372 2. add quota statistics
1374 Severity : enhancement
1376 Description: enable quota support for HEAD.
1380 Description: Properly propagate oinfo flags from lov to osc for statfs
1381 Details : restore missing copy oi_flags to lov requests.
1383 Severity : enhancement
1385 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1386 : l_getgroups(8), lst(8), routerstat(8)
1387 Details : included man pages for llobdstat(8), llstat(8),
1388 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1390 Severity : enhancement
1392 Description: configure's --enable-quota should check the
1393 : kernel .config for CONFIG_QUOTA
1394 Details : configure is terminated if --enable-quota is passed but
1395 : no quota support is in kernel
1399 Description: Remove portals compatibility
1400 Details : Remove portals compatibility, not interoperable with releases
1405 Description: Resolve device initialization race
1406 Details : Prevent proc handler from accessing devices added to the
1407 obd_devs array but yet be intialized.
1409 Severity : enhancement
1411 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1413 Severity : enhancement
1415 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1420 Description: Using wrong pointer in osc_brw_prep_request
1421 Details : Access to array[-1] can produce panic if kernel compiled with
1422 CONFIG_PAGE_ALLOC enabled
1426 Description: Client runs out of low memory
1427 Details : Consider only lowmem when counting initial number of llap pages
1431 Description: *optional* service tags registration
1432 Details : if the "service tags" package is installed on a Lustre node
1433 When the filesystem is mounted, a local-node service tag will
1434 be created. See http://inventory.sun.com/ for more information
1435 about the Service Tags asset management system.
1439 Description: Kernel BUG tries to release flock
1440 Details : Lustre does not destroy flock lock before last reference goes
1441 away. So always drop flock locks when client is evicted and
1442 perform unlock regardless of successfulness of speaking to MDS.
1446 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1450 Description: sanity test 65a fails if stripecount of -1 is set
1451 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1456 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1457 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1460 Severity : enhancement
1462 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1466 Description: do not process already freed flock
1467 Details : flock can possibly be freed by another thread before it reaches
1468 to ldlm_flock_completion_ast.
1472 Description: LBUG during stress test
1473 Details : Need properly lock accesses the flock deadlock detection list.
1477 Description: oops in page fault handler
1478 Details : kernel page fault handler can return two special 'pages' in error
1479 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1483 Description: timeout with invalidate import.
1484 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1485 should be handled by ptlrpcd. This produce long age waiting and
1486 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1488 Severity : enhancement
1490 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1494 Description: enable MGS and MDT services start separately
1495 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1496 a co-located MGS without starting the MGS, which is a complement
1497 to 'nosvc' mount option.
1501 Frequency : after recovery
1502 Description: precreate to many object's after del orphan.
1503 Details : del orphan st in oscc last_id == next_id and this triger growing
1504 count of precreated objects. Set flag LOW to skip increase count
1505 of precreated objects.
1509 Frequency : rare, on clear nid stats
1510 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1511 Details : when clean nid stats sometimes try destroy live entry,
1512 and this produce panic in free.
1516 Description: Stack overflow during MDS log replay
1517 ease stack pressure by using a thread dealing llog_process.
1521 Description: wait until IO finished before start new when do lock cancel.
1522 Details : VM protocol want old IO finished before start new, in this case
1523 need wait until PG_writeback is cleared until check dirty flag and
1524 call writepages in lock cancel callback.
1526 Severity : enhancement
1528 Description: using special macro for print time and cleanup in includes.
1532 Description: mds_mfd_close() ASSERTION(rc == 0)
1533 Details : In mds_mfd_close(), we need protect inode's writecount change
1534 within its orphan write semaphore to prevent possible races.
1538 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1542 Frequency : rare, on shutdown ost
1543 Description: don't hit live lock with umount ost.
1544 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1545 use shrink_dcache_sb instead.
1549 Description: don't panic with use echo client
1550 Details : echo client pass NULL as client nid pointer and this produce null
1551 pointer dereference.
1555 Description: fix build on ppc32
1556 Details : compile code with -m64 flag produce wrong object file for ppc32.
1560 Description: add message levels for liblustreapi
1564 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1565 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1566 will result in an -EAGAIN and in the exit call path we call
1567 alloc_rr() which will with increasing aggressiveness attempt to
1568 aquire precreated objects on the minimum number of required OSCs.
1572 Description: Use old size assignment to avoid deadlock
1573 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1574 the scheduling while holding a spinlock. We do not need locking
1575 for size in ll_update_inode() because size is only updated from
1576 the MDS for directories or files without objects, so there is no
1577 other place to do the update, and concurrent access to such inodes
1578 are protected by the inode lock.
1582 Description: resolve "_IOWR redefined" build error on SLES10
1586 Description: dump the memory debugging after all modules are unloaded to
1587 suppress false negative in conf_sanity test 39
1589 Severity : enhancement
1591 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1596 Description: fix SLES kernel versioning
1597 Details : the kernel version for our SLES 10 kernel did not include a "-"
1598 before the "smp" at the end. while this was not a problem in
1599 general, it did mean that software trying to use the kernel
1600 version to try to detect a vendor specific kernel would fail.
1601 this was most evident by the OFED build scripts.
1605 Description: Don't update lov_desc members until making sure they are valid
1606 Details : When updating lov_desc members via proc fs, need fix their
1607 validities before doing the real update.
1611 Description: don't put request into delay list while invalidate in flight.
1612 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1613 import in flight. this produce timeout for invalidate and sometimes
1614 can cause stale data.
1618 Frequency : on ppc only
1619 Description: not convert ost objects for directory because it's not exist.
1620 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1624 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1626 Severity : enhancement
1628 Description: Update to support 2.6.22.14 vanilla kernel.
1632 Frequency : rare, on recovery
1633 Description: read procfs can produce deadlock in some situation
1634 Details : Holding lprocfs lock which send rpc can produce block for destroy
1635 obd objects and this also block reconnect with -EALREADY. This
1636 isn't fix all lprocfs bugs - but make it rare.
1638 Severity : enhancement
1640 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1643 Frequency : frequent on X2 node
1645 Description: mdc_set_open_replay_data LBUG
1646 Details : Set replay data for requests that are eligible for replay.
1650 Description: lustre_mgs: operation 101 on unconnected MGS
1651 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1652 MGC, and late on MGC cannot successfully connect to MGS and a lot
1653 of the error messages complaining that MGS is not connected.
1657 Frequency : on network error
1658 Description: panic with double free request if network error
1659 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1660 but it's true only for synchronus enqueue, for async enqueue
1661 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1664 Severity : enhancement
1666 Description: client-side metadata stat-ahead during readdir(directory readahead)
1667 Details : perform client-side metadata stat-ahead when the client detects
1668 readdir and sequential stat of dir entries therein
1671 Frequency : on start mds
1673 Description: Implement get_info(last_id) in obdfilter.
1676 Frequency : occasional
1678 Description: Correctly check stale fid, not start epoch if ost not support SOM
1679 Details : open with flag O_CREATE need set old fid in op_fid3 because
1680 op_fid2 was overwritten with new generated fid, but mds can answer
1681 with one of these two fids and both is not stale. Setattr
1682 incorrectly started an epoch and assume will be called
1683 done_writing, but without SOM done_writing ever being called.
1686 Frequency : rare, depends on device drivers and load
1688 Description: MDS or OSS nodes crash due to stack overflow
1689 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1690 In some cases, in conjunction with device drivers that use a lot
1691 of stack the MDS (or possibly OSS) service threads could overflow
1692 the stack. One change which was identified to consume additional
1693 stack has been reworked to avoid the extra stack usage.
1696 Frequency : occasional
1698 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1699 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1700 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1701 oscc_has_objects and see OSC has no objests and call
1702 oscc_internal_create to resend create request.
1704 Severity : enhancement
1706 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1708 Severity : enhancement
1710 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1713 Frequency : very rare
1715 Description: Fix replay if there is an un-replied request and open
1716 Details : In some cases, older replay request will revert the
1717 mcd->mcd_last_xid on MDS which is used to record the client's
1718 latest sent request.
1720 Severity : enhancement
1722 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1724 Severity : enhancement
1726 Description: Add rhel5 support to HEAD.
1728 Serverity : enhancement
1730 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1735 Description: Don't allow skipping OSTs if index has been specified.
1736 Details : Don't allow skipping OSTs if index has been specified, make
1737 locking in internal create lots better.
1741 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1742 Details : access to bitfield in structure is always rounded to long
1743 and this produce problem with not atomic change any bit.
1747 Description: Lustre make rpms failed.
1748 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1749 builds Lustre rpms from tarball.
1752 Frequency : rare on shutdown ost
1754 Description: If llog cancel was not send before clean_exports phase, this can
1755 produce deadlock in llog code.
1756 Details : If llog thread has last reference to obd and call class_import_put
1757 this produce deadlock because llog_cleanup_commit_master wait when
1758 last llog_commit_thread exited, but this never success because was
1759 called from llog_commit_thread.
1763 Description: allow userland application know is lost one of stripes.
1764 Details : fill lvb_blocks with error code on ost and return it to
1765 application if error flag found.
1769 Description: NULL lov_tgts causing MDS oops
1770 Details : more safe checks for NULL lov_tgts for avoid oops.
1772 Severity : enhancement
1774 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1778 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1779 Details : If an OST is down the MDS will hang indefinitely in
1780 obd_statfs_rqset() waiting for the statfs data. While for
1781 MDS QOS usage of statfs, it should not stuck in waiting.
1783 Severity : enhancement
1785 Description: remote_acl support
1786 Details : Support ACL-based permission check for remote user.
1787 Support setfacl/getfacl for remote user with the utils
1788 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1789 the system "{s,g}etfacl" utils.
1791 Severity : enhancement
1793 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1795 Severity : enhancement
1797 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1801 Description: make mgs_setparam() handle fsname containing dash
1802 Details : fsname containing a dash does not work with lctl conf_param
1804 Severity : enhancement
1806 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1808 Severity : enhancement
1810 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1814 Frequency : rare, at shutdown
1815 Description: access already free / zero obd_namespace.
1816 Details : if client_disconnect_export was called without force flag set,
1817 and exist connect request in flight, this can produce access to
1818 NULL pointer (or already free pointer) when connect_interpret
1819 store ocd flags in obd_namespace.
1823 Frequency : only at startup
1824 Description: not alloc memory with spinlock held.
1825 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1826 if any spinlock held.
1828 Severity : enhancement
1830 Description: make lustre randomly fail allocating memory
1831 Details : Make lustre randomly failed allocating memory for testing purpose.
1833 Severity : enhancement
1835 Description: lost problems with lov objid file
1836 Details : Fixes some scability and access to not inited memory problems
1837 in work with lov objdid file.
1842 Description: lfs find does not continue on file error
1843 Details : Continue other files processing when a file/dir is absent.
1847 Description: Inconsistent usage of lustre_pack_reply()
1848 Details : Standardize the usage of lustre_pack_reply() such that it
1849 always generate a CERROR on failure.
1852 Frequency : occasional
1854 Description: MDS hang or stay in waiting lock
1855 Details : If client receive lock with CBPENDING flag ldlm need send lock
1856 cancel as separate rpc, to avoid situation when cancel request
1857 can't processed due all i/o threads stay in wait lock.
1861 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1865 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1869 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1873 Description: Update to RHEL4 latest kernel.
1875 Severity : enhancement
1877 Description: Build SLES10 patchless client fails
1878 Details : The configure was broken by run ./configure with
1879 --with-linux-obj=.... argument for patchless client. When the
1880 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1881 can't search header adequately. Use absolute path such as
1882 -I($LINUX)/include instead.
1886 Description: interrupt oig_wait produce painc on resend.
1887 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1888 private set, and this produce situation when rq_ptlrpcd_data not
1889 copyed to new allocated request and triggered LBUG on assert
1890 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1891 ptlrpcd set if request is changed and can be safety changed to use
1896 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1897 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1898 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1899 these macros and replace them with their difinitions instead.
1903 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1904 Details : Update which_patch & target file for SLES10 latest kernel.
1907 Description: Remove client patches from SLES 10 kernel.
1908 Details : This causes SLES 10 clients to behave as patchless clients
1909 even on a Lustre-patched (server) kernel.
1911 Severity : enhancement
1913 Description: self-adjustable client's lru lists
1914 Details : use adaptive algorithm for managing client cached locks lru
1915 lists according to current server load, other client's work
1916 pattern, memory activities, etc. Both, server and client
1917 side namespaces provide number of proc tunables for controlling
1920 Severity : enhancement
1922 Description: light-weight GSS support
1923 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1924 and reduce performance overhead.
1926 Severity : enhancement
1928 Description: Linux keyring support
1929 Details : Support using service of Linux keyring for Lustre GSS internal
1930 context refresh/cache mechanism.
1934 Description: Fix errors in lfs documentation
1935 Details : Fixes man pages
1939 Description: don't use GFP_* in generic Lustre code.
1940 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1944 Description: obdclass is limited by single OBD_ALLOC(idarray)
1945 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1949 Description: warnings with build patchless client with vanila 2.6.19 and up
1950 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1951 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1955 Frequency : Only for SLES
1956 Description: sanity_quota fail test_1
1957 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1958 file, which makes TSTID[2] inunique.
1962 Frequency : Only for SLES9
1963 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1964 Details : Update target/ChangeLog/which_patch .
1968 Frequency : Only for test_10 in sanity-quota.sh
1969 Description: a bug in quota test code
1970 Details : checking a test flag in a wrong place
1974 Frequency : Only for test_13 in sanity-quota.sh
1975 Description: a bug in quota test script
1976 Details : shouldn't check overfull allocated quota
1980 Description: jbd statistics
1981 Details : Port older jbd statistics patch for sles10
1983 Severity : enhancement
1985 Description: Add printing inode info into message about error in writepage.
1990 Description: reply_lock_interpret crash due to race with it and lock cancel.
1991 Details : Do not replay locks that are being cancelled. Do not reference
1992 locks by their address during replay, just by their handle.
1996 Frequency : When flocks are used.
1997 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1998 Details : Flock locks might destroy just granted lock if it could be merged
1999 with another existing flock, this is done in completion handler,
2000 so teach ldlm_cli_enquque_fini that this is a valid case for
2006 Description: Oops in read and write path when failing to allocate lock.
2007 Details : Check if lock allocation failed and return error back.
2011 Description: lstripe command fails for valid OST index
2012 Details : The stripe offset is compared to lov->desc.ld_tgt_count
2013 instead of lov->desc.ld_active_tgt_count.
2017 Description: sanity.sh failed test 103
2018 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
2023 Description: df doesn't work properly if diskfs blocksize != 4K
2024 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
2028 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
2029 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
2031 Severity : enhancement
2033 Description: Adaptive timeouts
2034 Details : RPC timeouts adapt to changing server load and network
2035 conditions to reduce resend attempts and improve recovery time.
2039 Description: llapi_file_create() does not allow some changes
2040 Details : add llapi_file_open() that allows specifying the mode and
2041 open flags, and also returns an open file handle.
2045 Description: merge and cleanup kernel patches.
2046 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2050 Description: Add journal checksum support.(Kernel part)
2051 Details : The journal checksum feature adds two new flags i.e
2052 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2053 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2054 indicates that the commit block contains the checksum for
2055 the blocks described by the descriptor blocks. Now commit
2056 record can be sent to disk without waiting for descriptor
2057 blocks to be written to disk. This behavior is controlled
2058 using JBD2_FEATURE_ASYNC_COMMIT flag.
2062 Description: OSS needs mutliple precreate threads
2063 Details : Add ability to start more than one create thread per OSS.
2067 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2068 Details : execute lfs setstripe on client
2072 Description: mds_obd_create error creating tmp object
2073 Details : When the user sets quota on root, llog will be affected and can't
2074 create files and write files.
2077 Frequency : Always on ia64 patchless client, and possibly others.
2079 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2080 Details : This allows the patchless client to be loaded on architectures
2081 without this export.
2085 Description: RedHat Update kernel for RHEL5
2086 Details : Add the kernel config file for RHEL5.
2090 Description: RedHat Update kernel for RHEL5
2091 Details : Modify the kernel config file more closer RHEL5.
2095 Description: Build failure against Centos5 (RHEL5)
2096 Details : Define PAGE_SIZE when it isn't present.
2100 Description: test_7 on sanity-quota.sh can't be used on separate machines
2101 Details : add facet to handle it
2105 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2106 Details : with NFS, the anon dentry's parent was set to itself in
2107 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2108 corresponding dentry other than use rec->ur_name.
2110 Severity : enhancement
2112 Description: lfs setstripe enhancement
2113 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2117 Description: enable data checksumming by default
2118 Details : enable checksum by default, allow --disable-checksum
2119 configure option and "-o nochecksum" mount option. Checksums
2120 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2124 Description: lustre support for RHEL5
2125 Details : Add support for RHEL5.
2129 Description: Client eviction due to failover config
2130 Details : after a connection loss, the lustre client should attempt to
2131 reconnect to the last active server first before trying the
2132 other potential connections.
2136 Description: when mds and osts use different quota unit(32bit and 64bit),
2137 quota will be released repeatly.
2138 Details : void sending multiple quota reqs to mds, which will keep the
2139 status between the reqs.
2143 Description: osts not allocated evenly to files
2144 Details : change the condition to increase offset_idx
2147 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2149 Description: Data corruption for OSTs that are formatted larger than 2TB
2151 Details : When generating the bio request for lustre file writes the
2152 sector number would overflow a temporary variable before being
2153 used for the IO. The data reads correctly from Lustre (which
2154 will overflow in a similar manner) but other file data or
2155 filesystem metadata may be corrupted in some cases.
2159 Description: Tune the kernel for good SCSI performance.
2160 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2161 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2166 Description: rewrite ext2-derived code in obdclass/uuid.c
2167 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2168 from scratch preserving functionality.
2172 Description: "lfs find -obd UUID" prints directories
2173 Details : "lfs find -obd UUID" will return all directory names instead
2174 of just file names. It is incorrect because the directories
2175 do not reside on the OSTs.
2179 Description: use i_size_read and i_size_write in 2.6 port
2180 Details : replace inode->i_size access with i_size_read/write()
2183 Frequency : when removing large files
2185 Description: scheduling issue during removal of large Lustre files
2186 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2187 It causes scheduling issues when removing large files (17TB in the
2191 Frequency : only with liblustre clients on XT3
2193 Description: evictions taking too long
2194 Details : allow llrd to evict clients directly on OSTs
2197 Frequency : only on ppc
2199 Description: /proc/fs/lustre/devices broken on ppc
2200 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2201 We should make sure the type of 'index' is loff_t and then cast
2202 explicitly as needed below. Do not assign an explicitly cast
2206 Frequency : only for rhel5
2208 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2209 Details : Modify the target file & which_kernel.
2211 Severity : enhancement
2213 Description: omit set fsid for export NFS
2214 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2215 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2219 Frequency : Always, for kernels after 2.6.16
2220 Description: Fix warning idr_remove called for id=.. which is not allocated.
2221 Details : Last kernels save old s_dev before kill super and not allow
2222 to restore from callback - restore it before call kill_anon_super.
2226 Description: buffer overruns could theoretically occur
2227 Details : llapi_semantic_traverse() modifies the "path" argument by
2228 appending values to the end of the origin string, and a
2229 overrun may occur. Adding buffer overrun check in liblustreapi.
2233 Description: Fix error on 'ls .' at the top of the Lustre mount.
2234 Details : Don't revalidate dentry if it is a root dentry.
2238 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2239 Details : Modify vm-tunables-rhel4.patch.
2243 Description: Kernel config for 2.6.18-vanilla.
2244 Details : Modify targets/2.6-vanilla.target.in.
2245 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2246 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2247 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2248 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2252 Description: improve handling recoverable errors
2253 Details : if request processig with error which can be recoverable on server
2254 request should be resend, otherwise page released from cache and
2259 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2260 Details : Modify target file & which_patch.
2261 A flaw was found in the IA32 system call emulation provided
2262 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2263 value could be stored in the %RAX register, which could trigger an
2264 out-of-bounds system call table access. An untrusted local user
2265 could exploit this flaw to run code in the kernel
2266 (ie a root privilege escalation). (CVE-2007-4573).
2270 Description: O_DIRECT bypasses client statistics.
2271 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2272 read_ahead_stats were updated. copied the stats section from
2273 osc_send_oap_rpc() into async_internal().
