3 * See https://wiki.hpdd.intel.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 3.10.0-862.3.2.el7 (RHEL7.5)
12 3.0.101-0.47.71 (SLES11 SP3)
13 3.0.101-107 (SLES11 SP4)
14 3.12.74-60.64.40 (SLES12 SP1)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.132-94.33 (SLES12 SP3)
17 vanilla linux 4.6.7 (ZFS only)
18 * Client known to build on unpatched kernels:
19 2.6.32-431.29.2.el6 (RHEL6.5)
20 2.6.32-504.30.3.el6 (RHEL6.6)
21 2.6.32-573.26.1.el6 (RHEL6.7)
22 2.6.32-642.15.1.el6 (RHEL6.8)
23 2.6.32-696.30.1.el6 (RHEL6.9)
24 2.6.32-754.2.1.el6 (RHEL6.10)
25 3.10.0-862.3.2.el7 (RHEL7.5)
26 3.0.101-0.47.71 (SLES11 SP3)
27 3.0.101-107 (SLES11 SP4)
28 3.12.74-60.64.40 (SLES12 SP1)
29 4.4.120-92.70 (SLES12 SP2)
30 4.4.133-94.33 (SLES12 SP3)
31 4.4.0-116 (Ubuntu 16.04)
33 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
34 * Recommended ZFS / SPL version: 0.7.9
35 * Tested with ZFS / SPL version: 0.7.9
36 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
37 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
38 stack overflow. For more information, please refer to bugzilla 17630.
39 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
40 the re-exporting nodes to be mounted with "32bitapi" mount option
42 05-22-2018 Intel Corporation
44 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
45 for currently supported client and server kernel versions.
46 * Server known to build on patched kernels:
47 2.6.32-431.29.2.el6 (RHEL6.5)
48 2.6.32-504.30.3.el6 (RHEL6.6)
49 2.6.32-573.26.1.el6 (RHEL6.7)
50 2.6.32-642.15.1.el6 (RHEL6.8)
51 2.6.32-696.30.1.el6 (RHEL6.9)
52 3.10.0-862.9.1.el7 (RHEL7.5)
53 3.0.101-0.47.71 (SLES11 SP3)
54 3.0.101-107 (SLES11 SP4)
55 3.12.74-60.64.40 (SLES12 SP1)
56 4.4.120-92.70 (SLES12 SP2)
57 4.4.132-94.33 (SLES12 SP3)
58 vanilla linux 4.6.7 (ZFS only)
59 * Client known to build on unpatched kernels:
60 2.6.32-431.29.2.el6 (RHEL6.5)
61 2.6.32-504.30.3.el6 (RHEL6.6)
62 2.6.32-573.26.1.el6 (RHEL6.7)
63 2.6.32-642.15.1.el6 (RHEL6.8)
64 2.6.32-696.30.1.el6 (RHEL6.9)
65 3.10.0-862.9.1.el7 (RHEL7.5)
66 3.0.101-0.47.71 (SLES11 SP3)
67 3.0.101-107 (SLES11 SP4)
68 3.12.74-60.64.40 (SLES12 SP1)
69 4.4.120-92.70 (SLES12 SP2)
70 4.4.133-94.33 (SLES12 SP3)
71 4.4.0-116 (Ubuntu 16.04)
73 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
74 * Recommended ZFS / SPL version: 0.7.9
75 * Tested with ZFS / SPL version: 0.7.9
76 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
77 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
78 stack overflow. For more information, please refer to bugzilla 17630.
79 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
80 the re-exporting nodes to be mounted with "32bitapi" mount option
82 01-31-2018 Intel Corporation
84 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
85 for currently supported client and server kernel versions.