2277 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2281 Description: change order of libsysio includes
2282 Details : '#include sysio.h' should always come before '#include xtio.h'
2286 Description: handle "serious error: objid * already exists" more gracefully
2287 Details : If LAST_ID value on disk is smaller than the objects existing in
2288 the O/0/d* directories, it indicates disk corruption and causes an
2289 LBUG(). If the object is 0-length, then we should use the existing
2290 object. This will help to avoid a full fsck in most cases.
2292 Severity : enhancement
2294 Description: adapt the lustre_config script to support the upgrade case
2295 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2296 server targets to 1.6 in parallel.
2300 Description: To avoid grant space > avaible space when the disk is almost
2301 full. Without this patch you might see the error "grant XXXX >
2302 available" or some LBUG about grant, when the disk is almost
2304 Details : In filter_check_grant, for non_grant cache write, we should
2305 check the left space by if (*left > ungranted + bytes), instead
2306 of (*left > ungranted), because only we are sure the left space
2307 is enough for another "bytes", then the ungrant space should be
2308 increase. In client, we should update cl_avail_grant only there
2309 is OBD_MD_FLGRANT in the reply.
2313 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2314 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2315 described in CVE-2007-4573. This problem could allow untrusted
2316 local users to gain root access.
2319 Frequency : when using O_DIRECT and quotas
2321 Description: Incorrect file ownership on O_DIRECT output files
2322 Details : block usage reported by 'lfs quota' does not take into account
2323 files that have been written with O_DIRECT.
2328 Description: touch file failed when fs is not full
2329 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2330 otherwise we can get ENOSP while fs is not full.
2334 Description: parallel lock callbacks
2335 Details : Instead of sending blocking and completion callbacks as separated
2336 requests, adding them to a set and sending in parallel.
2339 Frequency : only for Cray XT3
2340 Bugzilla : 12829/13455
2341 Description: Changing primary group doesn't change the group lustre assigns to
2343 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2344 overridden with the primary group provided by the group upcall,
2345 whereas the supplied fsgid can be trusted if it is in the list of
2346 supplementary groups returned by the group upcall.
2348 Severity : enhancement
2350 Description: Allow masking D_WARNING, D_ERROR messages from console
2351 Details : Console messages can now be disabled via lnet.printk.
2355 Description: User code with malformed file open parameter crashes client node
2356 Details : Before packing join_file req, all the related reference should be
2357 checked carefully in case some malformed flags cause fake
2358 join_file req on client.
2362 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2363 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2364 mode is equal to the granted mode. However, it should also check
2365 that LDLM_FL_CP_REQD is not set.
2369 Description: Heavy nfs access might result in deadlocks
2370 Details : After ELC code landed, it is now improper to enqueue any mds
2371 locks under och_sem, because enqueue might want to decide to
2372 cancel open locks for same inode we are holding och_sem for.
2376 Description: Client eviction while running blogbench
2377 Details : A lot of unlink operations with concurrent I/O can lead to a
2378 deadlock causing evictions. To address the problem, the number of
2379 oustanding OST_DESTROY requests is now throttled to
2380 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2381 callbacks are processed in priority.
2385 Description: enable ACLs on MDS by default
2386 Details : ACLs must be enabled on MDS by default.
2389 Frequency : PPC/PPC64 only
2391 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2392 Details : fix duplicated definitions between asm-ppc64/types.h and
2393 lustre_types.h on PPC.
2396 Frequency : PPC/PPC64 only
2398 Description: asm-ppc/segment.h does not exist
2399 Details : fix compile issue on PPC.
2403 Description: better handle error messages in extents code
2406 Frequency : RHEL4 only
2408 Description: mkfs is very slow on IA64/RHEL4
2409 Details : A performance regression has been discovered in the MPT Fusion
2410 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2411 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2412 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2414 Severity : enhancement
2416 Description: SNMP support enhancement
2417 Details : Adding total number of sampled request for an MDS node in snmp
2420 Severity : enhancement
2422 Description: Optimize ldlm waiting list processing for PR extent locks
2423 Details : When processing waiting list for read extent lock and meeting read
2424 lock that is same or wider to it that is not contended, skip
2425 processing rest of the list and immediatelly return current
2426 status of conflictness, since we are guaranteed there are no
2427 conflicting locks in the rest of the list.
2431 Description: Time out and refuse to reconnect
2432 Details : When the failover node is the primary node, it is possible
2433 to have two identical connections in imp_conn_list. We must
2434 compare not conn's pointers but NIDs, otherwise we can defeat
2435 connection throttling.
2439 Description: port llog fixes from b1_6 into HEAD
2440 Details : Port llog reference couting and some llog cleanups from b1_6
2441 (bug 10800) into HEAD, for protect from panic and access to already
2442 free llog structures.
2446 Description: Detect stride IO mode in read-ahead
2447 Details : When a client does stride read, read-ahead should detect that and
2448 read-ahead pages according to the detected stride pattern.
2452 Description: data checksumming impacts single node performance
2453 Details : add support for several checksum algorithm. Currently, only CRC32
2454 and Adler-32 are supported. The checksum type can be changed on
2455 the fly via /proc/fs/lustre/osc/*/checksum_type.
2459 Description: use adler32 for page checksums
2460 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2465 Description: build for x2 fails
2466 Details : fix compile issue on Cray systems.
2470 Description: Properly match for duplicate locks
2471 Details : Due to different lock order from skiplists code, we need to
2472 traverse entire list for now
2475 Frequency : only on PPC/SLES10
2477 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2478 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2479 be defined. Add a hack in mkfs_lustre.c to work around this compile
2484 Description: LASSERT on MDS when client holding flock lock dies
2485 Details : ldlm pool logic depends on number of granted locks equal to
2486 number of released locks which is not true for flock locks, so
2487 just exclude such locks from consideration.
2491 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2492 Details : Use fsfilt_sync() for both the whole filesystem sync and
2493 individual file sync to eliminate dangerous inode locking
2494 with I_LOCK that can lead to a deadlock.
2498 Description: performance in 1.6.3
2499 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2500 sure that 1MB requests are not fragmented by the block layer.
2502 Severity : enhancement
2504 Description: organize the server-side client stats on per-nid basis
2505 Details : Change the structure of stats under obdfilter and mds to
2513 The "uuid"s file would list the uuids of _active_ exports.
2514 And the clear entry is to clear all stats and stale nids.
2516 Severity : enhancement
2518 Description: eliminate client locks in face of contention
2519 Details : file contention detection and lockless i/o implementation
2520 for contended files.
2524 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2527 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2529 Description: MDS may lose file striping (and hence file data) in some cases
2530 Details : If there are additional extended attributes stored on the MDS,
2531 in particular ACLs, SELinux, or user attributes (if user_xattr
2532 is specified for the client mount options) then there is a risk
2533 of attribute loss. Additionally, the Lustre file striping
2534 needs to be larger than default (e.g. striped over all OSTs),
2535 and an additional attribute must be stored initially in the
2536 inode and then increase in size enough to be moved to the
2537 external attribute block (e.g. ACL growing in size) for file
2542 Description: skiplist implementation simplification
2543 Details : skiplists are used to group compatible locks on granted list
2544 that was implemented as tracking first and last lock of each
2545 lock group the patch changes that to using doubly linked lists
2549 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2550 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2551 because LASSERT on both the data supplied by a client, and the
2552 data on disk is dangerous and incorrect.
2554 Severity : enhancement
2556 Description: Slow truncate/writes to huge files at high offsets.
2557 Details : Directly associate cached pages to lock that protect those pages,
2558 this allows us to quickly find what pages to write and remove
2559 once lock callback is received.
2563 Description: more ldlm soft lockups
2564 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2565 starve other threads from the resource lock for a long time in
2566 case of long waiting queue, so change the debug level from
2567 D_OTHER to the less frequently used D_INFO.
2569 Severity : enhancement
2571 Description: add -gid, -group, -uid, -user options to lfs find
2575 Description: Hung threads in invalidate_inode_pages2_range
2576 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2577 once one is completed. As a result, some RPCs are stuck in the
2578 queue and are never sent.
2582 Description: filter threads hungs on waiting journal commit
2583 Details : Cleanup filter group llog code, then only filter group llog will
2584 be only created in the MDS/OST syncing process.
2588 Description: Procfs and llog threads access destoryed import sometimes.
2589 Details : Sync the import destoryed process with procfs and llog threads by
2590 the import refcount and semaphore.
2592 Severity : enhancement
2594 Description: openlock cache of b1_6 port to HEAD
2599 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2600 Details : If insertion of an extent fails, then discard the inode
2601 preallocation and free data blocks else it can lead to duplicate
2606 Description: don't always update ctime in ext3_xattr_set_handle()
2607 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2608 In some cases the ctime should not be updated, for example for
2609 2.0->1.8 compatibility it is necessary to delete an xattr and it
2610 should not update the ctime.
2614 Bugzilla : 15713/16362
2615 Description: Assertion in iopen_connect_dentry in 1.6.3
2616 Details : looking up an inode via iopen with the wrong generation number can
2617 populate the dcache with a disconneced dentry while the inode
2618 number is in the process of being reallocated. This causes an
2619 assertion failure in iopen since the inode's dentry list contains
2620 both a connected and disconnected dentry.
2624 Description: assertion failure in ldlm_handle2lock()
2625 Details : fix a race between class_handle_unhash() and class_handle2object()
2626 introduced in lustre 1.6.5 by bug 13622.
2631 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2632 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2637 Description: oops in sd_iostats_seq_show()
2638 Details : unloading/reloading the scsi low level driver triggers a kernel
2639 bug when trying to access the sd iostat file.
2644 Description: Kernel panics during QLogic driver reload
2645 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2646 patch, causing memory corruption.
2651 Description: journal_dev option does not work in b1_6
2652 Details : pass mount option during pre-mount.
2654 Severity : enhancement
2656 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2657 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2658 extent information of a file. It can be used to map logical blocks
2659 in a file to physical blocks in the block device.
2663 Description: LDLM soft lockups - improvement
2664 Details : It is be possible to send the lock handle along with each read
2665 or write request because the client is already doing a lock match
2666 itself so there isn't any reason the OST should have to re-do that
2670 Frequency : only on Cray X2
2672 Description: X2 build failures
2673 Details : fix build failures on Cray X2.
2677 Description: xid & resent requests
2678 Details : Initialize RPC XID from clock at startup (randomly if clock is
2681 Severity : enhancement
2683 Description: Add lustre_start utility to start or stop multiple Lustre servers
2688 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2689 Details : In case of memory pressure, list_del() can be called twice on
2690 req->rq_history_list, causing a kernel oops.
2694 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2695 Details : fix stack overflow in the distributed lock manager by defering
2696 export eviction after a failed AST to the elt thread instead of
2697 handling it in the dlm interpret routine.
2701 Description: Convert some comments to new format.
2702 Details : Update documenting comments to match doxygen conventions.
2706 Description: Grammar fixes.
2707 Details : A couple of trivial spelling fixes.
2711 Description: OSD_COUNTERS-mandatory
2712 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2713 to simplify the code.
2717 Description: simplify cmm_device freeing logic.
2718 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2722 Description: Add lockdep support to dt_object_operations locking interface.
2723 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2724 parameter indicating lock ordering. Update mdd code to use new
2729 Description: Introduce failloc constants for lockless IO tests.
2730 Details : Add two new failloc constants to test lockless IO. Only one of
2731 them in implemented---another is checked in yet to be landed
2736 Description: Add lockdep support for inode mutex.
2737 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2738 in the situations where Lustre has to lock more than one inode
2743 Description: Add optional invariants checking support.
2744 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2745 configure switch. This macro is to be used for consistency and
2746 sanity checks that are too expensive to be left in `production'
2751 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2752 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2757 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2758 Details : Use spin_lock_nested() in (the only) situation where more than
2759 one ldlm_lock is locked simultaneously. Also, fix possible
2760 dead-lock in ldlm_lock_change_resource() by enforcing particular
2765 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2766 Details : Instead of specifying each ldlm_lock call-back through separate
2767 parameter, wrap them into struct ldlm_callback_suite.
2771 Description: Kill join_lru obd method and its callers.
2772 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2773 in memory---a requirement implemented through ->o_join_lru() obd
2774 method in HEAD. Remove this method and its users.
2778 Description: Add asynchronous ldlm ENQUEUE completion handler.
2779 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2780 case that stock ldlm_completion_ast() cannot handle as it waits
2781 until lock is granted. Introduce new ldlm_completion_ast_async()
2782 for this. Also comment ldlm_completion_ast().
2786 Description: ldlm_error <-> errno conversion.
2787 Details : Add functions to map (rather arbitrary) between LDLM error codes
2788 and standard errno values. CLIO needs this to prevent LDLM specific
2789 constants from escaping ldlm and osc.
2793 Description: Kill unused ldlm_handle2lock_ns() function.
2794 Details : Kill unused ldlm_handle2lock_ns() function.
2798 Description: Add lu_ref support to ldlm_lock
2799 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2800 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2801 and ldlm_resource. LDLM interface has to be changed, because code that
2802 releases a reference on a lock, has to "know" what reference this is.
2803 In the most frequent case
2805 lock = ldlm_handle2lock(handle);
2807 LDLM_LOCK_PUT(lock);
2809 no changes are required. When any other reference (received _not_ from
2810 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2811 instead of LDLM_LOCK_PUT().
2813 Arguably, changes are pervasive, and interface requires some discipline
2814 for proper use. On the other hand, it was very instrumental in finding
2815 a few leaked lock references.
2819 Description: Add ldlm_lock_addref_try().
2820 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2821 attempts to addref a lock that might be being canceled
2826 Description: Add ldlm_weigh_callback().
2827 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2828 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2829 hard-coded `number of pages' logic.
2833 Description: Add lockdep annotations to llog code.
2834 Details : Use appropriately tagged _nested() locking calls in the places
2835 where llog takes more than one ->lgh_lock lock.
2839 Description: Add loi_kms_set().
2840 Details : Wrap kms updates into a helper function.
2844 Description: Constify instances of struct lsm_operations.
2845 Details : Constify instances of struct lsm_operations.
2849 Description: lu_conf support.
2850 Details : On a server, a file system object is uniquely identified
2851 by a fid, which is sufficient to locate and load all object
2852 state (inode). On a client, on the other hand, more data are
2853 necessary instantiate an object. Change lu_object_find() and
2854 friends to take additional `lu_conf' argument describing object.
2855 Typically this includes layout information.
2859 Description: lu_context fixes.
2860 Details : Introduce new lu_context functions that are needed on the client
2861 side, where some system threads (ptlrpcd) are shared by multiple
2862 modules, and so cannot be stopped during module shutdown.
2866 Description: Add start and stop methods to lu_device_type_operations.
2867 Details : Introduce two new methods in lu_device_type_operations, that are
2868 invoked when first instance of a given type is created and last one
2869 is destroyed respectively. This is need by CLIO.
2873 Description: Add lu_ref support to struct lu_device.
2874 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2875 track leaked references.
2879 Description: Introduce lu_kmem_descr.
2880 Details : lu_kmem_descr and its companion interface allow to create
2881 and destroy a number of kmem caches at once.
2885 Description: Fix lu_object finalization race.
2886 Details : Fix a race between lu_object_find() finding an object and its
2887 concurrent finalization. This race is (most likely) not possible
2888 on the server, but might happen on the client.
2892 Description: Introduce lu_ref interface.
2893 Details : lu_ref is a debugging module allowing to track references to
2894 a given object. It is quite cpu expensive, and has to be
2895 explicitly enabled with --enable-lu_ref. See usage description
2900 Description: Factor lu_site procfs stats into a separate function.
2901 Details : Separate lu_site stats printing code into a separate function
2902 that can be reused on a client.
2906 Description: Constify instances of struct {lu,dt,md}_device_operations.
2907 Details : Constify instances of struct {lu,dt,md}_device_operations.
2911 Description: Introduce struct md_site and move meta-data specific parts of
2912 struct lu_site here.
2913 Details : Move md-specific fields out of struct lu_site into special struct
2914 md_site, so that lu_site can be used on a client.
2918 Description: Kill mdd_lov_destroy().
2919 Details : Remove unused mdd code.
2923 Description: Add st_block checking to multistat.c.
2924 Details : Add st_block checking to multistat.c.
2928 Description: Add lu_ref support to struct obd_device.
2929 Details : Add lu_ref tracking to obd_device.
2933 Description: Kill obd_set_fail_loc().
2934 Details : Remove unused code.
2938 Description: Add special type for ptlrpc_request interpret functions.
2939 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2940 there. Actual usage will be in CLIO.
2944 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2945 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2946 doesn't work with lockdep.
2950 Description: Add rwv.c test program.
2951 Details : New testing program exercising readv(2) and writev(2) (Qian).
2955 Description: Add sendfile.c test program.
2956 Details : New testing program exercising sendfile(2) (Jay).
2960 Description: Ratelimit a message that can be very frequent.
2961 Details : Ratelimit a memory allocation failure message that can
2966 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2967 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2971 Description: Liblustre build fixes.
2972 Details : Liblustre build fixes.
2976 Description: libcfs: add cfs_{need,cond}_resched() interface.
2977 Details : libcfs: add cfs_{need,cond}_resched() definition and
2978 implementations for Linux, NT, and liblustre.
2980 Severity : enhancement
2982 Description: More exported tunables for mballoc
2983 Details : Add support for tunable preallocation window and new tunables for
2984 large/small requests
2988 Description: Detect corruption of block bitmap and checking for preallocations
2989 Details : Checks validity of on-disk block bitmap. Also it does better
2990 checking of number of applied preallocations. When corruption is
2991 found, it turns filesystem readonly to prevent further corruptions.
2995 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2996 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2997 window to 0. If it does hit cache miss, and read-pattern is still
2998 stride-io mode, does not reset the stride window, but also does not
2999 increase the stride window length in this case.
3003 Frequency : only for big-endian servers
3004 Description: Check if system is big-endian while mounting fs with extents feature
3005 Details : Mounting a filesystem with extents feature will fail on big-endian
3006 systems since ext3-based ldiskfs is not supported on big-endian
3007 systems. This can be over-riden with "bigendian_extents" mount option.
3009 Severity : enhancement
3011 Description: The root squash functionality
3012 Details : A security feature, which is to prevent users from being able
3013 to mount lustre on their desktop, run as root, and delete
3014 all of the files in the filesystem. The goal is accomplished by
3015 remapping user id (UID) and group id (GID) of the root user to
3016 a UID and GID specified by the system administartor via Lustre
3017 configuration management server (MGS). The functionality also
3018 allows to specify sets of clients for which the remapping does
3023 Description: Excessive recovery window
3024 Details : With AT enabled, the recovery window can be excessively long (6000+
3025 seconds). To address this problem, we no longer use
3026 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
3027 timeout no longer depends on the service time, it is set to
3028 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
3029 time via pb_service_time.
3033 Description: Watchdog triggered on MDS failover
3034 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
3035 the OSTs know that the MDS "UUID" can be reused for the same export
3036 from a different NID, so we do not need to wait for the export to be
3040 Frequency : rare, only if using MMP with Linux RAID
3042 Description: MMP doesn't work with Linux RAID
3043 Details : While using HA for Lustre servers with Linux RAID, it is possible
3044 that MMP will not detect multiple mounts. To make this work we
3045 need to unplug the device queue in RAID when the MMP block is being
3046 written. Also while reading the MMP block, we should read it from
3047 disk and not the cached one.
3049 Severity : enhancement
3051 Description: open file using fid
3052 Details : A file can be opened using just its fid, like
3053 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3056 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3059 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3060 Details : add proper locking for creation of jbd_4k slab cache
3064 Description: MMP check in ext3_remount() fails without displaying any error
3065 Details : When multiple mount protection fails during remount, proper error
3068 Severity : enhancement
3070 Description: Allow stripe size to be up to 4G-64k
3071 Details : Fix math logic to allow large stripe sizes.
3075 Description: add check for >8TB ldiskfs filesystems
3076 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3077 Don't allow >8TB ldiskfs filesystems to be mounted without
3078 force_over_8tb mount option
3083 Descriptoin: OSS complains "too many blocks" and crashes
3084 Details : In filter_commitrw_write():
3085 1. lquota_chkquota();
3086 2. obdo_from_inode();
3087 3. lquota_pending_commit();
3088 "2." has changed uid and gid in this case, then caused
3089 lquota_chkquota() and lquota_pending_commit() to use
3090 different uid and gid. That is the root of the bug.