86 * Server known to build on patched kernels:
87 2.6.32-431.29.2.el6 (RHEL6.5)
88 2.6.32-504.30.3.el6 (RHEL6.6)
89 2.6.32-573.26.1.el6 (RHEL6.7)
90 2.6.32-642.15.1.el6 (RHEL6.8)
91 2.6.32-696.23.1.el6 (RHEL6.9)
92 3.10.0-693.21.1.el7 (RHEL7.4)
93 3.0.101-0.47.71 (SLES11 SP3)
94 3.0.101-107 (SLES11 SP4)
95 3.12.74-60.64.40 (SLES12 SP1)
96 4.4.120-92.70 (SLES12 SP2)
97 4.4.120-94.17 (SLES12 SP3)
98 vanilla linux 4.6.7 (ZFS only)
99 * Client known to build on unpatched kernels:
100 2.6.32-431.29.2.el6 (RHEL6.5)
101 2.6.32-504.30.3.el6 (RHEL6.6)
102 2.6.32-573.26.1.el6 (RHEL6.7)
103 2.6.32-642.15.1.el6 (RHEL6.8)
104 2.6.32-696.23.1.el6 (RHEL6.9)
105 3.10.0-693.21.1.el7 (RHEL7.4)
106 3.0.101-0.47.71 (SLES11 SP3)
107 3.0.101-107 (SLES11 SP4)
108 3.12.74-60.64.40 (SLES12 SP1)
109 4.4.120-92.70 (SLES12 SP2)
110 4.4.120-94.17 (SLES12 SP3)
111 4.4.0-116 (Ubuntu 16.04)
113 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
114 * Recommended ZFS / SPL version: 0.7.5
115 * Tested with ZFS / SPL version: 0.7.5
116 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
117 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
118 stack overflow. For more information, please refer to bugzilla 17630.
119 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
120 the re-exporting nodes to be mounted with "32bitapi" mount option
122 --------------------------------------------------------------------------------
124 12-11-2017 Intel Corporation
126 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
127 for currently supported client and server kernel versions.
128 * Server known to build on patched kernels:
129 2.6.32-431.29.2.el6 (RHEL6.5)
130 2.6.32-504.30.3.el6 (RHEL6.6)
131 2.6.32-573.26.1.el6 (RHEL6.7)
132 2.6.32-642.15.1.el6 (RHEL6.8)
133 2.6.32-696.20.1.el6 (RHEL6.9)
134 3.10.0-693.17.1.el7 (RHEL7.4)
135 3.0.101-0.47.71 (SLES11 SP3)
136 3.0.101-107 (SLES11 SP4)
137 3.12.74-60.64.40 (SLES12 SP1)
138 4.4.103-92.56 (SLES12 SP2)
139 4.4.103-6.38 (SLES12 SP3)
140 vanilla linux 4.6.7 (ZFS only)
141 * Client known to build on unpatched kernels:
142 2.6.32-431.29.2.el6 (RHEL6.5)
143 2.6.32-504.30.3.el6 (RHEL6.6)
144 2.6.32-573.26.1.el6 (RHEL6.7)
145 2.6.32-642.15.1.el6 (RHEL6.8)
146 2.6.32-696.20.1.el6 (RHEL6.9)
147 3.10.0-693.17.1.el7 (RHEL7.4)
148 3.0.101-0.47.71 (SLES11 SP3)
149 3.0.101-107 (SLES11 SP4)
150 3.12.74-60.64.40 (SLES12 SP1)
151 4.4.103-92.56 (SLES12 SP2)
152 4.4.103-6.38 (SLES12 SP3)
154 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
155 * Recommended ZFS / SPL version: 0.7.3
156 * Tested with ZFS / SPL version: 0.7.3
157 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
158 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
159 stack overflow. For more information, please refer to bugzilla 17630.
160 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
161 the re-exporting nodes to be mounted with "32bitapi" mount option
163 --------------------------------------------------------------------------------
165 10-02-2017 Intel Corporation
167 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
168 for currently supported client and server kernel versions.