3094 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3095 Details : Detect and print out the 'stripe count = -1' case properly.
3099 Description: Out or order replies might be lost on replay
3100 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3101 request with tid smaller than one currently being added, add it
3102 to the start, not end of the list.
3104 --------------------------------------------------------------------------------
3106 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3108 * Support for kernels:
3109 2.6.5-7.283 (SLES 9),
3110 2.6.9-55.EL (RHEL 4),
3111 2.6.16.46-0.14 (SLES 10),
3112 2.6.18.8 vanilla (kernel.org)
3113 * Client support for unpatched kernels:
3114 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3115 2.6.16 - 2.6.22 vanilla (kernel.org)
3116 * Due to recently discovered recovery problems, we do not recommend
3117 using patchless RHEL 4 clients with this or any earlier release.
3118 * Recommended e2fsprogs version: 1.39.cfs8
3119 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3120 * Starting with this release, the ldiskfs backing filesystem required
3121 by Lustre is now in its own package, lustre-ldiskfs. This package
3122 should be installed. It is versioned separately from Lustre and
3123 may be released separately in future.
3127 Description: block reactivating mgc import until all deactivates complete
3128 Details : Fix race when failing back MDT/MGS to itself (testing)
3131 Frequency : at statup only
3133 Description: mds_lov_synchronize race leads to various problems
3134 Details : simultaneous MDT->OST connections at startup can cause the
3135 sync to abort, leaving the OSC in a bad state.
3137 Severity : enhancement
3139 Description: add optional extra BUILD_VERSION info
3140 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3141 to override the lustre version.
3144 Frequency : 2.6.18 servers only
3146 Description: ll_kern_mount() doesn't release the module reference
3147 Details : The ldiskfs module reference count never drops down to 0
3148 because ll_kern_mount() doesn't release the module reference.
3153 Description: server LBUG when using old ost_num_threads parameter
3154 Details : Accept the old ost_num_threads parameter but warn that it
3155 is deprecated, and fix an off-by-one error that caused an LBUG.
3160 Description: Transient SCSI error results in persistent IO issue
3161 Details : iobuf->dr_error is not reinitialized to 0 between two
3165 Frequency : sometimes when underlying device returns I/O errors
3167 Description: OSTs not going read-only during write failures
3168 Details : OSTs are not remounted read-only when the journal commit threads
3169 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3170 instead of the ext3 wrappers.
3174 Description: poor connect scaling with increasing client count
3175 Details : Don't run filter_grant_sanity_check for more than 100 exports
3176 to improve scaling for large numbers of clients.
3179 Frequency : SLES10 only
3181 Description: sanity-quota.sh quotacheck failed: rc = -22
3182 Details : Quotas cannot be enabled on SLES10.
3185 Frequency : liblustre clients only
3187 Description: getdirentries does not give error when run on compute nodes
3188 Details : getdirentries does not fail when the size specified as an argument
3189 is too small to contain at least one entry
3191 Severity : enhancement
3193 Description: Add LNET router traceability for debug purposes
3194 Details : If a checksum failure occurs with a router as part of the
3195 IO path, the NID of the last router that forwarded the bulk data
3196 is printed so it can be identified.
3201 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3202 Details : Due to a race condition, liblustre clients were occasionally
3203 evicted incorrectly.
3205 Severity : enhancement
3207 Description: lfs setstripe use optional parameters instead of postional
3210 Severity : enhancement
3212 Description: Nanosecond timestamp support for ldiskfs
3213 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3214 resolution timestamps. There is not yet support for this at
3215 the Lustre filesystem level.
3218 Frequency : during server recovery
3220 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3221 Details : request with rq_no_resend flag not awake l_wait_event if they get
3225 Frequency : nfs export on patchless client
3227 Description: connectathon hang when test nfs export over patchless client
3228 Details : Disconnected dentry cannot be found with lookup, so we do not need
3229 to unhash it or make it invalid
3232 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3233 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3234 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3239 Description: open req refcounting wrong on reconnect
3240 Details : If reconnect happened between getting open reply from server and
3241 call to mdc_set_replay_data in ll_file_open, we will schedule
3242 replay for unreferenced request that we are about to free.
3243 Subsequent close will crash in variety of ways.
3244 Check that request is still eligible for replay in
3245 mdc_set_replay_data().
3250 Description: disable writes to filesystem when reading health_check file
3251 Details : the default for reading the health_check proc file has changed
3252 to NOT do a journal transaction and write to disk, because this
3253 can cause reads of the /proc file to hang and block HA state
3254 checking on a healthy but otherwise heavily loaded system. It
3255 is possible to return to the previous behaviour during configure
3256 with --enable-health-write.
3258 Severity : enhancement
3260 Description: 64-bit inode version
3261 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3262 to the inode. This will be required for version-based recovery.
3267 Description: MDS fails to start if a duplicate client export is detected
3268 Details : in some rare cases it was possible for a client to connect to
3269 an MDS multiple times. Upon recovery the MDS would detect this
3270 and fail during startup. Handle this more gracefully.
3272 Severity : enhancement
3274 Description: Add -o localflock option to simulate old noflock
3276 Details : This will achieve local-only flock/fcntl locks
3282 Description: log_commit_thread vs filter_destroy race leads to crash
3283 Details : Take import reference before releasing llog record semaphore
3288 Description: Wrong request locking in request set processing
3289 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3290 imp_delayed_list, in this place should be used imp_lock.
3293 Frequency : when reconnection
3295 Description: Grant Leak when osc reconnect to OST
3296 Details : When osc reconnect ost, OST(filter) should check whether it
3297 should grant more space to client by comparing fed_grant and
3298 cl_avail_grant, and return the granted space to client instead
3299 of "new granted" space, because client will call osc_init_grant
3300 to update the client grant space info.
3303 Frequency : when client reconnect to OST
3305 Description: Grant Leak when osc do resend and replay bulk write
3306 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3307 bulk write request, because the grant info while be sync between
3308 OSC and OST when reconnect, and we should ignore the grant info
3309 these of resend/replay write req.
3314 Description: Grant space more than available left space sometimes.
3315 Details : When then OST is about to be full, if two bulk writing from
3316 different clients came to OST. Accord the avaliable space of the
3317 OST, the first req should be permitted, and the second one
3318 should be denied by ENOSPC. But if the seconde arrived before
3319 the first one is committed. The OST might wrongly permit second
3320 writing, which will cause grant space > available space.
3323 Frequency : when client is evicted
3325 Description: Grant might be wrongly erased when osc is evicted by OST
3326 Details : when the import is evicted by server, it will fork another
3327 thread ptlrpc_invalidate_import_thread to invalidate the
3328 import, where the grant will be set to 0. While the original
3329 thread will update the grant it got when connecting. So if
3330 the former happened latter, the grant will be wrongly errased
3331 because of this race.
3336 Description: Checking Stale with correct fid
3337 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3338 to check whether it is stale, because sometimes, we want the
3339 enqueue happened anyway, and op_data.fid2 will not be initialized.
3341 Severity : enhancement
3343 Description: update patchless client
3344 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3347 Frequency : only with 2.4 kernel
3349 Description: random memory corruption
3350 Details : size of struct ll_inode_info is to big for union inode.u and this
3351 can be cause of random memory corruption.
3356 Description: Memory leak in recovery
3357 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3358 It should also check obd_fail before fsfilt_start, otherwise if
3359 fsfilt_start return -EROFS,(failover mds during mds recovery).
3360 then the req will return with repmsg->transno = 0 and rc = EROFS.
3361 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3362 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3363 be freed no matter whether fsfilt_commit success or not.
3366 Frequency : only with huge count clients
3368 Description: Prevents from taking the superblock lock in llap_from_page for
3370 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3371 for llap_from_page call in ll_removepage prevents from taking the
3372 superblock lock for a soon died page.
3377 Description: Not check open intent error before release open handle
3378 Details : in some rare cases, the open intent error is not checked before
3379 release open handle, which may cause
3380 ASSERTION(open_req->rq_transno != 0), because it tries to release
3381 the failed open handle.
3386 Description: Set cat log bitmap only after create log success.
3387 Details : in some rare cases, the cat log bitmap is set too early. and it
3388 should be set only after create log success.
3392 Description: Accessing a block bevice can re-enable I/O when Lustre is
3393 tearing down a device.
3394 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3399 Description: service threads may hog cpus when there are a lot of requests
3400 Details : Insert cond_resched to give other threads a chance to use some CPU
3405 Description: the cat log was not initialized in recovery
3406 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3407 unlink log on mds will not be initialized until mds post
3408 recovery. And also in mds post recovery, the unlink log will
3409 initialization will be done asynchronausly, so there will be race
3410 between add unlink log and unlink log initialization.
3414 Description: brw_stats were being printed incorrectly
3415 Details : brw_stats were being printed as log2 but all of them were not
3416 recorded as log2. Also remove some code duplication arising from
3417 filter_tally_{read,write}.
3421 Frequency : rare, only in recovery.
3422 Description: ASSERTION(req->rq_type != LI_POISON) failed
3423 Details : imp_lock should be held while iterating over imp_sending_list for
3424 prevent destroy request after get timeout in ptlrpc_queue_wait.
3428 Description: replay-single.sh test 52 fails
3429 Details : A lock's skiplist need to be cleanup when it being unlinked
3430 from its resource list.
3434 Description: Short directio read returns full requested size rather than
3436 Details : Direct I/O operations should return actual amount of bytes
3437 transferred rather than requested size.
3439 Severity : enhancement
3441 Description: metadata RPC reduction (e.g. for rm performance)
3442 Details : decrease the amount of synchronous RPC between clients and servers
3443 by canceling conflicing lock before the operation on the client
3444 and packing thier handles into the main operation RPC to server.
3446 Severity : enhancement
3448 Description: Async OSC create to avoid the blocking unnecessarily.
3449 Details : If an OST has no remaining object, system will block on the
3450 creation when it needs to create a new object on this OST. Now,
3451 ways use pre-created objects when available, instead of blocking on
3452 an empty osc while others are not empty. If we must block, we
3453 block for the shortest possible period of time.
3457 Description: improve handling recoverable errors
3458 Details : if request processig with error which can be recoverable on server
3459 request should be resend, otherwise page released from cache and
3462 Severity : enhancement
3464 Description: refine locking for avoid write wrong info into lov_objid file
3465 Details : fix possible races with add new target and write/update data in
3468 --------------------------------------------------------------------------------
3470 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3475 Frequency : on some architectures
3477 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3478 Details : Uninitialized flags sometimes cause configuration commands to
3482 Frequency : patchless clients only
3484 Description: missing __iget() symbol export
3485 Details : The __iget() symbol export is missing. To avoid the need for
3486 this on patchless clients the deathrow inode reaper is turned
3487 off, and we depend on the VM to clean up old inodes. This
3488 dependency was during via the fix for bug 12181.
3490 --------------------------------------------------------------------------------
3492 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3494 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3495 INTEROPERATE with older versions automatically. In many cases a
3496 special upgrade step is needed. Please read the
3497 user documentation before upgrading any part of a live system.
3498 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3499 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3500 * WARNING: Lustre configuration and startup changes are required with
3501 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3505 Severity : enhancement
3507 Description: MountConf
3508 Details : Lustre configuration is now managed via mkfs and mount
3509 commands instead of lmc and lconf. New obd types (MGS, MGC)
3510 are added for dynamic configuration management. See
3511 https://mail.clusterfs.com/wikis/lustre/MountConf for
3514 Severity : enhancement
3516 Description: dynamic OST addition
3517 Details : OSTs can now be added to a live filesystem
3519 Severity : enhancement
3521 Description: startup order invariance
3522 Details : MDTs and OSTs can be started in any order. Clients only
3523 require the MDT to complete startup.
3525 Severity : enhancement
3527 Description: parallel, asynchronous orphan cleanup
3528 Details : orphan cleanup is now performed in separate threads for each
3529 OST, allowing parallel non-blocking operation.
3531 Severity : enhancement
3533 Description: optimized stripe assignment
3534 Details : stripe assignments are now made based on ost space available,
3535 ost previous usage, and OSS previous usage, in order to try
3536 to optimize storage space and networking resources.
3538 Severity : enhancement
3540 Description: Permanently set tunables
3541 Details : All writable /proc/fs/lustre tunables can now be permanently
3542 set on a per-server basis, at mkfs time or on a live system.
3544 Severity : enhancement
3546 Description: Lustre message v2
3547 Details : Add lustre message format v2.
3549 Severity : enhancement
3551 Description: client OST exclusion list
3552 Details : Clients can be started with a list of OSTs that should be
3553 declared "inactive" for known non-responsive OSTs.
3557 Description: SPEC SFS validation failure on NFS v2 over lustre.
3558 Details : Changes the blocksize for regular files to be 2x RPC size,
3559 and not depend on stripe size.
3561 Severity : enhancement
3563 Description: Multiple MD RPCs in flight.
3564 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3565 To avoid overly-overloading MDS, introduce a limit on number of
3566 MDS RPCs in flight for a single client and add /proc controls
3567 to adjust this limit.
3569 Severity : enhancement
3571 Description: client read/write statistics
3572 Details : Add client read/write call usage stats for performance
3573 analysis of user processes.
3574 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3575 file access. extents_stats shows chunk size distribution.
3576 extents_stats_per_process show chunk size distribution per
3579 Severity : enhancement
3581 Description: mds statistics
3582 Details : Add detailed mds operations statistics in
3583 /proc/fs/lustre/mds/*/stats.
3587 Description: Failure of copying files with lustre special EAs.
3588 Details : Client side always return success for setxattr call for lustre
3589 special xattr (currently only "trusted.lov").
3594 Description: Refcount LNET uuids
3595 Details : The global LNET uuid list grew linearly with every startup;
3596 refcount repeated list entries instead of always adding to
3599 Severity : enhancement
3601 Description: Dynamic service threads
3602 Details : Within a small range, start extra service threads
3603 automatically when the request queue builds up.
3606 Frequency : mixed-endian client/server environments
3608 Description: mixed-endian crashes
3609 Details : The new msg_v2 system had some failures in mixed-endian
3612 Severity : enhancement
3614 Description: Easy OST removal
3615 Details : OSTs can be permanently deactivated with e.g. 'lctl
3616 conf_param lustre-OST0001.osc.active=0'
3618 Severity : enhancement
3620 Description: MGS proc entries
3621 Details : Added basic proc entries for the MGS showing what filesystems
3624 Severity : enhancement
3626 Description: provide MGS failover
3627 Details : Added config lock reacquisition after MGS server failover.
3629 Severity : enhancement
3631 Description: add Linux 2.4 support
3632 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3636 Description: a large application tries to do I/O to the same resource and dies
3637 in the middle of it.
3638 Details : Check the req->rq_arrival time after the call to
3639 ost_brw_lock_get(), but before we do anything about
3640 processing it & sending the BULK transfer request. This
3641 should help move old stale pending locks off the queue as
3642 quickly as obd_timeout.
3645 Frequency : when an incorrect nid is specified during startup
3647 Description: ptlrpc connect to non-existant node causes kernel crash
3648 Details : LNET can't be re-entered from an event callback, which
3649 happened when we expire a message after the export has been
3650 cleaned up. Instead, hand the zombie cleanup off to another
3653 Severity : enhancement
3655 Description: plain/inodebits lock performance improvement
3656 Details : Grouping plain/inodebits in granted list by their request modes
3657 and bits policy, thus improving the performance of search through
3661 Frequency : only if OST filesystem is corrupted
3663 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3664 Details : for a short time RPCs with bulk IO are in the replay list,
3665 but replay of bulk IOs is unimplemented. If the OST filesystem
3666 is corrupted due to disk cache incoherency and then replay is
3667 started it is possible to trip an assertion. Avoid putting
3668 committed RPCs into the replay list at all to avoid this issue.
3671 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3673 Description: System hang on startup
3674 Details : This bug allowed the liblustre (e.g. catamount) client to
3675 return to the app before handling all startup RPCs. This
3676 could leave the node unresponsive to lustre network traffic
3677 and manifested as a server ptllnd timeout.
3679 Severity : enhancement
3681 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3682 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3683 boolean to control whether to print peer debug info when a
3684 client's RPC times out.
3687 Frequency : only for kernels with patches from Lustre below 1.4.3
3689 Description: Remove old rdonly API
3690 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3693 Frequency : only for devices with external journals
3695 Description: Set external device read-only also
3696 Details : During a commanded failover stop, we set the disk device
3697 read-only while the server shuts down. We now also set any
3698 external journal device read-only at the same time.
3701 Frequency : when upgrading from 1.4 while trying to change parameters
3703 Description: The wrong (new) MDC name was used when setting parameters for
3704 upgraded MDT's. Also allows changing of OSC (and MDC)
3705 parameters if --writeconf is specified at tunefs upgrade time.
3708 Frequency : when setting specific ost indicies
3710 Description: QOS code breaks on skipped indicies
3711 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3712 created with --index need not be sequential.
3714 Severity : enhancement
3716 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3717 Details : The uninit_groups feature works in conjunction with the kernel
3718 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3719 up the pass1 processing of e2fsck. This is a read-only feature
3720 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3721 mount filesystems that have had this feature enabled.
3723 Severity : enhancement
3725 Description: Improve multi-block allocation algorithm to avoid fragmentation
3726 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3727 allocation locality and avoid filesystem fragmentation.
3729 ------------------------------------------------------------------------------
3731 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3733 * Support for kernels:
3734 2.4.21-47.0.1.EL (RHEL 3)
3735 2.6.5-7.283 (SLES 9)
3736 2.6.9-42.0.10.EL (RHEL 4)
3737 2.6.12.6 vanilla (kernel.org)
3738 2.6.16.27-0.9 (SLES 10)
3739 * Recommended e2fsprogs version: 1.39.cfs5
3741 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3745 Frequency : occasional, depends on client load and configuration
3746 Bugzilla : 12181, 12203
3747 Description: data loss for recently-modified files
3749 Details : In some cases it is possible that recently written or created
3750 files may not be written to disk in a timely manner (this should
3751 normally be within 30s unless client IO load is very high).
3752 The problem appears as zero-length files or files that are a
3753 multiple of 1MB in size after a client crash or client eviction
3754 that are missing data at the end of the file.
3756 This problem is more likely to be hit on clients where files are
3757 repeatedly created and unlinked in the same directory, clients
3758 have a large amount of RAM, have many CPUs, the filesystem has
3759 many OSTs, the clients are rebooted frequently, and/or the files
3760 are not accessed by other nodes after being written.
3762 The presence of the problem can be detected by looking at
3763 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3764 smaller than the second (nr_unused) then dirty files will not
3765 be flushed automatically to disk. "sync; sleep 10" should be
3766 run several times on the node before unmounting it to update
3767 Lustre (this is also safe to run on nodes without this problem).
3769 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3770 that can cause this same problem, so customers using that kernel
3771 also need to update the kernel in addition to Lustre. In order
3772 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3774 It is normal that files written just before a client crash (less
3775 than 30s) may not yet have been flushed to disk, even for local
3779 Frequency : frequent on thin XT3 nodes
3781 Description: UUID collision on thin XT3 Linux nodes
3782 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3783 randomly, since we relied on an insufficiently-seeded PRNG.
3788 Description: OSS hangs after "All ost request buffers busy"
3789 Details : A deadlock between quota and journal operations caused OSS
3790 hangs after printing "All ost request buffers busy."
3793 Frequency : always on liblustre builds
3795 Description: Cleanup compiler warnings on liblustre
3798 Frequency : always on liblustre builds on XT3
3800 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3804 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3805 > 32000 subdirectories
3806 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3807 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3808 subdirectory count crosses 32000. This will aid e2fsck to
3809 correctly handle more than 32000 subdirectories.
3812 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3814 Description: System hang on startup
3815 Details : This bug allowed the liblustre (e.g. catamount) client to
3816 return to the app before handling all startup RPCs. This
3817 could leave the node unresponsive to lustre network traffic
3818 and manifested as a server ptllnd timeout.
3820 Severity : enhancement
3822 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3823 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3824 boolean to control whether to print peer debug info when a
3825 client's RPC times out.
3830 Description: make O_SYNC working on 2.6 kernels
3831 Details : 2.6 kernels use different method for mark pages for write,
3832 so need add a code to lustre for O_SYNC work.
3837 Description: Failure to close file and release space on NFS
3838 Details : Put inode details into lock acquired in ll_intent_file_open.
3839 Use mdc_intent_lock in ll_intent_open to properly
3840 detect all kind of errors unhandled by mdc_enqueue.
3845 Description: proc file read during shutdown sometimes raced obd removal,
3847 Details : Add lock to prevent obd access after proc file removal.
3850 Frequency : Only for files larger than 4GB on 32-bit clients.
3852 Description: improperly doing page alignment of locks
3853 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3854 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3859 Description: RPCs being resent when they shouldn't be
3860 Details : Some RPCs that should not be resent are being resent. This
3861 can cause inconsistencies in the RPC state machine. Do not
3862 resend such requests.
3865 Frequency : rare, only with NFS export
3867 Description: Crash on NFS re-export node
3868 Details : under very unusual load conditions an assertion is hit in
3869 ll_intent_file_open()
3872 Frequency : only if OST filesystem is corrupted
3874 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3875 Details : for a short time RPCs with bulk IO are in the replay list,
3876 but replay of bulk IOs is unimplemented. If the OST filesystem
3877 is corrupted due to disk cache incoherency and then replay is
3878 started it is possible to trip an assertion. Avoid putting
3879 committed RPCs into the replay list at all to avoid this issue.
3884 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3885 Details : Large single O_DIRECT read and write calls can fail to allocate
3886 a sufficiently large buffer to process the request. In case of
3887 allocation failure the allocation is retried with a smaller
3888 buffer and broken into smaller requests.
3890 Severity : enhancement
3892 Description: Add -o localflock option to simulate old noflock behaviour.
3893 Details : This will achieve local-only flock/fcntl locks coherentness.
3898 Description: versioning check is incomplete
3899 Details : Checking the version difference of client vs. server, report
3900 error if the gap is too big.
3905 Description: add support PG_writeback bit
3906 Details : add support for PG_writeback bit for Lustre, for more carefull
3907 work with page cache in 2.6 kernel. This also fix some deadlocks
3908 and remove hack for work O_SYNC with 2.6 kernel.
3910 Severity : enhancement
3912 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3913 Details : The uninit_groups feature works in conjunction with the kernel
3914 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3915 up the pass1 processing of e2fsck. This is a read-only feature
3916 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3917 mount filesystems that have had this feature enabled.
3919 Severity : enhancement
3921 Description: Improve multi-block allocation algorithm to avoid fragmentation
3922 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3923 allocation locality and avoid filesystem fragmentation.
3925 ------------------------------------------------------------------------------
3927 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3929 * Support for kernels:
3930 2.6.9-42.0.3.EL (RHEL 4)
3931 2.6.5-7.276 (SLES 9)
3932 2.4.21-47.0.1.EL (RHEL 3)
3933 2.6.12.6 vanilla (kernel.org)
3934 2.6.16.21-0.8 (SLES10)
3935 * Recommended e2fsprogs version: 1.39.cfs2-0
3937 * The backwards-compatible /proc/sys/portals symlink has been removed
3938 in this release. Before upgrading, please ensure that you change
3939 any configuration scripts or /etc/sysctl.conf files that access
3940 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3941 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3942 in advance of the upgrade on any system running Lustre 1.4.6 or
3943 newer, since /proc/sys/lnet was added in that version.
3944 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3948 Frequency : only when quota is used
3950 Description: avoid scanning export list for quota master
3951 Details : Change the algorithms to avoid scanning export list in order
3952 to improve the efficiency.
3955 Frequency : MDS failover only, very rarely
3957 Description: "went back in time" messages on mds failover
3958 Details : The greatest transno may be lost when the current operation
3959 finishes with an error (transno==0) and the client's last_rcvd
3960 record is over-written. Save the greatest transno in the
3961 mds_last_transno for this case.
3964 Frequency : always for specific kernels and striping counts
3966 Description: client may get "Matching packet too big" without ACL support
3967 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3968 when trying to access files in certain configurations. The
3969 clients should in fact be denied when mounting because they do
3970 not understand ACLs.
3973 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3975 Description: many clients connecting with IO in progress causes connect timeouts
3976 Details : Avoid synchronous journal commits to avoid delays caused by many
3977 clients connecting/disconnecting when bulk IO is in progress.
3978 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3979 OST_IO_PORTAL to avoid delays behind potentially many pending
3983 Frequency : occasionally with multiple writers to a single file
3985 Description: shared writes to file may result in wrong size reported by stat()
3986 Details : Allow growing of kms when extent lock is cancelled
3989 Frequency : always with random mmap IO to multi-striped file
3991 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3992 Details : Only if the hole is at the end of OST object so that kms is too
3993 small. Fix is to increase kms accordingly in ll_nopage.
3996 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3998 Description: writes to a missing object would leak memory on the OST
3999 Details : If there is an inconsistency between the MDS and OST filesystems,
4000 such that the MDS references an object that doesn't exist, writes
4001 to that object will leak memory due to incorrect cleanup in the
4002 error handling path, eventually running out of memory on the OST.
4007 Description: Creating too long symlink causes lustre errors
4008 Details : Check symlink and name lengths before sending requests to MDS.
4011 Frequency : only if flock is enabled (not on by default)
4013 Description: posix locks not released on fd closure on 2.6.9+
4014 Details : We failed to add posix locks to list of inode locks on 2.6.9+
4015 kernels, this caused such locks not to be released on fd close and
4016 then assertions on fs unmount about still used locks.
4019 Frequency : MDS failover only, very rarely
4021 Description: clients may get ASSERTION(granted_lock != NULL)
4022 Details : When request was taking a long time, and a client was resending
4023 a getattr by name lock request. The were multiple lock requests
4024 with the same client lock handle and
4025 mds_getattr_name->fixup_handle_for_resent_request found one of the
4026 lock handles but later failed with ASSERTION(granted_lock != NULL).
4031 Description: handle->h_buffer_credits > 0, assertion failure
4032 Details : h_buffer_credits is zero after truncate, causing assertion
4033 failure. This patch extends the transaction or creates a new
4037 Frequency : NFS re-export or patchless client
4038 Bugzilla : 11179, 10796
4039 Description: Crash on NFS re-export node (__d_move)
4040 Details : We do not want to hash the dentry if we don't have a lock.
4041 But if this dentry is later used in d_move, we'd hit uninitialised
4042 list head d_hash, so we just do this to init d_hash field but
4043 leave dentry unhashed.
4046 Frequency : NFS re-export or patchless client
4048 Description: NFS exports has problem with symbolic link
4049 Details : lustre client didn't properly install dentry when re-exported
4050 to NFS or running patchless client.
4053 Frequency : NFS re-export or patchless client
4055 Description: Various nfs/patchless fixes.
4056 Details : fixes reuse disconected alias for lookup process - this fixes
4057 warning "find_exported_dentry: npd != pd",
4058 fix permission error with open files at nfs.
4059 fix apply umask when do revalidate.
4062 Frequency : occasional
4064 Description: Crash on NFS re-export node
4065 Details : calling clear_page() on the wrong pointer triggered oops in
4066 generic_mapping_read().
4069 Frequency : rarely, using O_DIRECT IO
4071 Description: unaligned directio crashes client with LASSERT
4072 Details : check for unaligned buffers before trying any requests.
4075 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4077 Description: stale data returned from RAID cache
4078 Details : If only a small amount of IO is done to the RAID device before
4079 reading it again it is possible to get stale data from the RAID
4080 cache instead of reading it from disk.
4083 Frequency : always for sles10 kernel
4085 Description: sles10 support
4086 Details : ll_follow_link: compile fixes and using of nd_set_link
4087 under newer kernels.
4090 Frequency : depends on arch, kernel and compiler version, always on sles10
4093 Description: recursive or deep enough symlinks cause stack overflow
4094 Details : getting rid of large stack-allocated variable in
4098 Frequency : depends on hardware
4100 Description: lustre write performance loss in the SLES10 kernel
4101 Details : the performance loss is caused by using of write barriers in the
4102 ext3 code. The SLES10 kernel turns barrier support on by
4103 default. The fix is to undo that change for ldiskfs.
4105 ------------------------------------------------------------------------------
4107 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4109 * Support for kernels:
4110 2.6.9-42.0.3EL (RHEL 4)
4111 2.6.5-7.276 (SLES 9)
4112 2.4.21-47.0.1.EL (RHEL 3)
4113 2.6.12.6 vanilla (kernel.org)
4117 Frequency : quota enabled and large files being deleted
4119 Description: releasing more than 4GB of quota at once hangs OST
4120 Details : If a user deletes more than 4GB of files on a single OST it
4121 will cause the OST to spin in an infinite loop. Release
4122 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4127 Description: statfs data retrieved from /proc may be stale or zero
4128 Details : When reading per-device statfs data from /proc, in the
4129 {kbytes,files}_{total,free,avail} files, it may appear
4130 as zero or be out of date.
4133 Frequency : systems with MD RAID1 external journal devices
4135 Description: lconf's call to blkid is confused by RAID1 journal devices
4136 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4137 of returning all block devices that match the journal UUID.
4140 Frequency : always, for aggregate stripe size over 4GB
4142 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4143 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4144 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4147 Frequency : always if "lfs find" used on a local file/directory
4149 Description: "lfs find" segfaults if used on a local file/directory
4150 Details : The case where a directory component was not specified wasn't
4151 handled correctly. Handle this properly.
4154 Frequency : always on ppc64
4156 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4157 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4158 perform endianness conversion.
4161 Frequency : rarely (truncate to non-zero file size after write under load)
4162 Bugzilla : 10730, 10687
4163 Description: Files padded with zeros to next 4K multiple
4164 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4165 it is possible that files that are truncated to a non-zero size
4166 immediately after being written are filled with zero bytes beyond
4167 the truncated size. No file data is lost.
4169 Severity : enhancement
4171 Description: Allow recovery/failover for liblustre clients.
4172 Details : liblustre clients were unaware of failover configurations until
4175 Severity : enhancement
4177 Description: user file locks should fail when not mounting with flock option
4178 Details : Set up an error-returning stub in ll_file_operations.lock field
4179 to prevent incorrect behaviour when client is mounted without
4180 flock option. Also, set up properly f_op->flock field for
4184 Frequency : always on ia64
4186 Description: "lfs df" loops on printing out MDS statfs information
4187 Details : The obd_ioctl_data was not initialized and in some systems
4188 this caused a failure during the ioctl that did not return
4189 an error. Initialize the struct and return an error on failure.
4192 Frequency : SLES 9 only
4194 Description: Error of copying files with lustre special EAs as root
4195 Details : Client side always return success for setxattr call for lustre
4196 special xattr (currently only "trusted.lov").
4199 Frequency : rarely on clusters with both ia64+i386 clients
4201 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4202 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4203 clients are doing concurrent writes to the same file it is
4204 possible that the ia64 clients may LASSERT because the OST
4205 extent locks are not PAGE_SIZE aligned. Ensure that grown
4206 locks are always aligned on the request boundary.
4209 Frequency : specific use, occasional
4211 Description: Overwriting in use executable truncates on-disk binary image
4212 Details : If one node attempts to overwrite an executable in use by
4213 another node, we now correctly return ETXTBSY instead of
4214 truncating the file.
4216 Severity : enhancement
4218 Description: Async OSC create to avoid the blocking unnecessarily.
4219 Details : If an OST has no remaining object, system will block on the creating
4220 when need to create a new object on this OST. Now, ways use
4221 pre-created objects when available, instead of blocking on an
4222 empty osc while others are not empty. If we must block, we block
4223 for the shortest possible period of time.
4228 Description: chmod on Lustre root is propagated to other clients
4229 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4230 invalid by the follow_mount time.
4233 Frequency : liblustre clients only
4235 Description: Race in 'instant cancel' lock handling could lead to such locks
4236 never to be granted in case of SMP MDS
4237 Details : Do not destroy not yet granted but cbpending locks in
4241 Frequency : replay/resend of open
4243 Description: non null lock assetion failure in mds_intent_policy
4244 Details : Trying to replay/resend lockless open requests resulted in
4245 mds_open() returning 0 with no lock. Now it sets a flag if
4246 a lock is going to be returned.
4248 Severity : enhancement
4250 Description: Checksum enhancements
4251 Details : New checksum enhancements allow for resending RPCs that failed
4254 Severity : enhancement
4256 Description: Tunables on number of dirty pages in cacche
4257 Details : Allow to set limit on number of dirty pages cached.
4262 Description: client crash on unmount - lock still has references
4263 Details : In some error handling cases it was possible to leak a lock
4264 reference on a client while accessing a file. This was not
4265 harmful to the client during operation, but would cause the
4266 client to crash when the filesystem is unmounted.
4269 Frequency : specific case, rare
4271 Description: ETXTBSY on mds though file not in use by client
4272 Details : ETXTBSY is no longer incorrectly returned when attempting to
4273 chmod or chown a directory that the user previously tried to
4274 execute or a currently-executing binary.
4277 Frequency : extremely rare except on liblustre-based clients
4279 Description: Lustre space not freed when files are deleted
4280 Details : Clean up open-unlinked files after client eviction. Previously
4281 the unlink was skipped and the files remained as orphans.
4286 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4287 Details : In some cases it was possible to send a blocking callback to
4288 a client doing a glimpse, even though that client didn't get
4289 a lock granted. When the glimpse lock is cancelled on the OST
4290 the freed lock is left on the waiting list and corrupted the list.
4293 Frequency : all core dumps
4295 Description: Broke core dumps to lustre
4296 Details : Negative dentry may be unhashed if parent does not have UPDATE
4297 lock, but some callers, e.g. do_coredump, expect dentry to be
4298 hashed after successful create, hash it in ll_create_it.
4300 ------------------------------------------------------------------------------
4302 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4304 * Support for kernels:
4305 2.6.9-42.0.2.EL (RHEL 4)
4306 2.6.5-7.276 (SLES 9)
4307 2.4.21-40.EL (RHEL 3)
4308 2.6.12.6 vanilla (kernel.org)
4312 Frequency : always on RHEL 3
4314 Description: Number of open files grows over time
4315 Details : The number of open files grows over time, whether or not
4316 Lustre is started. This was due to a filp leak introduced
4317 by one of our kernel patches.
4319 ------------------------------------------------------------------------------
4321 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4323 * Support for kernels:
4324 2.6.9-42.EL (RHEL 4)
4325 2.6.5-7.276 (SLES 9)
4326 2.4.21-40.EL (RHEL 3)
4327 2.6.12.6 vanilla (kernel.org)
4332 Bugzilla : 5719, 9635, 9792, 9684
4333 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4334 Details : If a server is under heavy load and cannot reply to new
4335 connection requests before the client resends the (re)connect,
4336 the connection handling code can behave badly if two service
4337 threads are concurrently handing separate (re)connections from
4338 the same client. Add better locking to the connection handling
4339 code, and ensure that only a single connection will be processed
4340 for a given client UUID, even if the lock is dropped.
4342 Severity : enhancement
4344 Description: add TCP zero-copy support to kernel
4345 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4346 sends if the hardware supports scatter-gather and checksumming.
4347 This allows socklnd to do client-write and server-read more
4348 efficiently and reduce CPU utilization from skbuf copying.
4351 Frequency : only if NFS exporting from client
4353 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4354 Details : If setting attributes on a file created under NFS that had
4355 never been opened it would be possible to oops the client
4356 if the file had no objects.
4359 Frequency : always for liblustre
4361 Description: liblustre client does MDS+OSTs setattr RPC for each write
4362 Details : When doing a write from a liblustre client, the client
4363 incorrectly issued an RPC to the MDS and each OST the file was
4364 striped over in order to update the timestamps. When writing
4365 with small chunks and many clients this could overwhelm the MDS
4366 with RPCs. In all cases it would slow down the write because
4367 these RPCs are unnecessary.
4369 Severity : enhancement
4371 Description: allow number of MDS service threads to be changed at module load
4372 Details : It is now possible to change the number of MDS service threads
4373 running. Adding "options mds mds_num_threads={N}" to the MDS's
4374 /etc/modprobe.conf will set the number of threads for the next
4375 time Lustre is restarted (assuming the "mds" module is also
4376 reloaded at that time). The default number of threads will
4377 stay the same, 32 for most systems.
4382 Description: OST crash if filesystem is unformatted or corrupt
4383 Details : If an OST is started on a device that has never been formatted
4384 or if the filesystem is corrupt and cannot even mount then the
4385 error handling cleanup routines would dereference a NULL pointer.
4390 Description: NULL pointer deref in llap_from_page.
4391 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4392 incorrect) mapping if the page was truncated/reclaimed while it was
4393 searched for. Check for this condition and skip such pages when
4394 doing readahead. Introduce extra check to llap_from_page() to
4395 verify page->mapping->host is non-NULL (so page is not anonymous).
4398 Frequency : Sometimes when using sys_sendfile
4400 Description: "page not covered by a lock" warnings from ll_readpage
4401 Details : sendfile called ll_readpage without right page locks present.
4402 Now we introduced ll_file_sendfile that does necessary locking
4403 around call to generic_file_sendfile() much like we do in
4407 Frequency : with certain MDS communication failures at client mount time
4409 Description: NULL pointer deref after failed client mount
4410 Details : a client connection request may delayed by the network layer
4411 and not be sent until after the PTLRPC layer has timed out the
4412 request. If the client fails the mount immediately it will try
4413 to clean up before the network times out the request. Add a
4414 reference from the request import to the obd device and delay
4415 the cleanup until the network drops the request.
4418 Frequency : occasionally during client (re)connect
4420 Description: assertion failure during client (re)connect
4421 Details : processing a client connection request may be delayed by the
4422 client or server longer than the client connect timeout. This
4423 causes the client to resend the connection request. If the
4424 original connection request is replied in this interval, the
4425 client may trip an assertion failure in ptlrpc_connect_interpret()
4426 which thought it would be the only running connect process.
4429 Frequency : only with obd_echo servers and clients that are rebooted
4431 Description: kernel BUG accessing uninitialized data structure
4432 Details : When running an obd_echo server it did not start the ping_evictor
4433 thread, and when a client was evicted an uninitialized data
4434 structure was accessed. Start the ping_evictor in the RPC
4435 service startup instead of the OBD startup.
4437 Severity : enhancement
4438 Bugzilla : 10193 (patchless)
4439 Description: Remove dependency on various unexported kernel interfaces.
4440 Details : No longer need reparent_to_init, exit_mm, exit_files,
4441 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4444 Frequency : rare (only users of deprecated and unsupported LDAP config)
4446 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4447 Details : LDAP apparently contains 'lustreName' attributes instead of
4448 'name'. A simple remapping of the name is sufficient.
4451 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4453 Description: waiting_locks_callback trips kernel BUG if client is evicted
4454 Details : Running with the dump_on_timeout debug flag turned on makes
4455 it possible that the waiting_locks_callback() can try to dump
4456 the Lustre kernel debug logs from an interrupt handler. Defer
4457 this log dumping to the expired_lock_main() thread.
4459 Severity : enhancement
4461 Description: Support NFS exporting on 2.6 kernels.
4462 Details : Implement non-rawops metadata methods for NFS server to use without
4463 changing NFS server code.
4466 Frequency : very rare (synthetic metadata workload only)
4468 Description: two racing renames might cause an MDS thread to deadlock
4469 Details : Running the "racer" program may cause one MDS thread to rename
4470 a file from being the source of a rename to being the target of
4471 a rename at exactly the same time that another thread is doing
4472 so, and the second thread has already enqueued these locks after
4473 doing a lookup of the target and is trying to relock them in
4474 order. Ensure that we don't try to re-lock the same resource.
4477 Frequency : only very large systems with liblustre clients
4479 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4480 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4481 all OSTs in parallel. This allows the eviction of stale liblustre
4482 clients to proceed much faster than if they were done in series,
4483 and also offers similar improvements for other set_info RPCs.
4488 Description: excessive CPU usage during initial read phase on client
4489 Details : During the initial read phase on a client, it would agressively
4490 retry readahead on the file, consuming too much CPU and impacting
4491 performance (since 1.4.5.8). Improve the readahead algorithm
4492 to avoid this, and also improve some other common cases (read
4493 of small files in particular, where "small" is files smaller than
4494 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4499 Description: MDS crash when receiving packet with unknown intent.
4500 Details : Do not LBUG in unknown intent case, just return -EFAULT
4502 Severity : enhancement
4503 Bugzilla : 9293, 9385
4504 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4505 Details : Do not serialize getattr (non-intent version) and statfs.