169 * Server known to build on patched kernels:
170 2.6.32-431.29.2.el6 (RHEL6.5)
171 2.6.32-504.30.3.el6 (RHEL6.6)
172 2.6.32-573.26.1.el6 (RHEL6.7)
173 2.6.32-642.15.1.el6 (RHEL6.8)
174 2.6.32-696.10.3.el6 (RHEL6.9)
175 3.10.0-693.2.2.el7 (RHEL7.4)
176 3.0.101-0.47.71 (SLES11 SP3)
177 3.0.101-107 (SLES11 SP4)
178 3.12.74-60.64.40 (SLES12 SP1)
179 4.4.74-92.35 (SLES12 SP2)
180 vanilla linux 4.6.7 (ZFS only)
181 * Client known to build on unpatched kernels:
182 2.6.32-431.29.2.el6 (RHEL6.5)
183 2.6.32-504.30.3.el6 (RHEL6.6)
184 2.6.32-573.26.1.el6 (RHEL6.7)
185 2.6.32-642.15.1.el6 (RHEL6.8)
186 2.6.32-696.10.3.el6 (RHEL6.9)
187 3.10.0-693.2.2.el7 (RHEL7.4)
188 3.0.101-0.47.71 (SLES11 SP3)
189 3.0.101-107 (SLES11 SP4)
190 3.12.74-60.64.40 (SLES12 SP1)
191 4.4.74-92.35 (SLES12 SP2)
193 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
194 * Recommended ZFS / SPL version: 0.7.1
195 * Tested with ZFS / SPL version: 0.7.1
196 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
197 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
198 stack overflow. For more information, please refer to bugzilla 17630.
199 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
200 the re-exporting nodes to be mounted with "32bitapi" mount option
202 --------------------------------------------------------------------------------
204 2017-07-05 Intel Corporation
206 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
207 for currently supported client and server kernel versions.
208 * Server known to build on patched kernels:
209 2.6.32-431.29.2.el6 (RHEL6.5)
210 2.6.32-504.30.3.el6 (RHEL6.6)
211 2.6.32-573.26.1.el6 (RHEL6.7)
212 2.6.32-642.15.1.el6 (RHEL6.8)
213 2.6.32-696.3.1.el6 (RHEL6.9)
214 3.10.0-514.21.1.el7 (RHEL7.3)
215 3.0.101-0.47.71 (SLES11 SP3)
216 3.0.101-100 (SLES11 SP4)
217 3.12.74-60.64.40 (SLES12 SP1)
218 4.4.59-92.17 (SLES12 SP2)
219 vanilla linux 4.6.7 (ZFS only)
220 * Client known to build on unpatched kernels:
221 2.6.32-431.29.2.el6 (RHEL6.5)
222 2.6.32-504.30.3.el6 (RHEL6.6)
223 2.6.32-573.26.1.el6 (RHEL6.7)
224 2.6.32-642.15.1.el6 (RHEL6.8)
225 2.6.32-696.3.1.el6 (RHEL6.9)
226 3.10.0-514.21.1.el7 (RHEL7.3)
227 3.0.101-0.47.71 (SLES11 SP3)
228 3.0.101-100 (SLES11 SP4)
229 3.12.74-60.64.40 (SLES12 SP1)
230 4.4.59-92.17 (SLES12 SP2)
232 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
233 * Recommended ZFS / SPL version: 0.6.5.9
234 * Tested with ZFS / SPL version: 0.6.5.9
235 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
236 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
237 stack overflow. For more information, please refer to bugzilla 17630.
238 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
239 the re-exporting nodes to be mounted with "32bitapi" mount option
241 --------------------------------------------------------------------------------
243 12-07-2016 Intel Corporation
245 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
246 for currently supported client and server kernel versions.