4508 Frequency : occasional, when OST network is overloaded/intermittent
4510 Description: client evicted by OST after bulk IO timeout
4511 Details : If a client sends a bulk IO request (read or write) the OST
4512 may evict the client if it is unresposive to its data GET/PUT
4513 request. This is incorrect if the network is overloaded (takes
4514 too long to transfer the RPC data) or dropped the OST GET/PUT
4515 request. There is no need to evict the client at all, since
4516 the pinger and/or lock callbacks will handle this, and the
4517 client can restart the bulk request.
4520 Frequency : Always when mmapping file with no objects
4522 Description: client crashes when mmapping file with no objects
4523 Details : Check that we actually have objects in a file before doing any
4524 operations on objects in ll_vm_open, ll_vm_close and
4530 Description: Request leak when working with deleted CWD
4531 Details : Introduce advanced request refcount tracking for requests
4532 referenced from lustre intent.
4534 Severity : Enhancement
4536 Description: Cache open file handles on client.
4537 Details : MDS now will return special lock along with openhandle, if
4538 requested and client is allowed to hold openhandle, even if unused,
4539 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4540 closing files for every read/write openration.
4542 Severity : Enhancement
4544 Description: Cache open negative dentries on client when possible.
4545 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4546 negative dentries on lock revocation.
4551 Description: Remounting a client read-only wasn't possible with a zconf mount
4552 Details : It wasn't possible to remount a client read-only with llmount.
4554 Severity : enhancement
4555 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4556 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4557 MPICH is included to add Lustre-specific ADIO interfaces.
4558 This is based closely on the UFS ADIO layer and only differs
4559 in file creation, in order to allow the OST striping to be set.
4560 This is user-contributed code and not supported by CFS.
4565 Description: extended inode attributes (immutable, append-only) work improperly
4566 when 2.4 and 2.6 kernels are used on client/server or vice versa
4567 Details : Introduce kernel-independent values for these flags.
4569 Severity : enhancement
4572 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4573 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4574 granularity. Now, Fractional MB granularity is be supported,
4575 this is very useful for low memory system.
4577 Severity : enhancement
4579 Description: Getattr by fid
4580 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4581 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4582 have locks and avoids getattr rpc requests for it on every stat.
4585 Frequency : Always, for filesystems larger than 2TB
4587 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4588 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4589 filesystems larger than 2TB to be mounted with mballoc enabled.
4592 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4594 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4595 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4596 filesystems larger than 2TB is mounted then the kernel will
4597 silently corrupt the start of the filesystem. CONFIG_LBD is
4598 enabled for all CFS-supported kernels, but the possibility of
4599 this happening with a modified kernel config exists.
4601 Severity : enhancement
4603 Description: add client O_DIRECT support for 2.6 kernels
4604 Details : It is now possible to do O_DIRECT reads and writes to files
4605 in the Lustre client mountpoint on 2.6 kernel clients.
4607 Severity : enhancement
4609 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4610 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4611 parallel, not waiting for response from every OST before sending
4612 a rpc to the next OST.
4617 Description: setattr vs write race when updating file timestamps
4618 Details : Client processes that update a file timestamp into the past
4619 right after writing to the file (e.g. tar) it is possible that
4620 the updated file modification time can be reset to the current
4621 time due to a race between processing the setattr and write RPC.
4623 Severity : enhancement
4625 Description: Bring 'lfs find' closer in line with regular Linux find.
4626 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4627 -print0 options and obtains all the needed info through the lustre
4630 Severity : enhancement
4632 Description: support up to 1024 configured devices on one node
4633 Details : change obd_dev array from statically allocated to dynamically
4634 allocated structs as they are first used to reduce memory usage
4639 Description: Flush dirty partially truncated pages during truncate
4640 Details : Immediatelly flush partially truncated pages in filter_setattr,
4641 this way we completely avoid having any pages in page cache on OST
4642 and can retire ugly workarounds during writes to flush such pages.
4647 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4648 Details : protect inode from truncation within vfs_unlink() context
4649 just take a reference before calling vfs_unlink() and release it
4650 when parent's i_sem is free.
4655 Description: last_id value checked outside lock on OST caused LASSERT failure
4656 Details : If there were multiple MDS->OST object precreate requests in
4657 flight, it was possible that the OST's last object id was checked
4658 outside a lock and incorrectly tripped an assertion. Move checks
4659 inside locks, and discard old precreate requests.
4662 Frequency : always, if extents are used on OSTs
4664 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4665 Details : OSTs using the extents format would not zero the high 16 bits of
4666 the index physical block number. This is not a problem for any
4667 OST filesystems smaller than 16TB, and no kernels support ext3
4668 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4669 new/modified files) and can be fixed for existing filesystems
4670 with e2fsprogs-1.39-cfs1.
4675 Description: import connection selection may be incorrect if timer wraps
4676 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4677 import connections to be ignored if the 32-bit jiffies counter
4678 wraps. Use a 64-bit jiffies counter.
4681 Frequency : very large clusters immediately after boot
4683 Description: LNET request buffers exhausted under heavy short-term load
4684 Details : If a large number of client requests are generated on a service
4685 that has previously never seen so many requests it is possible
4686 that the request buffer growth cannot keep up with the spike in
4687 demand. Instead of dropping incoming requests, they are held in
4688 the LND until the RPC service can accept more requests.
4691 Frequency : Sometimes during replay
4693 Description: Assertion failure in ll_local_open after replay.
4694 Details : If replay happened on an open request reply before we were able
4695 to set replay handler, reply will become not swabbed tripping the
4696 assertion in ll_local_open. Now we set the handler right after
4697 recognising of open request
4700 Frequency : very rare
4702 Description: kernel reports "badness in vsnprintf"
4703 Details : Reading from the "recovery_status" /proc file in small chunks
4704 may cause a negative length in lprocfs_obd_rd_recovery_status()
4705 call to vsnprintf() (which is otherwise harmless). Exit early
4706 if there is no more space in the output buffer.
4708 Severity : enhancement
4710 Description: clear OBD RPC statistics by writing to them
4711 Details : It is now possible to clear the OBD RPC statistics by writing
4712 to the "stats" file.
4717 Description: Client mtime is not the same on different clients after utimes
4718 Details : In some cases, the client was using the utimes() syscall on
4719 a file cached on another node. The clients now validate the
4720 ctime from the MDS + OSTs to determine which one is right.
4725 Description: Inability to activate failout mode
4726 Details : lconf script incorrectly assumed that in pythong string's numeric
4727 value is used in comparisons.
4730 Frequency : always with multiple stripes per file
4732 Description: Inefficient object allocation for mutli-stripe files
4733 Details : When selecting which OSTs to stripe files over, for files with
4734 a stripe count that divides evenly into the number of OSTs,
4735 the MDS is always picking the same starting OST for each file.
4736 Return the OST selection heuristic to the original design.
4741 Description: mount failures may take full timeout to return an error
4742 Details : Under some heavy load conditions it is possible that a
4743 failed mount can wait for the full obd_timeout interval,
4744 possibly several minutes, before reporting an error.
4745 Instead return an error as soon as the status is known.
4747 Frequency : quota enabled and large files being deleted
4749 Description: releasing more than 4GB of quota at once hangs OST
4750 Details : If a user deletes more than 4GB of files on a single OST it
4751 will cause the OST to spin in an infinite loop. Release
4752 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4757 Description: statfs data retrieved from /proc may be stale or zero
4758 Details : When reading per-device statfs data from /proc, in the
4759 {kbytes,files}_{total,free,avail} files, it may appear
4760 as zero or be out of date.
4763 Frequency : systems with MD RAID1 external journal devices
4765 Description: lconf's call to blkid is confused by RAID1 journal devices
4766 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4767 of returning all block devices that match the journal UUID.
4770 Frequency : always, for aggregate stripe size over 4GB
4772 Description: assertion fails when trying to use 4GB stripe size
4773 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4774 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4777 Frequency : always on ppc64
4779 Description: the first write on an ext3 filesystem with mballoc got stuck
4780 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4781 perform endianness conversion.
4783 ------------------------------------------------------------------------------
4785 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4787 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4788 INTEROPERATE with older versions automatically. Please read the
4789 user documentation before upgrading any part of a live system.
4790 * WARNING: Lustre networking configuration changes are required with
4791 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4794 * Support for kernels:
4795 2.6.9-22.0.2.EL (RHEL 4)
4796 2.6.5-7.244 (SLES 9)
4797 2.6.12.6 vanilla (kernel.org)
4800 Severity : enhancement
4801 Bugzilla : 7981/8208
4802 Description: Introduced Lustre Networking (LNET)
4803 Details : LNET is new networking infrastructure for Lustre, it includes
4804 a reorganized network configuration mode (see the user
4805 documentation for full details) as well as support for routing
4806 between different network fabrics. Lustre Networking Devices
4807 (LNDs) for the supported network fabrics have also been
4808 created for this new infrastructure.
4810 Severity : enhancement
4811 Description: Introduced Access control lists
4812 Details : clients can set ACLs on files and directories in order to have
4813 more fine-grained permissions than the standard Unix UGO+RWX.
4814 The MDS must be started with the "-o acl" mount option.
4816 Severity : enhancement
4817 Description: Introduced filesystem quotas
4818 Details : Administrators may now establish per-user quotas on the
4821 Severity : enhancement
4823 Description: Configuration change for the XT3
4824 The PTLLND is now used to run Lustre over Portals on the XT3
4825 The configure option(s) --with-cray-portals are no longer used.
4826 Rather --with-portals=<path-to-portals-includes> is used to
4827 enable building on the XT3. In addition to enable XT3 specific
4828 features the option --enable-cray-xt3 must be used.
4833 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4834 Details : A race between allocating a new llcd and re-getting the llcd_lock
4835 allowed another thread to grab newly-allocated llcd.
4837 Severity : enhancement
4839 Description: 2.6 OST async journal commit and locking fix to improve performance
4840 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4841 2.6 kernels are now async as they already were in 2.4 kernels so
4842 that they can commit concurrently with the network bulk transfer.
4843 For block-allocated files the filter allocation semaphore is held
4844 to avoid filesystem fragmentation during allocation. BKL lock
4845 removed for 2.6 xattr operations where it is no longer needed.
4850 Description: lconf incorrectly determined whether two IP networks could talk
4851 Details : In some more complicated routing and multiple-network
4852 configurations, lconf will avoid trying to make a network
4853 connection to a disjoint part of the IP space. It was doing the
4854 math incorrectly for one set of cases.
4859 Description: Fix for potential infinite loop processing records in an llog.
4860 Details : If an llog record is corrupted/zeroed, it is possible to loop
4861 forever in llog_process(). Validate the llog record length
4862 and skip the remainder of the block on error.
4865 Frequency : occasional (liblustre only)
4867 Description: liblustre could not open files whose last component is a symlink
4868 Details : sysio_path_walk() would incorrectly pass the open intent to
4869 intermediate path components.
4872 Frequency : rare (liblustre only with non-standard tuning)
4873 Bugzilla : 7201 (7350)
4874 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4875 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4876 after it has been released, only do this for referenced locks
4878 Severity : enhancement
4880 Description: specifying an (invalid) directory default stripe_size of -1
4881 would reset the directory default striping
4882 Details : stripe_size -1 was used internally to signal directory stripe
4883 removal, now use "all default" to signal dir stripe removal
4884 as a directory striping of "all default" is not useful
4887 Frequency : common for large clusters running liblustre clients
4889 Description: doing an ls when liblustre clients are running is slow
4890 Details : sending a glimpse AST to a liblustre client waits for every AST
4891 to time out, as liblustre clients will not respond. Since they
4892 cannot cache data we refresh the OST lock LVB from disk instead.
4894 Severity : enhancement
4896 Description: doing an ls at the same time as file IO can be slow
4897 Details : enqueue and other "small" requests can be blocked behind many
4898 large IO requests. Create a new OST IO portal for non-IO
4899 requests so they can be processed faster.
4902 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4904 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4905 created file with mode 000
4906 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4907 allow an ftruncate() as a non-root user to a file with mode 000.
4908 Lustre now respects this flag to disable mode checks when
4909 truncating a file owned by the user
4912 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4915 Description: Revoking locks from clients that went dead or catatonic might take
4917 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4918 cancellation of such locks instant on servers without waiting for
4919 any reply from clients. Clients drops these locks when cancel
4920 notification from server is received without replying.
4923 Frequency : liblustre-only, when liblustre client dies or becomes busy
4925 Description: Doing ls on Linux clients can take a long time with active
4927 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4928 granting such locks to it in the first place if possible. Locks
4929 are taken by proxy on the OST during the read or write and
4930 dropped immediately afterward. Add connect flags handling, do
4931 not grant locks to liblustre clients for glimpse ASTs.
4933 Severity : enhancement
4935 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4936 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4937 sequential reads and behaves badly with applications doing random
4938 reads. Improve it by reading ahead at least read region, and
4939 avoiding excessive large RPC for small reads.
4941 Severity : enhancement
4943 Description: Creating more than 1000 files for a single job may cause a load
4944 imbalance on the OSTs if there are also a large number of OSTs.
4945 Details : qos_prep_create() uses an OST index reseed value that is an
4946 even multiple of the number of available OSTs so that if the
4947 reseed happens in the middle of the object allocation it will
4948 still utilize the OSTs as uniformly as possible.
4953 Description: OST or MDS may oops in ping_evictor_main()
4954 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4955 but doesn't restart at beginning of obd_exports_timed list
4958 Severity : enhancement
4960 Description: improve by-nid export eviction on the MDS and OST
4961 Details : allow multiple exports with the same NID to be evicted at one
4962 time without re-searching the exports list.
4965 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4967 Description: MDS may oops in groups_free()
4968 Details : in rare race conditions a newly allocated group_info struct is
4969 freed again, and this can be NULL. The 2.4 compatibility code
4970 for groups_free() checked for a NULL pointer, but 2.6 did not.
4973 Frequency : common for liblustre clients doing little filesystem IO
4974 Bugzilla : 9352, 7313
4975 Description: server may evict liblustre clients accessing contended locks
4976 Details : if a client is granted a lock or receives a completion AST
4977 with a blocking AST already set it would not reply to the AST
4978 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4981 Frequency : lfs setstripe, only systems with more than 160 OSTs
4983 Description: unable to set striping with a starting offset beyond OST 160
4984 Details : llapi_create_file() incorrectly limited the starting stripe
4985 index to the maximum single-file stripe count.
4988 Frequency : LDAP users only
4990 Description: lconf did not handle in-kernel recovery with LDAP properly
4991 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4993 Severity : enhancement
4995 Description: bind OST threads to NUMA nodes to improve performance
4996 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4997 node and do their allocations there to localize memory access
4999 Severity : enhancement
5001 Description: llmount can determine client NID directly from Myrinet (GM)
5002 Details : the client NID code from gmnalnid was moved directly into
5003 llmount, removing the need to use this or specifying the
5004 client NID explicitly when mounting GM clients with zeroconf
5007 Frequency : if client is started with down MDS
5009 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
5010 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
5011 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
5012 ptlrpc_import_delayed_req() because we didn't check if the
5013 request was interrupted, and we also didn't break out of the
5014 event loop if there was no timeout
5019 Description: data loss during non-page-aligned writes to a single file from
5020 both multiple nodes and multiple threads on one node at same time
5021 Details : updates to KMS and lsm weren't protected by common lock. Resulting
5022 inconsistency led to false short-reads, that were cached and later
5023 used by ->prepare_write() to fill in partially written page,
5024 leading to data loss.
5027 Frequency : always, if lconf --abort_recovery used
5029 Description: lconf --abort_recovery fails with 'Operation not supported'
5030 Details : lconf was attempting to abort recovery on the MDT device and not
5033 Severity : enhancement
5035 Description: remove cleanup logs
5036 Details : replace lconf-generated cleanup logs with lustre internal
5037 cleanup routines. Eliminates the need for client-cleanup and
5040 Severity : enhancement
5042 Description: add support for EAs (user and system) on lustre filesystems
5043 Details : it is now possible to store extended attributes in the Lustre
5044 client filesystem, and with the user_xattr mount option it
5045 is possible to allow users to store EAs on their files also
5047 Severity : enhancement
5049 Description: Add possibility (config option) to show minimal available OST free
5051 Details : When compiled with --enable-mindf configure option, statfs(2)
5052 (and so, df) will return least minimal free space available from
5053 all OSTs as amount of free space on FS, instead of summary of
5054 free spaces of all OSTs.
5056 Severity : enhancement
5058 Description: do not expand extent locks acquired on OST-side
5059 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5060 by server: they are not cached anyway.
5063 Frequency : when mmap is used/binaries executed from Lustre
5065 Description: Unmmap pages before throwing them away from read cache.
5066 Details : llap_shrink cache now attempts to unmap pages before discarding
5067 them (if unmapping failed - do not discard). SLES9 kernel has
5068 extra checks that trigger if this unmapping is not done first.
5073 Description: lconf didn't resolve symlinks before checking to see whether a
5074 given mountpoint was already in use
5077 Frequency : when migrating failover services
5078 Bugzilla : 6395, 9514
5079 Description: When migrating a subset of services from a node (e.g. failback
5080 from a failover service node) the remaining services would
5081 time out and evict clients.
5082 Details : lconf --force (implied by --failover) sets the global obd_timeout
5083 to 5 seconds in order to quickly disconnect, but this caused
5084 other RPCs to time out too quickly. Do not change the global
5085 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5087 Severity : enhancement
5088 Frequency : if MDS is started with down OST
5089 Bugzilla : 9439,5706
5090 Description: Allow startup/shutdown of an MDS without depending on the
5091 availability of the OSTs.
5092 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5093 Add appropriate locking and lov-osc refcounts for safe
5094 cleaning. Add osc abort_inflight calls in case the
5095 synchronize never started.
5098 Frequency : occasional (Cray XT3 only)
5100 Description: root not authorized to access files in CRAY_PORTALS environment
5101 Details : The client process capabilities were not honoured on the MDS in
5102 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5103 been accessed by an authorized user then root was able to access
5104 the file on the local client also. The root user capabilities
5105 are now allowed on the MDS, as this environment has secure UID.
5108 Frequency : occasional
5110 Description: ldiskfs "too long searching" message happens too often
5111 Details : A debugging message (otherwise harmless) prints too often on
5112 the OST console. This has been reduced to only happen when
5113 there are fragmentation problems on the filesystem.
5118 Description: Division by zero in statfs when all OSCs are inactive
5119 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5120 lov_statfs divides by value returned by lov_get_stripecnt().
5124 Bugzilla : 9489, 3273
5125 Description: First write from each client to each OST was only 4kB in size,
5126 to initialize client writeback cache, which caused sub-optimal
5127 RPCs and poor layout on disk for the first written file.
5128 Details : Clients now request an initial cache grant at (re)connect time
5129 and so that they can start streaming writes to the cache right
5130 away and always do full-sized RPCs if there is enough data.
5131 If the OST is rebooted the client also re-establishes its grant
5132 so that client cached writes will be honoured under the grant.
5137 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5138 Details : Now I/O RPCs go to different portal number and (presumably) fast
5139 lock requests (and glimses) and other RPCs get their own service
5140 threads pool that should be able to service those RPCs
5143 Severity : enhancement
5145 Description: Ability to exchange lustre version between client and servers and
5146 issue warnings at client side if client is too old. Also for
5147 liblustre clients there is ability to refuse connection of too old
5149 Details : New 'version' field is added to connect data structure that is
5150 filled with version info. That info is later checked by server and
5154 Frequency : rare, liblustre only.
5155 Bugzilla : 9296, 9581
5156 Description: Two simultaneous writes from liblustre at offset within same page
5157 might proceed at the same time overwriting eachother with stale
5159 Details : I/O lock withing llu_file_prwv was released too early, before data
5160 actually was hitting the wire. Extended lock-holding time until
5161 server acknowledges receiving data.
5164 Frequency : extremely rare. Never observed in practice.
5166 Description: avoid generating lustre_handle cookie of 0.
5167 Details : class_handle_hash() generates handle cookies by incrementing
5168 global counter, and can hit 0 occasionaly (this is unlikely, but
5169 not impossible, because initial value of cookie counter is
5170 selected randonly). Value of 0 is used as a sentinel meaning
5171 "unassigned handle" --- avoid it. Also coalesce two critical
5172 sections in this function into one.
5174 Severity : enhancement
5176 Description: allow liblustre clients to delegate truncate locking to OST
5177 Details : To avoid overhead of locking, liblustre client instructs OST to
5178 take extent lock in ost_punch() on client's behalf. New connection
5179 flag is added to handle backward compatibility.
5181 Severity : enhancement
5182 Bugzilla : 4928, 7341, 9758
5183 Description: allow number of OST service threads to be specified
5184 Details : a module parameter allows the number of OST service threads
5185 to be specified via "options ost ost_num_threads={N}" in the
5186 OSS's /etc/modules.conf or /etc/modprobe.conf.
5190 Bugzilla : 6146, 9635, 9895
5191 Description: servers crash with bad pointer in target_handle_connect()
5192 Details : In rare cases when a client is reconnecting it was possible that
5193 the connection request was the last reference for that export.
5194 We would temporarily drop the export reference and get a new
5195 one, but this may have been the last reference and the export
5196 was just destroyed. Get new reference before dropping old one.
5198 Severity : enhancement
5199 Frequency : if client is started with failover MDS
5201 Description: Allow multiple MDS hostnames in the mount command
5202 Details : Try to read the configuration from all specified MDS
5203 hostnames during a client mount in case the "primary"
5206 Severity : enhancement
5208 Description: Stop sending data to evicted clients as soon as possible.
5209 Details : Check if the client we are about to send or are sending data to
5210 was evicted already. (Check is done every second of waiting,
5211 for which l_wait_event interface was extended to allow checking
5212 of exit condition at specified intervals).
5215 Frequency : rare, normally only when NFS exporting is done from client
5217 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5219 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5222 Frequency : always, if srand() is called before liblustre initialization
5224 Description: Liblustre uses system PRNG disturbing its usage by user application
5225 Details : Introduce internal to lustre fast and high-quality PRNG for
5226 lustre usage and make liblustre and some other places in generic
5227 lustre code to use it.
5229 Severity : enhancement
5230 Bugzilla : 9477, 9557, 9870
5231 Description: Verify that the MDS configuration logs are updated when xml is
5232 Details : Check if the .xml configuration logs are newer than the config
5233 logs stored on the MDS and report an error if this is the case.
5234 Request --write-conf, or allow starting with --old_conf.
5236 Severity : enhancement
5238 Description: Handle symlinks in the path when checking if Lustre is mounted.
5239 Details : Resolve intermediate symlinks when checking if a client has
5240 mounted a filesystem to avoid duplicate client mounts.
5245 Description: lconf can hit an error exception but still return success.
5246 Details : The lconf command catches the Command error exception at the top
5247 level script context and will exit with the associated exit
5248 status, but doesn't ensure that this exit status is non-zero.
5253 Description: failure of ptlrpc thread startup can cause oops
5254 Details : Starting a ptlrpc service thread can fail if there are a large
5255 number of threads or the server memory is very fragmented.
5256 Handle this without oopsing.
5259 Frequency : always, only if liblustre and non-default acceptor port was used
5261 Description: liblustre cannot connect to servers with non-default acceptor port
5262 Details : tcpnal_set_default_params() was not called and was therefore
5263 ignoring the environment varaible TCPNAL_PORT, as well as other
5264 TCPNAL_ environment variables
5269 Description: two objects could be created on the same OST for a single file
5270 Details : If an OST is down, in some cases it was possible to create two
5271 objects on a single OST for a single file. No problems other
5272 than potential performance impact and spurious error messages.
5276 Bugzilla : 5681, 9562
5277 Description: Client may oops in ll_unhash_aliases
5278 Details : Client dcache may become inconsistent in race condition.
5279 In some cases "getcwd" can fail if the current directory is
5285 Description: Inode refcounting problems in NFS export code
5286 Details : link_raw functions used to call d_instantiate without obtaining
5287 extra inode reference first.
5291 Bugzilla : 9942, 9903
5292 Description: Referencing freed requests leading to crash, memleaks with NFS.
5293 Details : We used to require that call to ll_revalidate_it was always
5294 followed by ll_lookup_it. Also with revalidate_special() it is
5295 possible to call ll_revalidate_it() twice for the same dentry
5296 even if first occurence returned success. This fix changes semantic
5297 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5298 reference on a request referred from the intent.
5299 ll_intent_release() then releases such a request.
5302 Frequency : rare, normally benchmark loads only
5304 Description: unlinked inodes were kept in memory on the client
5305 Details : If a client is repeatedly creating and unlinking files it
5306 can accumulate a lot of stale inodes in the inode slab cache.
5307 If there is no other client load running this can cause the
5308 client node to run out of memory. Instead flush old inodes
5309 from client cache that have the same inode number as a new inode.
5312 Frequency : SLES9 2.6.5 kernel and long filenames only
5313 Bugzilla : 9969, 10379
5314 Description: utime reports stale NFS file handle
5315 Details : SLES9 uses out-of-dentry names in some cases, which confused
5316 the lustre dentry revalidation. Change it to always use the
5320 Frequency : rare, unless heavy write-truncate concurrency is continuous
5321 Bugzilla : 4180, 6984, 7171, 9963, 9331
5322 Description: OST becomes very slow and/or deadlocked during object unlink
5323 Details : filter_destroy() was holding onto the parent directory lock
5324 while truncating+unlinking objects. For very large objects this
5325 may block other threads for a long time and slow overall OST
5326 responsiveness. It may also be possible to get a lock ordering
5327 deadlock in this case, or run out of journal credits because of
5328 the combined truncate+unlink. Solution is to do object truncate
5329 first in one transaction without parent lock, and then do the
5330 final unlink in a new transaction with the parent lock. This
5331 reduces the lock hold time dramatically.
5334 Frequency : rare, 2.4 kernels only
5336 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5337 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5338 before unmounting their filesystem in order to avoid deadlock.
5339 The kernel_locked() function in 2.4 kernels only checks whether
5340 the kernel lock is held, not whether it is this process that is
5341 holding it as 2.6 kernels do.
5346 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5347 Details : The client ptlrpc code may be trying to reconnect to a down
5348 server before a previous connection attempt has timed out.
5349 Increase the reconnect interval to be longer than the connection
5350 timeout interval to avoid sending duplicate connections to
5354 Frequency : echo_client brw_test command
5356 Description: fix echo_client to work with OST preallocated code
5357 Details : OST preallocation code (5137) didn't take echo_client IO path
5358 into account: echo_client calls filter methods outside of any
5359 OST thread and, hence, there is no per-thread preallocated
5360 pages and buffers to use. Solution: hijack pga pages for IO. As
5361 a byproduct, this avoids unnecessary data copying.
5365 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5366 Description: Client can oops in mdc_commit_close() after open replay
5367 Details : It was possible for the MDS to return an open request with no
5368 transaction number in mds_finish_transno() if the client was
5369 evicted, but without actually returning an error. Clients
5370 would later try to replay that open and may trip an assertion
5371 Simplify the client close codepath, and always return an error
5372 from the MDS in case the open is not successful.
5375 Frequency : rare, 2.6 OSTs only
5377 Description: OST may deadlock under high load on fragmented files
5378 Details : If there was a heavy load and highly-fragmented OST filesystems
5379 it was possible to have all the OST threads deadlock waiting on
5380 allocation of biovecs, because the biovecs were not released
5381 until the entire RPC IO was completed. Instead, release biovecs
5382 as soon as they are complete to ensure forward IO progress.
5384 Severity : enhancement
5386 Description: Support for specifying external journal device at mount
5387 Details : If an OST or MDS device is formatted with an external journal
5388 device, this device major/minor is stored in the ext3 superblock
5389 and may not be valid for failover. Allow detecting and
5390 specifying the external journal at mount time.
5395 Description: Mounting an MDS with pending unlinked files may cause oops
5396 Details : target_finish_recovery() calls mds_postrecov() which returned
5397 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5398 considers this an error and immediately begins cleaning up the
5399 lov, just after starting the mds_lov process
5401 Severity : enhancement
5403 Description: Implement 'lfs df' to report actual free space on per-OST basis
5404 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5405 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5406 usage of objects; '-h' to report in human readable format.
5408 ------------------------------------------------------------------------------
5410 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5417 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5418 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5419 at startup during orphan unlinking. Instead, skip buddy update
5420 before setup, as it will be regenerated after recovery is complete.
5425 Description: If an OST is inactive, its locks might reference stale inodes.
5426 Details : lov_change_cbdata() must iterate over all namespaces, even if
5427 they are inactive to clear inode references from the lock.
5429 Severity : enhancement
5430 Frequency : occasional, if non-standard max_dirty_mb used
5432 Description: Client will block write RPCs if not enough grant
5433 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5434 then the client will block writes while waiting for another RPC
5435 to complete instead of consuming its dirty limit. With change
5436 we get improved performance when max_dirty_mb is small.
5438 Severity : enhancement
5439 Bugzilla : 3389, 6253
5440 Description: Add support for supplementary groups on the MDS.
5441 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5442 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5443 lookups for user supplementary groups into a cache.
5447 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5448 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5449 for liblustre O_EXCL is enforced by MDS.
5453 Description: readdir never returns NULL in liblustre.
5454 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5455 dentry in struct dirent.
5459 Frequency : liblustre only, depends on application IO pattern
5460 Description: liblustre clients evicted if not contacting servers
5461 Details : Don't put liblustre clients into the ping_evictor list, so
5462 they will not be evicted by the pinger ever.
5464 Severity : enhancement
5466 Description: Add ability to evict clients by NID from MDS.
5467 Details : By echoing "nid:$NID" string into
5468 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5469 $NID would be instantly evicted from this MDS and from all active
5470 OSTs connected to it.
5474 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5475 Details : lookup_it_finish() used to query file size from OSTs that was not
5480 Description: service threads change working directory to that of init
5481 Details : Starting lustre service threads may pin the working directory
5482 of the parent thread, making that filesystem busy. Threads
5483 now change to the working directory of init to avoid this.
5487 Frequency : during shutdown only
5488 Description: shutdown with a failed MDS or OST can cause unmount to hang
5489 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5494 Frequency : 2.6 only
5495 Description: chmod/chown may include an extra supplementary group
5496 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5497 supplementary group and if none is specified this is used.
5500 Bugzilla : 5479 (6816)
5502 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5503 Details : If lookup is in progress on a file that is unlinked we might try
5504 to revalidate the inode and fail in revalidate after lookup is
5505 complete and ll_file_open() enqueues the open again but
5506 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5509 Frequency : always, if lconf --abort_recovery used
5511 Description: lconf --abort_recovery fails with 'Operation not supported'
5512 Details : lconf was attempting to abort recovery on the MDT device and not
5515 ------------------------------------------------------------------------------
5517 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5522 Frequency : rare (only unsupported configurations with a node running as an
5524 Bugzilla : 6514, 5137
5525 Description: Mounting a Lustre file system on a node running as an OST could
5527 Details : OSTs now preallocates memory needed to write out data at
5528 startup, instead of when needed, to avoid having to
5529 allocate memory in possibly low memory situations.
5530 Specifically, if the file system is mounted on on OST,
5531 memory pressure could force it to try to write out data,
5532 which it needed to allocate memory to do. Due to the low
5533 memory, it would be unable to do so and the node would
5534 become unresponsive.
5536 Severity : enhancement
5538 Description: Addition of lconf --service command line option
5539 Details : lconf now accepts a '--service <arg>' option, which is
5540 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5542 Severity : enhancement
5544 Description: Failover mode is now the default for OSTs.
5545 Details : By default, OSTs will now run in failover mode. To return to
5546 the old behaviour, add '--failout' to the lmc line for OSTs.
5548 Severity : enhancement
5550 Description: Health checks are now provided for MDS and OSTs
5551 Details : Additional detailed health check information on MSD and OSTs
5552 is now provided through the procfs health_check value.
5555 Frequency : occasional, depends on IO load
5557 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5558 after numerous create/delete cycles
5559 Details : The ext3 inode allocation policy would not allocate new inodes
5560 very well on the OSTs because there are no new directories
5561 being created. Instead we look for groups with free space if
5562 the parent directories are nearly full.
5567 Description: Network or server problems during mount may cause partially
5568 mounted clients instead of returning an error.
5569 Details : The config llog parsing code may overwrite the error return
5570 code during mount error handling, returning success instead
5576 Description: MDS can fail to allocate large reply buffers
5577 Details : After long uptimes the MDS can fail to allocate large reply
5578 buffers (e.g. zconf client mount config records) due to memory
5579 fragmentation or consumption by the buffer cache. Preallocate
5580 some large reply buffers so that these replies can be sent even
5581 under memory pressure.
5585 Frequency : rare (liblustre)
5586 Description: fsx running with liblustre complained that using truncate() to
5587 extend the file doesn't work. This patch corrects that issue.
5588 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5589 ATTR_SIZE and lsm use in llu_setattr_raw.
5593 Frequency : rare, only 2.6 kernels
5594 Description: Unusual file access patterns on the MDS may result in inode
5595 data being lost in very rare circumstances.
5596 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5597 code in the 2.6 kernel caused the inode and/or EA data not to
5598 be read from disk, causing single-file corruption.
5602 Frequency : rare, only 2.6 filesystems using extents
5603 Description: Heavy concurrent write and delete load may cause data corruption.
5604 Details : It was possible under high-load situations to have an extent
5605 metadata block in the block device cache from a just-unlinked
5606 file overwrite a newly-allocated data block. We now unmap any
5607 metadata buffers that alias just-allocated data blocks.
5611 Frequency : filesystems with default stripe_count larger than 77
5612 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5613 Details : lconf specifies an inode size of 4096 bytes when the default
5614 stripe_count is larger than 77. This conflicts with the default
5615 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5616 this case to avoid pinning too much memory for large EAs.
5618 ------------------------------------------------------------------------------
5620 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5625 Frequency : rare (extremely heavy IO load with hundreds of clients)
5627 Description: Client is evicted, gets IO error writing to file
5628 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5629 caused clients to be evicted for AST timeouts. The fixes in
5630 bug 5192 mean we no longer need to have such short AST timeouts
5631 so ldlm_timeout has been increased.
5634 Frequency : occasional during --force or --failover shutdown under load
5635 Bugzilla : 5949, 4834
5636 Description: Server oops/LBUG if stopped with --force or --failover under load
5637 Details : a collection of import/export refcount and cleanup ordering
5638 issues fixed for safer force cleanup
5641 Frequency : only filesystems larger than 120 OSTs
5642 Bugzilla : 5990, 6223
5643 Description: lfs getstripe would oops on a very large filesystem
5644 Details : lov_getconfig used kfree on vmalloc'd memory
5647 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5648 Bugzilla : 6242, 6243
5649 Description: reading from files that had been truncated to a non-zero size
5650 but never opened returned no data
5651 Details : ll_file_read() reads zeros from no-object files to EOF
5656 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5657 Details : There's little harm in aborting MDS/OSS recovery and letting it
5658 try again, so I removed the LASSERT and return an error instead.
5660 Severity : enhancement
5662 Description: New debugging infrastructure for tracking down data corruption
5663 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5664 (b) cover more of the client-side code path, and (c) try to narrow
5665 down where problems occurred
5669 Bugzilla : 3819, 4364, 4397, 6313
5670 Description: Racing close and eviction MDS could cause assertion in mds_close
5671 Details : It was possible to get multiple mfd references during close and
5672 client eviction, leading to one thread referencing a freed mfd.
5674 Severity: : enhancement
5675 Bugzilla : 3262, 6359
5676 Description: Attempts to reconnect to servers are now more aggressive.
5677 Details : This builds on the enhanced upcall-less recovery that was added
5678 in 1.4.2. When trying to reconnect to servers, clients will
5679 now try each server in the failover group every 10 seconds. By
5680 default, clients would previously try one server every 25 seconds.
5685 Description: After recovery, certain operations trigger a failed
5686 assertion on a client.
5687 Details : Failing over an mds, using lconf -d --failover, while a
5688 client was doing a readdir() call would cause the client to
5689 LBUG after recovery completed and the readdir() was resent.
5691 Severity : enhancement
5693 Description: Default groups are now added by lconf
5694 Details : You can now run lconf --group <servicename> without having to
5695 manually add groups with lmc.
5698 Frequency : occasional
5700 Description: Nodes with an elan id of 0 trigger a failed assertion
5703 Frequency : always when accessing e.g. tty/console device nodes
5705 Description: tty and some other devices nodes cannot be used on lustre
5706 Details : file's private_data field is used by device data and lustre
5707 values in there got lost. New field was added to struct file to
5708 store fs-specific private data.
5711 Frequency : when exporting Lustre via NFS
5713 Description: NFSD failed occasionally when looking up a path component
5714 Details : NFSD is looking up ".." which was broken in ext3 directories
5715 that had grown large enough to become hashed.
5718 Frequency : Clusters with multiple interfaces not on the same subnet
5720 Description: Nodes will repeatedly try to reconnect to an interface which it
5721 cannot reach and report an error to the log.
5722 Details : Extra peer list entries will be created by lconf with some peers
5723 unreachable. lconf now validates the peer before adding it.
5726 Frequency : Only if a default stripe is set on the filesystem root.
5728 Description: Setting a default stripe on the filesystem root prevented the
5729 filesystem from being remounted.
5730 Details : The client was sending extra request flags in the root getattr
5731 request and did not allocate a reply buffer for the dir EA.
5734 Frequency : occasional, higher if lots of files are accessed by one client
5735 Bugzilla : 6159, 6097
5736 Description: Client trips assertion regarding lsm mismatch/magic
5737 Details : While revalidating inodes the VFS looks up inodes with ifind()
5738 and in rare cases can find an inode that is being freed.
5739 The ll_test_inode() code will free the lsm during ifind()
5740 when it finds an existing inode and then the VFS later attaches
5741 this free lsm to a new inode.
5745 Bugzilla : 6422, 7030
5746 Description: MDS deadlock between mkdir and client eviction
5747 Details : Creating a new file via mkdir or mknod (starting a transaction
5748 and getting the ns lock) can deadlock with client eviction
5749 (gets ns lock and trying to finish a synchronous transaction).
5752 Frequency : occasional
5753 Description: While starting a server, the fsfilt_ext3 module could not be
5755 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5756 kernels. Previously, lconf would still use the ext3 name
5757 when trying to load modules. Now, it will correctly use
5758 ext3 on 2.4 and ldiskfs on 2.6.
5760 Severity : enhancement
5761 Description: The default stripe count has been changed to 1
5762 Details : The interpretation of the default stripe count (0, to lfs
5763 or lmc) has been changed to mean striping across a single
5764 OST, rather than all available. For general usage we have
5765 found a stripe count of 1 or 2 works best.
5767 Severity : enhancement
5768 Description: Add support for compiling against Cray portals.
5769 Details : Conditional compiling for some areas that are different
5773 Frequency : occasional
5774 Bugzilla : 6409, 6834
5775 Description: Creating files with an explicit stripe count may lead to
5776 a failed assertion on the MDS
5777 Details : If some OSTs are full or unavailable, creating files may
5778 trigger a failed assertion on the MDS. Now, Lustre will
5779 try to use other servers or return an error to the
5783 Frequency : occasional
5785 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5786 caused lock timeout during readahead (since 1.4.2).
5787 Details : Processes doing readahead might match a lock that hasn't been
5788 granted yet if there are overlapping and conflicting lock
5789 requests. The readahead process waits on ungranted lock
5790 (original lock is CBPENDING), while OST waits for that process
5791 to cancel CBPENDING read lock and eventually evicts client.
5793 Severity : enhancement
5795 Description: Initial enabling of flock support for clients
5796 Details : Implements fcntl advisory locking and file status functions.
5797 This feature is provided as an optional mount flag (default
5798 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5799 locking are implemented yet, and those that are are not guaranteed
5800 to be completely correct in production environments.
5801 mount -t lustre -o [flock|noflock] ...
5804 Frequency : occasional
5806 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5807 assertion in the ext3 JBD (journaling) layer.
5808 Details : The b_committed_data struct is protected by the big kernel lock
5809 in 2.4 kernels, serializing journal_commit_transaction() and
5810 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5811 access to this struct. In 2.6 kernels there is finer grained
5812 locking to improve SMP performance of the JBD layer.
5816 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5821 Description: merge and cleanup kernel patches.
5824 -----------------------------------------------------------------------------
5826 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5828 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5829 versions of Lustre. Please update all servers and clients to
5830 version 1.4.2 or later at the same time. You must also run
5831 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5832 to update the configuration logs.