247 * Server known to build on patched kernels:
248 2.6.32-431.29.2.el6 (RHEL6.5)
249 2.6.32-504.30.3.el6 (RHEL6.6)
250 2.6.32-573.26.1.el6 (RHEL6.7)
251 2.6.32-642.11.1.el6 (RHEL6.8)
252 3.10.0-514.el7 (RHEL7.3)
253 3.0.101-0.47.71 (SLES11 SP3)
254 3.0.101-80 (SLES11 SP4)
255 3.12.62-60.64.8 (SLES12 SP1)
256 vanilla linux 4.5.7 (ZFS only)
257 * Client known to build on unpatched kernels:
258 2.6.32-431.29.2.el6 (RHEL6.5)
259 2.6.32-504.30.3.el6 (RHEL6.6)
260 2.6.32-573.26.1.el6 (RHEL6.7)
261 2.6.32-642.11.1.el6 (RHEL6.8)
262 3.10.0-514.el7 (RHEL7.3)
263 3.0.101-0.47.71 (SLES11 SP3)
264 3.0.101-80 (SLES11 SP4)
265 3.12.62-60.64.8 (SLES12 SP1)
267 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
268 * Recommended ZFS / SPL version: 0.6.5.8
269 * Tested with ZFS / SPL version: 0.6.5.8
270 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
271 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
272 stack overflow. For more information, please refer to bugzilla 17630.
273 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
274 the re-exporting nodes to be mounted with "32bitapi" mount option
276 --------------------------------------------------------------------------------
278 02-29-2016 Intel Corporation
280 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
281 for currently supported client and server kernel versions.
282 * Server known to build on patched kernels:
283 2.6.32-431.29.2.el6 (RHEL6.5)
284 2.6.32-504.30.3.el6 (RHEL6.6)
285 2.6.32-573.12.1.el6 (RHEL6.7)
286 3.10.0-327.3.1.el7 (RHEL7.2)
287 3.0.101-0.47.71 (SLES11 SP3)
288 3.0.101-68 (SLES11 SP4)
289 vanilla linux 4.2.1 (ZFS only)
290 * Client known to build on unpatched kernels:
291 2.6.32-431.29.2.el6 (RHEL6.5)
292 2.6.32-504.30.3.el6 (RHEL6.6)
293 2.6.32-573.12.1.el6 (RHEL6.7)
294 3.10.0-327.3.1.el7 (RHEL7.2)
295 3.0.101-0.47.71 (SLES11 SP3)
296 3.0.101-68 (SLES11 SP4)
299 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
300 * Recommended ZFS / SPL version: 0.6.4.2
301 * Tested with ZFS / SPL version: 0.6.4.2
302 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
303 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
304 stack overflow. For more information, please refer to bugzilla 17630.
305 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
306 the re-exporting nodes to be mounted with "32bitapi" mount option
308 --------------------------------------------------------------------------------
310 03-10-2015 Intel Corporation
312 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
313 for currently supported client and server kernel versions.
314 * Server known to build on patched kernels:
315 2.6.32-431.29.2.el6 (RHEL6.5)
316 2.6.32-504.8.1.el6 (RHEL6.6)
317 3.0.101-0.46 (SLES11 SP3)
318 * Client known to build on unpatched kernels:
319 2.6.32-431.29.2.el6 (RHEL6.5)
320 2.6.32-504.8.1.el6 (RHEL6.6)
321 3.10.0-123.20.1.el7 (RHEL7)
322 3.0.101-0.46 (SLES11 SP3)
323 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
324 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
325 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
326 stack overflow. For more information, please refer to bugzilla 17630.
327 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
328 the re-exporting nodes to be mounted with "32bitapi" mount option
330 Severity : enhancement
332 Description: control OST-index in IDIF via ROCOMPAT flag.
333 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
334 last_rcvd file. For new formatted OST device, it will be auto set;
335 for the case of upgrading from old OST device, you can enable it
336 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
337 enabled, for new created OST-object, its IDIF-in-LMA will contain
338 the OST-index; for the existing OST-object, the OSD will convert
339 old format IDIF as new format IDIF with OST-index stored in the
340 LMA EA when accessing such OST-object or via OI scrub. Once such
341 flag is enabled, it cannot be reverted back, so the system cannot
342 be downgraded to the orignal incompatible version.