5834 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5835 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5836 - swab llog records of type '0' so we get proper header size/idx (5861)
5837 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5838 - fix rename of one directory over another leaking an inode (5953)
5839 - avoid SetPageDirty on 2.6 (5981)
5840 - don't re-add just-being-destroyed locks to the waiting list (5653)
5841 - when creating new directories, inherit the parent's custom
5842 striping settings if present parent (3048)
5843 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5844 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5845 - don't hold client locks on temporary worklist from l_lru (5666)
5846 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5847 - automatically evict dead clients (5921)
5848 - Update file size properly in create+truncate+fstat case (6196)
5849 - Do not unhash mountpoint dentries, do not allow removal of
5851 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5852 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5853 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5854 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5855 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5856 - Correctly handle reads of files with no objects (6243)
5857 - lctl recover will also mark a device active if deactivate used (5933)
5859 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5860 - allow --write-conf on an MDS with different nettype than client (5619)
5861 - don't write config llogs to MDS for mounts not from that MDS (5617)
5862 - lconf should create multiple TCP connections from a client (5201)
5863 - init scripts are now turned off by default; run chkconfig --on
5864 lustre and chkconfig --on lustrefs to use them
5865 - upcalls are no longer needed for clients to recover to failover
5867 - add --abort-recovery option to lconf to abort recovery on device
5869 - add support for an arbitrary number of OSTs (3026)
5870 - Quota support protocol changes.
5871 - forward compatibility changes to wire structs (6007)
5872 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5873 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5874 - improved "lustre status" script
5875 - initialize blocksize for non-regular files (6062)
5876 - added --disable-server and --disable-client configure options (5782)
5877 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5878 - Vanilla 2.4.29 support
5879 - increase maximum number of obd devices to 520 (6242)
5880 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5881 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5882 SLES 9 2.6.5 kernels
5883 - limit stripes per file to 160 (the maximum EA size) (6093)
5885 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5888 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5889 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5890 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5891 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5892 - small ext3 extents cleanups and fixes (5733)
5893 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5894 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5895 - avoid needless client->OST connect, fix handle mismatch (5317)
5896 - fix DLM error path that led to out-of-sync client, long delays (5779)
5897 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5898 - fix several locking issues related to i_size (5492,5624,5654,5672)
5899 - don't move pending lock onto export if it is already evicted (5683)
5900 - fix kernel oops when creating .foo in unlinked directory (5548)
5901 - fix deadlock in obdfilter statistics vs. object create (5811)
5902 - use time_{before,after} to avoid timer jiffies wrap (5882)
5903 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5904 - Do not leak request if server was not able to process it (5154)
5905 - If mds_open unable to find parent dir, make that negative lookup(5154)
5906 - don't create new directories with extent-mapping (5909, 5936)
5908 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5909 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5910 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5911 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5912 - support for RHEL4 kernel on i686 (5773)
5913 - provide error messages when incompatible logs are encountered (5898)
5915 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5916 * version 1.4.0.10 (1.4.1 release candidate 1)
5918 - don't keep a lock reference when lock is not granted (4238)
5919 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5920 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5921 - drop import inflight refcount on signal_completed_replay error (5255)
5922 - unlock page after async write error during send (3677)
5923 - handle missing objects in filter_preprw_read properly (5265)
5924 - no transno return for symlink open, don't save no-trasno open (3440)
5925 - don't try to complete elan receive that already failed (4012)
5926 - free RPC server reply state on error (5406)
5927 - clean up thread from ptlrpc_start_thread() on error (5160)
5928 - readahead could read extra page into cache that wasn't ejected (5388)
5929 - prevent races in class_attach/setup/cleanup/detach (5260)
5930 - don't dereference de->d_inode after l_dput of de (5458)
5931 - use "int" for stripe value returned from lock_to_stripe (5544)
5932 - mballoc allocation and error-checking fixes in 2.6 (5504)
5933 - block device patches to fix I/O request sizes in 2.6 (5482)
5934 - look up hostnames for IB nals (5602)
5935 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5936 - don't start multiple acceptors for the same port (5277)
5937 - fix incorrect LASSERT in mds_getattr_name (5635)
5938 - export a proc file for general "ping" checking (5628)
5939 - fix "lfs check" to not block when the MDS is down (5628)
5941 - service request history (4965)
5942 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5943 - create an "evict_client" /proc entry on OSTs, like the MDS has
5944 - fix mount usage message, return errors per mount(8) (5168)
5945 - change grep [] to grep "[]" in tests so they work in more UMLs
5946 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5947 - remove ancient LOV_MAGIC_V0 EA support (5047)
5948 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5949 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5950 - allow readahead allocations to fail when low on memory (5383)
5951 - mmap locking landed again, after considerable improvement (2828)
5952 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5954 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5957 - send OST transaction number in read/write reply to free req (4966)
5958 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5959 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5960 - the watchdog thread now runs as interruptible (5246)
5961 - flock/lockf fixes (but it's still disabled, pending 5135)
5962 - don't use EXT3 constants in llite code (5094)
5963 - memory shortage at startup could cause assertion (5176)
5965 - reorganization of lov code
5966 - single portals codebase
5968 - add extents/mballoc support (5025)
5969 - direct I/O reads in the obdfilter (4048)
5970 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5972 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5975 - send OST transaction number in read/write reply to free req (4966)
5976 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5977 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5978 - the watchdog thread now runs as interruptible (5246)
5979 - handle missing objects in filter_preprw_read properly (5265)
5980 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5981 - drop import inflight refcount on signal_completed_replay error (5255)
5982 - unlock page after async write error during send (3677)
5983 - return original error code on reconstructed replies (3761)
5984 - no transno return for symlink open, don't save no-trasno open (3440)
5986 - add pid to ldlm debugging output (4922)
5987 - bump the watchdog timeouts -- we can't handle 30sec yet
5988 - extra debugging for orphan dentry/inode bug (5259)
5990 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5993 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5994 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5995 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5996 - fix NULL dereference in /proc/sys/portals/routes (4827)
5997 - allow failed mdc_close() operations to be interrupted (4561)
5998 - stop precreate on OST before MDS would time out on it (4778)
5999 - don't send partial-page writes before EOF from client (4410)
6000 - discard client grant for sub-page writes on large-page clients (4520)
6001 - don't free dentries not owned by NFS code, check generation (4806)
6002 - fix lsm leak if mds_create_objects() fails (4801)
6003 - limit debug_daemon file size, always print CERROR messages (4789)
6004 - use transno after validating reply (3892)
6005 - process timed out requests if import state changes (3754)
6006 - update mtime on OST during writes, return in glimpse (4829)
6007 - add mkfsoptions to LDAP (4679)
6008 - use ->max_readahead method instead of zapping global ra (5039)
6009 - don't interrupt __l_wait_event() during strace
6011 - add software watchdogs to catch hung threads quickly (4941)
6012 - make lustrefs init script start after nfs is mounted
6013 - fix CWARN/ERROR duplication (4930)
6014 - return async write errors to application if possible (2248)
6015 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
6016 - print NAL number in %x format (4645)
6017 - update barely-supported suse-2.4.21-171 series (4842)
6018 - support for sles 9 %post scripts
6019 - support for building 2.6 kernel-source packages
6020 - support for sles km_* packages
6022 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
6025 - ignore -ENOENT errors in osc_destroy (3639)
6026 - notify osc create thread that OSC is being cleaned up (4600)
6027 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
6028 - reconstruct ost_handle() like mds_handle() (4657)
6029 - create a new thread to do import eviction to avoid deadlock (3969)
6030 - let lconf resolve symlinked-to devices (4629)
6031 - don't unlink "objects" from directory with default EA (4554)
6032 - hold socknal file ref over connect in case target is down (4394)
6033 - allow more than 32000 subdirectories in a single directory (3244)
6034 - fix blocks count for O_DIRECT writes (3751)
6035 - OST returns ENOSPC from object create when no space left (4539)
6036 - don't send truncate RPC if file size isn't changing (4410)
6037 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
6038 - bind to privileged port in socknal and tcpnal (3689)
6040 - rate limit CERROR/CWARN console message to avoid overload (4519)
6041 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
6042 - basic mmap support (3918)
6043 - kernel patch series update from b1_4 (4711)
6045 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6048 - avoid crash during MDS cleanup with OST shut down (2775)
6049 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6050 - don't use bad inodes on the MDS (3744)
6051 - dynamic object preallocation to improve recovery speed (4236)
6052 - don't hold spinlock over lock dumping or change debug flags (4401)
6053 - don't zero obd_dev when it is force cleaned (3651)
6054 - print grants to console if they go negative (4431)
6055 - "lctl deactivate" will stop automatic recovery attempts (3406)
6056 - look for existing locks in ldlm_handle_enqueue() (3764)
6057 - don't resolve lock handle twice in recovery avoiding race (4401)
6058 - revalidate should check working dir is a directory (4134)
6060 - don't always mark "slow" obdfilter messages as errors (4418)
6062 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6065 - don't close LustreDB during write_conf until it is done (3860)
6066 - fix typo in lconf for_each_profile (3821)
6067 - allow dumping logs from multiple threads at one time (3820)
6068 - don't allow multiple threads in OSC recovery (3812)
6069 - fix debug_size parameters (3864)
6070 - fix mds_postrecov to initialize import for llog ctxt (3121)
6071 - replace config semaphore with spinlock (3306)
6072 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6073 - don't allow enqueue to complete on a destroyed export (3822)
6074 - down write_lock before checking llog header bitmap (3825)
6075 - recover from lock replay timeout (3764)
6076 - up llog sem before sending rpc (3652)
6077 - reduce ns lock hold times when setting kms (3267)
6078 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6079 - fix NULL deref and obd_dev leak on setup error (3312)
6080 - replace some LBUG about llog ops with error handling (3841)
6081 - don't match INVALID dentries from d_lookup and spin (3784)
6082 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6083 - fix invalid assertion in ptlrpc_set_wait (3880)
6085 - add libwrap support for the TCP acceptor (3996)
6086 - add /proc/sys/portals/routes for non-root route listing (3994)
6087 - allow setting MDS UUID in .xml (2580)
6088 - print the stack of a process that LBUGs (4228)
6090 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6093 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6094 - make sure to unset replay flag from failed open requests (3440)
6095 - if default stripe count is 0, use OST count for inode size (3636)
6096 - update parent mtime/ctime on client for create/unlink (2611)
6097 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6098 - free recovery state on server during a forced cleanup (3571)
6099 - unregister_reply for resent reqs (3063)
6100 - loop back devices mounting and status check on 2.6 (3563)
6101 - fix resource-creation race that can provoke i_size == 0 (3513)
6102 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6103 - more debugging for page-accounting assertion (3746)
6104 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6105 - avoid deadlock after precreation failure (3758)
6106 - fix race and lock order deadlock in orphan handling (3450, 3750)
6107 - add validity checks when grabbing inodes from l_ast_data (3599)
6109 - add /proc/.../recovery_status to obdfilter (3428)
6110 - lightweight CDEBUG infrastructure, debug daemon (3668)
6111 - change default OSC RPC parameters to be better on small clusters
6112 - turn off OST read cache for files smaller than 32MB
6113 - install man pages and include them in rpms (3100)
6114 - add new init script for (un)mounting lustre filesystems (2593)
6115 - run chkconfig in %post for init scripts (3701)
6116 - drop scimac NAL (unmaintained)
6118 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6121 - clean kiobufs before and after use (3485)
6122 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6123 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6124 - change init script to fail more clearly if not run as root (1528)
6125 - allow clients to reconnect during replay (1742)
6126 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6127 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6128 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6129 - fix cond_resched (3554)
6130 - only evict unfinished clients after recovery (3515)
6131 - allow bulk resend, prevent data loss (3570)
6132 - dynamic ptlrpc request buffer allocation (2102)
6133 - don't allow unlinking open directory if it isn't empty (2904)
6134 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6135 - remove extraneous obd dereference causing LASSERT failure (3334)
6136 - don't use get_cycles() when creating temp. files on the mds (3156)
6137 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6138 - handle EEXIST for set-stripe, set proper directory name (3336)
6140 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6141 - fix ksocknal_fmb_callback() error messages (2918)
6143 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6146 - don't copy lvb into (possibly NULL) reply on error (2983)
6147 - don't deref dentry after dput, don't free lvb on error (2922)
6148 - use the kms to determine writeback rpc length (2947)
6149 - increment oti_logcookies when osc is inactive (2948)
6150 - update client's i_blocks count via lvb messages (2543)
6151 - handle intent open/close of special files properly (1557)
6152 - mount MDS with errors=remount-ro, like obdfilter (2009)
6153 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6154 - don't use cancelling-locks' kms values (2947)
6155 - use highest lock extent for kms, not last one (2925)
6156 - don't dereference ERR_PTR() dentry in error handling path (3107)
6157 - fix thread race in portals_debug_dumplog() (3122)
6158 - create lprocfs device entries at setup instead of at attach (1519)
6159 - common AST error handler, don't evict client on completion race (3145)
6160 - zero nameidata in detach_mnt in 2.6 (3118)
6161 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6162 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6163 - initialize RPC timeout timer earlier for 2.6 (3219)
6164 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6165 - print nal/nid for unknown nid (3258)
6166 - additional checks for oscc recovery before doing precreate (3284)
6167 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6168 - don't crash in mdc_close for bad permissions on open (3285)
6169 - zero i_rdev for non-device files (3147)
6170 - clear page->private before handing to FS, better assertion (3119)
6171 - tune the read pipeline (3236)
6172 - fix incorrect decref of invalidated dentry (2350)
6173 - provide read-ahead stats and refine rpc in flight stats (3328)
6174 - don't hold journal transaction open across create RPC (3313)
6175 - update atime on MDS at close time (3265)
6176 - close LDAP connection when recovering to avoid server load (3315)
6177 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6178 - don't leak open file on MDS after open resend (3325)
6179 - serialize filter_precreate and filter_destroy_precreated (3329)
6180 - loop device shouldn't call sync_dev() for nul device (3092)
6181 - clear page cache after eviction (2766)
6182 - resynchronize MDS->OST in background (2824)
6183 - refuse to mount the same filesystem twice on same mountpoint (3394)
6184 - allow llmount to create routes for mounting behind routers (3320)
6185 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6186 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6187 - fix rare problem with rename on htree directories (3417)
6189 - allow default OST striping configuration per directory (1414)
6190 - fix compilation for qswnal for 2.6 kernels (3125)
6191 - increase maximum number of MDS request buffers for large systems
6192 - change liblustreapi to be useful for external progs like lfsck (3098)
6193 - increase local configuration timeout for slow disks (3353)
6194 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6196 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6199 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6200 - don't overwrite extent policy data in reply if lock was blocked (2901)
6201 - drop filter export grants atomically with removal from device (2663)
6202 - del obd_self_export from work_list in class_disconnect_exports (2908)
6203 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6204 - swab reply message in mdc_close, other PPC fixes (2464)
6205 - fix destroying of named logs (2325)
6206 - overwrite old logs when running lconf --write_conf (2264)
6207 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6208 - fix race in target_handle_connect (2898)
6209 - mds_reint_create() should take same inode create lock (2926)
6210 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6211 - don't close files for self_export to avoid uninitialized obd (2936)
6212 - allow MDS with the same name as client node (2939)
6213 - hold dentry reference for closed log files for unlink (2325)
6214 - reserve space for all logs during transactions (2059)
6215 - don't evict page beyond end of stripe extent (2925)
6216 - don't oops on a deleted current working directory (2399)
6217 - handle hard links to targets without a parent properly (2517)
6218 - don't dereference NULL lock when racing during eviction (2867)
6219 - don't grow lock extents when lots of conflicting locks (2919)
6221 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6224 - account for cache space usage on clients to avoid data loss (974)
6225 - lfsck support in lustre kernel code (2349)
6226 - reduce journal credits needed for BRW writes (2370)
6227 - orphan handling to avoid losing space on client/server crashes
6228 - ptlrpcd can be blocked, stopping ALL progress (2477)
6229 - use lock value blocks to assist in proper KMS, faster stat (1021)
6230 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6231 - recovery for initial connections (2355)
6232 - fixes for mds_cleanup_orphans (1934)
6233 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6234 - block all file creations until orphan recovery completes (1901)
6235 - client remove rq_connection from request struct (2423)
6236 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6237 - recovery timer race (2670)
6238 - mdc_close recovey bug (2532)
6239 - ptlrpc cleanup bug (2710)
6240 - mds timeout on local locks (2588)
6241 - namespace lock held during RPCs (2431)
6242 - handle interrupted sync write properly (2503)
6243 - don't try to handle a message that hasn't been replied to (2699)
6244 - client assert failure during cleanup after abort recovery (2701)
6245 - leak mdc device after failed mount (2712)
6246 - ptlrpc_check_set allows timedout requests to complete (2714)
6247 - wait for inflight reqs when ptlrpcd finishes (2710)
6248 - make sure unregistered services are removed from the srv_list
6249 - reset bulk XID's when resending them (caught by 1138 test)
6250 - unregister_bulk after timeout
6251 - fix lconf error (2694)
6252 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6253 - readahead locks pages, leaves pending causing memory pressure (2673)
6254 - increase OST request buffers to 4096 on large machines (2729)
6255 - fix up permission of existing directories in simple_mkdir (2661)
6256 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6257 - don't assign transno to errored transactions (2742)
6258 - don't delete objects on OST if given a bogus objid from MDS (2751)
6259 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6260 - if rq_no_resend, then timeout request after recovery (2432)
6261 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6262 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6263 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6264 - MDS recovery shouldn't delete objects that it has given out (2730)
6265 - if enqueue arrives after completion, don't clobber LVB (2819)
6266 - don't unlock pages twice when trigger_group_io returns error (2814)
6267 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6268 - don't write pages to disk if there was an error (1450)
6269 - don't ping imports that have recovery disabled (2676)
6270 - take buffered bytes into account when balancing socknal conn (2817)
6271 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6272 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6273 - remove little-endian swabbing of llog records (1987)
6274 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6275 - retry reposting request buffers if they fail (1191)
6276 - grow extent at grant time to avoid granting a revoked lock (2809)
6277 - lock revoke doesn't evict page if covered by a second lock (2765)
6278 - disable VM readahead to avoid reading outside lock extents (2805)
6280 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6281 - updated kernel patches for hp-2.4.20 kernel (2681)
6283 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6286 - fix truncated write corruption (2366)
6287 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6289 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6290 - ensure len(uuid) < 37 in lmc (1171)
6291 - fix ia64 OOPS in llog_test (2255)
6292 - zero end of page at obdfilter for partial page writes (2648)
6293 - don't leave stale dentries around after renames (bug 2428)
6294 - fix timeouts when evicting a client with a single lock held (2642)
6295 - set deadline for the initial HELLO message to drain (2634)
6296 - print out dotted-quad IP addresses in the socknal (2302)
6298 - additional debugging for MDS client eviction problem (2443)
6299 - fix mkfsoptions support for osts (2603, 2604)
6301 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6304 - add series for the vanilla 2.6.0 kernel
6305 - add series for the vanilla 2.4.24 kernel
6306 - add series for a cray x86/64 UL kernel drop
6307 - fix xattr patches for the vanilla 2.4.19 series
6309 - generate true UUIDs in lmc (1171)
6310 - have portals stack dumping break in UML (2466)
6311 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6312 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6313 - wake callback waiting threads on client eviction (2460)
6314 - Add --ptldebug and --subsystem to lmc (1719)
6315 - update assertion to allow safe interrupt allocation
6316 - set rq_no_resend for cancel requests (2432)
6317 - recalculate ptlrpcd timeout after resend (2494)
6318 - call vfs_rmdir when removing pending directories (2368)
6319 - fix renaming a file to itself (2429)
6320 - lmc creates a default one-stripe lov (2454)
6321 - expand procfs space to handle large clusters (2326)