344 --------------------------------------------------------------------------------
346 07-30-2014 Intel Corporation
348 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
349 for currently supported client and server kernel versions.
350 * Server known to build on patched kernels:
351 2.6.32-431.20.3.el6 (RHEL6.5)
352 3.0.101-0.31 (SLES11 SP3)
353 * Client known to build on unpatched kernels:
354 2.6.32-431.20.3.el6 (RHEL6.5)
355 3.0.101-0.31 (SLES11 SP3)
358 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
359 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
360 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
361 stack overflow. For more information, please refer to bugzilla 17630.
362 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
363 the re-exporting nodes to be mounted with "32bitapi" mount option
365 --------------------------------------------------------------------------------
367 10-31-2013 Intel Corporation
369 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
370 * Server builds for kernels:
371 2.6.32-358.18.1.el6 (RHEL6.4)
372 * Client builds for unpatched kernels:
373 2.6.32-358.18.1.el6 (RHEL6.4)
374 3.0.92-0.8 (SLES11 SP3)
376 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
377 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
378 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
379 stack overflow. For more information, please refer to bugzilla 17630.
380 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
381 the re-exporting nodes to be mounted with "32bitapi" mount option
383 --------------------------------------------------------------------------------
385 05-30-2013 Intel Corporation
387 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
388 * Server builds for kernels:
389 2.6.32-358.6.2.el6 (RHEL6.4)
390 * Client builds for unpatched kernels:
391 2.6.32-358.6.2.el6 (RHEL6.4)
392 3.0.74-0.6.6 (SLES11 SP2)
394 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
395 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
396 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
397 stack overflow. For more information, please refer to bugzilla 17630.
398 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
399 the re-exporting nodes to be mounted with "32bitapi" mount option
401 --------------------------------------------------------------------------------
403 9-30-2012 Whamcloud, Inc.
405 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
406 * Server support for kernels:
407 2.6.32-279.5.1.el6 (RHEL6)
408 * Client support for unpatched kernels:
409 2.6.32-279.5.1.el6 (RHEL6)
411 2.6.18-238.19.1.el5 (RHEL5)
412 2.6.16.60-0.69.1 (SLES10)
413 * Recommended e2fsprogs version: 1.41.90.wc4
414 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
415 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
416 stack overflow. For more information, please refer to bugzilla 17630.
417 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
418 the re-exporting nodes to be mounted with "32bitapi" mount option
420 Severity : enhancement
422 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
424 Severity : enhancement
426 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
428 Severity : enhancement
430 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
432 Severity : enhancement
434 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
436 Severity : enhancement
438 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
440 --------------------------------------------------------------------------------
442 04-30-2012 Whamcloud, Inc.
444 * See https://wiki.hpdd.intel.com/display/PUB/Lustre+Support+Matrix
445 * Server support for kernels:
446 2.6.32-220.4.2.el6 (RHEL6)
447 * Client support for unpatched kernels:
448 2.6.32-220.4.2.el6 (RHEL6)
450 2.6.18-238.19.1.el5 (RHEL5)
451 2.6.16.60-0.69.1 (SLES10)
452 * Recommended e2fsprogs version: 1.41.90.wc4
453 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
454 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
455 stack overflow. For more information, please refer to bugzilla 17630.
456 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
457 the re-exporting nodes to be mounted with "32bitapi" mount option
459 Severity : enhancement
461 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
463 Severity : enhancement
465 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
467 Severity : enhancement
469 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
471 Severity : enhancement
473 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
475 --------------------------------------------------------------------------------
477 09-30-2011 Whamcloud, Inc.
479 * Server support for kernels:
480 2.6.18-238.19.1.el5 (RHEL5)
481 2.6.32-131.6.1.el6 (RHEL6)
482 * Client support for unpatched kernels:
483 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
484 2.6.18-238.19.1.el5 (RHEL5)
485 2.6.32-131.6.1.el6 (RHEL6)
486 2.6.29.4-167.fc11 (FC11)
487 2.6.16.60-0.69.1 (SLES10)
489 * Recommended e2fsprogs version: 1.41.90.wc3
490 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
491 Lustre filesystem with 4K stack may cause a stack overflow. For more
492 information, please refer to bugzilla 17630.