6322 - increase UML stack to avoid overflow
6323 - update lconf's list of debug and subsystem masks
6324 - fix lfs find --obd (2510)
6325 - /proc tunable for disabling filter read caching (2591)
6326 - stop rpm packages from altering slapd.conf (2301)
6327 - disable nagle in the socknal under 0conf (2578)
6328 - choose mds inode size based on stripe count (2572)
6329 - fix kernel-source rpm problems (2516)
6331 - add --disable-doc to avoid pdf generation (2421)
6332 - update documentation, tests, type-os, comments
6333 - avoid format warnings on ia64
6334 - remove the TOE NAL
6335 - tiny code cleanups by removing unused fields
6337 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6340 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6341 - remove the most blatant lies from BUILDING file (2371)
6342 - change default debug level to reasonable production setting
6343 - reduce client side cache size to reduce cache flush time
6344 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6345 - make TCP zerocopy and pinger support enabled by default (2476)
6346 - sync writes completed after process exits caused crashes (2319)
6347 - maintain correct mount count on the MDS (2356)
6348 - backout 1557, because 2316 wasn't really fixed
6349 - better file I/O statistics gathering in /proc
6350 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6351 - another kernel patch to fix zero-copy TCP function export
6352 - don't take duplicate lock when processing re-sent getattr (2420)
6353 - lctl uses obd_self_export instead of creating new conn (2353)
6354 - MDS/OST recovery case which requires object creation asserted (2425)
6355 - move lfs from /usr/sbin to /usr/bin in packages
6356 - fix race between mds_client_add and mds_client_free (2417)
6357 - use kmalloc instead of slabs in portals (2430)
6358 - don't create duplicate records when a failover MDS is present (2442)
6359 - remove unnecessary mount age check (2332)
6360 - don't remove directory inodes from locks prematurely (2451)
6361 - don't break if MDS service name is the same as hostname (2103)
6362 - fix races in client write RPC generation when cache full (2482)
6364 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6367 - remove now-unused request->rq_obd (278)
6368 - if an allocation fails, print out how much memory we've used (1933)
6369 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6370 - add the "configurable stack size" patch to most series files (1256)
6371 - ability to write large log records, for 100+ OST configs (2306)
6372 - fix NULL deref when filter_prep fails (2314)
6373 - fix operator precedence error in filter_sync
6374 - dynamic allocation of socknal TX descriptors (2315)
6375 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6376 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6377 - ensure that CWARN messages in Portals always get to the syslog
6378 - __init/__exit are not for prototype decls (ldlm_init/exit)
6379 - x86-64 compile warning fixes
6380 - fix gateway LMC keyword conflict (2318)
6381 - fix MDS lock inversions in getattr/reint paths (1844)
6382 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6383 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6384 - DLM race condition prevented some lock evictions (2328)
6385 - ENOMEM detection and retry on socknal sends (2230)
6386 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6387 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6388 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6389 - better error messages when a client is rejected during recovery (1505)
6390 - avoid cancelling locks which were never granted, after failure (2330)
6391 - fix i_sem/journal inversion in mds_client_add (2333)
6392 - fix truncate/getattr lock cycle deadlock (2334)
6393 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6394 - fix two rare exit paths which could leak an l_lock() ref (2321)
6395 - fencepost error in MDS/OST orphan recovery (2226)
6396 - make log record alignment 8 bytes (1988)
6397 - lstripe now fails when requested offset > ost_count (2237)
6398 - ensure that all kernel series have a complete list.h (1607)
6399 - fix crashes in special-file operations (2316)
6400 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6402 - allow configurable automake binary, for testing new versions
6403 - small update to the lfs documentation
6405 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6407 * fix negative export reference count in fsfilt_sync (2312)
6409 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6410 * release candidate 0.9.1
6412 - orphans are moved into the PENDING directory for possible recovery
6413 - replayed opens now open by fid for orphan/rename safety (1042)
6414 - last close of an orphan inode generates a transno (683)
6415 - chdir() and mount() now pin the directory entry (1020)
6416 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6417 - discard very old requests without processing them (1502)
6418 - remove some common, well-understood CERRORs (1505)
6419 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6420 - clear "grant" flags in OST replies until OST grant code lands (1644)
6421 - fix read performance by not clobbering i_blksize on client (1598)
6422 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6423 - make LRU size a /proc tunable, clears locks when reduced (707)
6424 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6425 - server locks take references on exports now (1558)
6426 - build fixes for 2.4.20-rh trees (1663)
6427 - return an error from lov_create if all OSCs are inactive (1751)
6428 - fix import levels when a reconnect happens without a timeout (1597)
6429 - exit early from mds_open if we get a lookup error (1749)
6430 - partial page read at EOF wouldn't wait for disk before sending (1642)
6431 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6432 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6433 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6434 - write/truncate lock inversion (1639)
6435 - Don't auto-load obdclass, portals modules during cleanup (1495)
6436 - fix timestamps from jumping to "now" (1763)
6437 - extra journal assertions (1648)
6438 - add an extra multiunlink test (1771)
6439 - fix read_record/write_record API (1776)
6440 - fix leak of offset_extent, possible incorrect i_size later (1772)
6441 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6442 - Debugging for the kqswnal_get_idle_tx problems (1820)
6443 - Allow recovery to be attempted multiple times (1536)
6444 - Write out MDS last_rcvd file after it is first created (1600)
6445 - Fix tx_descriptor leak in failed transmit situations (1827)
6446 - ext3 journaling fixes for assertion failure after IO error (1871)
6447 - class_export_put() on freed export after completion AST error (1896)
6448 - Fix revalidate looping in VFS (1322)
6449 - Don't access a freed export during MDS_REINT timeout (1521)
6450 - Add open-unlink recovery support on the MDS (1673,1764)
6451 - Return an error if no MDS data was read from last_rcvd (1946)
6452 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6453 - Don't setup a disk device that doesn't match exported UUID (317)
6454 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6455 - avoid committing NULL handle in force close
6456 - local.sh is now a one-stripe LOV configuration
6457 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6458 - fix ext3 htree duplicate directory entry corruption (1516)
6459 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6460 - update to new LOV EA format (2097)
6461 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6463 2003-06-15 Phil Schwan <phil@clusterfs.com>
6466 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6467 - per-import/export recovery handling (958, 931, 959)
6468 - multiple last-rcvd slots, for serving multiple FSes (949)
6469 - connections are again shared between multiple imp/exports (963, 964)
6470 - "umount -f" would hang if any requests needed to be sent (393, 978)
6471 - avoid pinning large req buffer by copying for queued messages (989)
6472 - add "uuid" to "lctl device" command to help upcalls (991)
6473 - "open" RPCs with transnos would confuse recovery counters (1037)
6474 - do proper endian conversion of all wire messages (288, 340, 891)
6475 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6476 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6477 - fix ptlrpc_connection leak in target_handle_connect (1174)
6478 - fix import refcounting bug in OST and MDS cleanup (1134)
6479 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6480 - fix very unlikely obd_types race condition (501)
6481 - remove osc_open hack for echo_client (1187)
6482 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6483 - a failure in read_inode2 leads to deadlock (1139)
6484 - cancel ack-locks as soon as transaction is committed (1072)
6485 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6486 - make sure to commitrw after any preprw to avoid deadlock (1162)
6487 - failing to execute a file in a lustre FS would lock inode (1203)
6488 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6489 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6490 - obd_uuid2tgt would walk off the end of the list (1255)
6491 - on IA64 the osc would give portals incorrect bulk size (1258)
6492 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6493 - fix lock inversion caused by new llite matching code (1282)
6494 - limit the number of dirty pages on a client to 10MB (1286)
6495 - timed out locks were not being corrected cancelled (1289)
6496 - fix O_DIRECT above 4GB on IA-32 (1292)
6497 * major user-visible changes
6498 - fail out/fail over policy now controlled by the upcall (993)
6500 - add OBD_PING to check server availability and failure (954)
6501 - lustre messages are now sent in sending host order (288, 340, 891)
6502 - add eadatalen to MDS getattr reply (340)
6503 - OST read replies may contain second buffer, with per-page status (593)
6505 2003-03-11 Phil Schwan <phil@clusterfs.com>
6508 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6509 - failed open()s could cause deadlock; fixed (867, 869)
6510 - stop cancelling OST locks when files are closed (481)
6511 - overlapping XID spaces caused network corruption (851, 853)
6512 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6513 - setattr_raw added, to do single-RPC, server-side setattrs
6514 - lmc/lconf syntax change for OST UUIDs
6515 - fix crashy race condition between ptlrpc_free_req and osc_close
6516 - don't use request in mdc_enqueue if we hit a timeout (889)
6517 - don't set the inode i_size for regular files from the MDS (896)
6518 - handle out of order completion AST (842)
6519 - don't LBUG if a lock request times out after receiving AST (913)
6520 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6521 - if a bad lock AST arrives, send an error instead of dropping entirely
6522 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6523 - fix leak in bulk IO when only partially completed (899, 900, 926)
6524 - fix O_DIRECT for ia64 (55)
6525 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6526 - C-z after timeout could hang a process forever; fixed (977)
6528 - client-side I/O cache (678, 924, 929, 941, 970)
6530 - READPAGE and SETATTRs which don't take server-side locks get
6533 2003-02-11 Phil Schwan <phil@clusterfs.com>
6536 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6537 - Fix fsfilt_extN_readpage() to read a full page of directory
6538 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6539 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6540 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6541 - Most intent operations were removed, in favour of a new RPC mode
6542 that does a single RPC to the server and bypasses most of the VFS
6543 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6544 - Aggressively cancel local locks on DLM servers
6545 - mds_reint_unlink sends EA to the client if it's the last nlink.
6546 client uses that EA to unlink OST objects.
6547 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6548 - recursive symlinks were fixed (439)
6549 - fixed NULL deref in DEBUG_REQ
6550 - filter_update_lastobjid no longer calls sync, which annoyed extN
6551 - fixed multi-client small-writes to a single file problem (445)
6552 - fixed mtime updates during file writes (607)
6553 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6554 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6555 - fixed Linux OST size reporting problem (444, 656)
6556 - OST now updates object mtime with writes or setattr (607, 619)
6557 - client verifies file size before zeroing page past EOF (445)
6558 - OST now writes last allocated objid to disk with allocation (108)
6559 - LOV on echo now works (409)
6561 - mds_reint_unlink sends a new buffer, with the EA included. this
6562 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6563 set if a regular file was being unlinked, and it was the last link
6564 - use PtlGet from the target for bulk writes (315)
6565 - OST now updates object mtime with writes or setattr (607, 619)
6566 - LDLM now has a grant-time callback to revalidate locked items, if
6568 - Many MDS operations were reorganized to combat race conditions
6570 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6571 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6573 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6576 - Fully reactivate OST imports after reconnection (512, others)
6577 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6578 - More graceful error handling for truncating on dead OST (515)
6579 - Don't error out unless we're actually accessing dead stripes (474)
6580 - Fix garbage sizes when stripes are missing (410)
6581 - LRU counters were broken, causing constant lock purge (433, 432)
6582 - garbage on read from stripes with failed OSTs (441)
6583 - mark OSCs as active before reconnecting during recovery (438)
6584 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6585 - lfind did not preserve OST order in output (443)
6586 - symlinks cause hung clients, incorrect data (439)
6587 - stop dereferencing request after dropping refcount (457)
6588 - don't LASSERT(spin_is_locked) on non-SMP (455)
6589 - fixes for many rename() bugs
6590 - fstat didn't correctly synchronize attributes (399)
6591 - server must handle lock cancellation during blocking AST prep (487)
6592 - bulk descriptors were free()d too soon (511)
6593 - fix paths in lconf, which would load incorrect modules (451, 507)
6594 - fix confusing lconf 'host not found' error message (386)
6595 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6596 - fix race condition in mdc_blocking_ast() for inode access (526)
6597 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6598 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6599 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6600 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6601 - fix lstripe in several ways (kernel side) (527)
6602 - fix request leak in ldlm_cli_enqueue (262)
6603 - incorrect OSC was marked inactive after OST failure
6604 - call mds_fs_cleanup before unmounting filesystem (524)
6605 - fix races between taking ns_lock and ldlm_lock_change_resource
6606 - fix races updating LOV export open file list
6607 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6608 - fix recovery NULL deref in ldlm_cli_cancel_unused
6609 - fix some DLM races by using new hash table for lock handles (419)
6610 - permit the client to specify desired inodes, at replay
6611 - duplicate requests when we queue them for replay reintegration
6612 - fix last_rcvd offset calculation
6613 - sync after each recovered transaction, so we always make progress
6614 - never, not always, ERESTART requests without transnos
6615 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6617 - skip replay if the MDS says that the client is already connected
6618 - don't check for a recovery-enabled export to match lctl's UUID
6619 - don't INC_USE_COUNT for phantom exports
6620 - don't crash when cleaning up phantom exports (567)
6621 - don't double-finish or set replay data for errored mdc_open requests
6622 - abort requests when they time out, so we don't get old replies
6623 - send/receive replies for AST messages again
6624 - if the client says that it doesn't have the lock, cancel it on the
6626 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6627 mark it as destroyed, it will all work out when decref is called
6628 - fix module use counts (22, 581)
6630 - ASTs now expect a reply (server cancels lock on error reply)
6632 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6635 - fix many simultaneous client startup (392)
6636 - fix dentry->d_it clobbering
6637 - credentials weren't being shipped for readdir/getattr operations
6638 - remove invalid assertions triggered during some concurrent MD
6640 - proper Lustre versions added (336, 389)
6641 - fix memory leak for create error case (398)
6642 - fix LOV locking bug that would get cli/srv out of sync
6643 - fix echo client over LOV (409)
6644 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6645 - fix double-O_EXCL intent crash (424)
6646 - avoid sending multiple lock CANCELs (352)
6648 - MDS can do multi-client recovery (modulo bugs in new code)
6650 - many updates, edits, cleanups
6652 2002-11-18 Phil Schwan <phil@clusterfs.com>
6655 - fix null d_it dereference (346)
6656 - fix full OST/dbench hang (333)
6657 - fix permission problem with file removal (286)
6658 - fix removal of OSCs from LOV when they fail
6659 - fix NULL deref during bulk timeout (214)
6660 - fix problems related to multiple filesystems on one MDS (241)
6661 - fixed serious subtle metadata locking bugs
6662 - free locks on clients when inodes are removed due to memory
6664 - fix inode pointer in lock data (285)
6665 - partial support for multiple MDS on a single host (241)
6666 - data locks weren't cancelled at clear_inode time (290, 311)
6667 - intent locks could lead to unbounded lock growth (205)
6668 - added a maximum lock count, an LRU list, and a flusher
6669 - fix multiple rename (365)
6670 - properly abstracted the echo client
6671 - OSC locked 1 byte too many; fixed
6672 - rewrote brw callback code:
6673 - fixed recovery bugs related to LOVs (306)
6674 - fixed too-many-pages-in-one-write crash (191)
6675 - fixed (again) crash in sync_io_timeout (214)
6676 - probably fixed callback-related race (385)
6678 - Add capability to MDS protocol
6679 - LDLM cancellations and callbacks on different portals
6681 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6684 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6685 - defer open object destruction to close time (601981, 138)
6686 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6687 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6688 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6689 - don't call obd_setattr in ll_file_release for destroyed objects
6690 * protocol change to lustre_msg: move |version| and add |flags|
6691 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6692 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6693 * added replay of create, unlink, link and rename operations during
6694 MDS failover; recovery should be much more robust now
6695 * remove failed OSCs from LOVs (only lov_create uses this so far)
6696 * the lustre-HOWTO was brought (more) up to date (582544)
6698 2002-10-23 Phil Schwan <phil@clusterfs.com>
6701 - in-use dentries weren't being reused properly (617851)
6702 - prevent multiple LDLM setup (599178)
6703 - fix LOV size calculations for truncate (617853)
6704 - fix client handling of MDS intent errors (POSIX)
6705 - fix permission bug in lovstripe.c test (624321)
6706 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6707 - truncate past end of file could corrupt data
6708 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6709 - a race in recovery could return ETIMEDOUT to apps (623947)
6710 - building outside the source directory was fixed
6711 * the lustre-HOWTO was brought (more) up to date (582544)
6712 * major progress was made on recovery functionality
6714 2002-10-10 Phil Schwan <phil@clusterfs.com>
6717 - recovery deadlock fix
6718 - rm -rf causes LBUG fix (617817)
6719 - file open by multiple tasks fix (618962)
6720 - directory permissions bugs (602707 and 620007)
6721 - journal_stop fixed with locking (611313)
6722 - O_APPEND failures resolved (618273, perhaps 614459)
6723 - lconf PATH fix (619770)
6724 - IA64 build fix (621450)
6725 - RPC buffer sizes scale with amount of memory
6727 2002-10-01 Phil Schwan <phil@clusterfs.com>
6730 - locks would be cancelled without throwing away data pages,
6731 resulting in inconsistent data (605627)
6732 - inode attributes were not always being refreshed (605627, 612449)
6733 - lconf now continues to cleanup after lctl reports an error
6734 - MDS now enforces user permissions (602707)
6735 - lprocfs cleanup fixed, but not yet enabled (614157)
6736 - fixed infinite server hang, should a client not respond to an AST
6737 - avoid going into recovery if user calls readlink() with a buffer
6738 that's too small (613941)
6739 - AST RPCs no longer require replies (614867) -- this may be changed
6740 - don't crash server if client sends an IOV that's too big (611336)
6741 - fixed lock conversion deadlock (611892)
6742 - fixed the following of symlinks (614622)
6743 * recovery: the server can remove locks from a client that dies, other
6744 clients can make progress
6745 * more extN patch fixes
6746 * compile-time configurable ptlrpc buffer allocations
6748 - collaborative read cache document
6749 - Lustre Lite Performance CDR document-in-progress
6751 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6754 - fix typo in patch-2.4.18
6756 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6759 - clear ptlrpc request each time in handle_incoming_request()
6760 - unlink of files now destroys the object on the OST
6762 2002-09-19 Peter Braam <braam@clusterfs.com>
6764 * add hard link support
6765 * change obdfile creation method
6766 * kernel patch changed
6768 2002-09-19 Peter Braam <braam@clusterfs.com>
6771 - stack overflow bug in extN fixed
6773 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6775 * documentation updates
6776 - add man pages for config tools
6777 - update tests/README to describe testing with new config tools
6778 - finish metadata API descriptions
6779 * bug fixes and cleanups
6780 - statfs workaround for 16TB limit
6781 - LOV stripe allocation improved, can stripe on subset of OSTs
6782 - LOV file size/IO offset was wrong for files > 4GB in size
6783 - object EA data was being dropped, caused files to be unreadable
6784 - memory overflow with non-LOV OST caused memory corruption
6785 - fixed regression tests to work with new config tools, obdfilter
6786 - fixed bug when directory size became larger than 1 block
6787 - fixed bug (for single client case) when PWD was deleted
6788 - invalidate local directory pages when doing intent-based ops
6789 - avoid LDLM oops when lock callback contained bad data
6791 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6793 * documentation updates
6794 * bug fixes and cleanups
6795 - configuration tools
6798 - 64-bit compile warnings
6799 - 64-bit internal statfs data
6801 * test_brw on persistent OST devices
6803 * lprocfs (disabled)
6805 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6807 * documentation updates
6808 * bug fixes and cleanups
6809 * configuration tools
6811 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6814 * many small fixes to 0_5_4
6815 * io/network handling
6816 * thinkos in MDS operations
6818 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6821 * crucial basic fixes to 0.5.3
6822 * IOR, Iozone work over Elan
6825 2002-08-07 Phil Schwan <phil@clusterfs.com>
6826 * version 0_5_3, our first alpha
6827 * we use the new Portals iovs
6828 * documentation updates
6829 * bug fixes and cleanups
6830 * small changes in the DLM wire protocol
6832 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6833 * version 0_5_1 with some initial stability,
6834 * locking on MD and file I/O.
6835 * documentation updates
6836 * several bug fixes since 0.5.0
6837 * small changes in wire protocol
6839 2002-07-18 Phil Schwan <phil@clusterfs.com>
6841 * delivered as Lustre Light Alpha
6842 * fixed a crash after handling invalid MDS requests
6843 * fixed directory pages for architectures with non-4k pages sizes
6845 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6846 * release version v0_4_4
6847 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6849 * Fixes a number of interruption problems with OST operations.
6850 * Update documentation for portals header changes
6851 * Move all wire protocol structs/defines to lustre_idl.h
6852 * Fixes symlink length bug.
6853 * Add tcpdump to repository.
6855 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6856 * release version v0_4_3
6857 * Fixes statfs for inodes on extN.
6858 * Fixes bug in runtests which would delete /etc/hosts.
6859 * Use 64-bit object IDs wherever possible (not into VFS though)
6860 Remove ost_get_info, which is unused by lustre, and out of date.
6862 2002-07-03 Peter Braam <braam@clusterfs.com>
6863 * release version v0_4_2 Fixes a lookup error (type not passed)
6864 * move forward to head of Portals
6865 * move forward to latest Lustre kernel
6867 2002-06-25 Peter Braam <braam@clusterfs.com>
6868 * release version v0_4_1. Hopefully stable on single node use.