493 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
494 mounted with "32bitapi" mount option
496 Severity : enhancement
497 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
499 Severity : enhancement
500 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
502 Severity : enhancement
503 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
506 Frequency : only with SLES10
508 Description: Use OFED "KMP" provided by Novell
509 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
510 Lustre is now built against this package. That means you need to
511 install the ofed-kmp package from Novell for the patchless client.
512 Note that the ofed-kmp that Novell ships may not exactly match the
513 kernel version but should still be compatible.
515 Severity : enhancement
517 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
518 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
520 --------------------------------------------------------------------------------
522 2010-07-15 Oracle, Inc.
524 * Server support for kernels:
525 2.6.18-164.11.1.el5 (RHEL 5)
526 2.6.18-164.11.1.0.1.el5 (OEL 5)
527 * Client support for unpatched kernels:
528 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
529 2.6.18-164.11.1.el5 (RHEL 5),
530 2.6.18-164.11.1.0.1.el5 (OEL 5)
531 2.6.16.60-0.42.8 (SLES 10),
533 2.6.29.4-167.fc11 (FC11)
534 * Recommended e2fsprogs version: 1.41.10-sun2
536 Severity : enhancement
538 Description: Distribute lfs_migrate script with lustre
539 Details : The lfs_migrate mentioned in the manual is now included in the
540 lustre tree in lustre/scripts and will distributed in lustre RPMs.
542 Severity : enhancement
544 Description: Failover nid list added to import proc file.
546 Severity : enhancement
548 Description: "som_preview" mount option required to use SOM
549 Details : The clients will need to use som_preview mount option
553 Description: Lustre does not do 1MB IOs to HW RAID
554 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
556 Severity : enhancement
558 Description: lfs quota output is not very convenient for awk/sed-parsing
559 Details : Some positions in lfs quota output table could be empty or
560 non-empty which made it hard to parse it with scripts, now
561 a dash is put instead of space where there is not supposed
566 Description: Update to OFED 1.5.1
567 Details : OFED 1.5.1 will be used in environments that don't use
568 vendor supplied OFED.
572 Description: Remove lnet. prefix from lctl params display
573 Details : Also renamed lnet.memused to lnet_memused to
578 Description: Use conf_param -d to remove permanant settings.
582 Description: Add list_param -R
583 Details : lctl list_param now supports recursive mode.
586 Bugzilla : 21829, 22312
587 Description: Fix llobdstat.
588 Details : 1. Now updating the write_bytes counter
589 2. Port script fixes from 1.8.
593 Description: decrease the usage of memory on clients.
594 Details : 1. On clients, recycle dentries and inodes unused.
595 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
600 Description: llobdstat fix and enhancement
601 Details : add a counter to set a limit to how many samples will be returned
602 fix a wildcard in the path to limit to obdfilter stats only
606 Description: lfs check only allowed for root
607 Details : Code cleanups and added a regresion test of non-root use of
612 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
613 Details : Fix hr->hr_index overload.
617 Description: Fix FIEMAP functionality.
621 Description: FC11 patchless client support added
625 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
626 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
630 Description: New acc-sm script and various test-framework enhancments.
634 Description: Previously landed patch was reverted.
638 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
642 Description: Add ldiskfs maxdirsize mount option.
646 Description: Fix mkfs.lustre support for 16TB LUNs.
650 Description: Print mmp_check_interval
651 Details : print mmp_check_interval and make it possible to abort
652 mount operation in case it takes too long.
656 Description: Add cascading_rw to lustre mpi tests.
660 Description: Store SOM data in EA.
661 Details : Also several SOM related cleanups.
665 Description: Add refcount to md_open_data
669 Description: Use refcount on lov_request_set to prevent memory corruption.
673 Description: Statahead: use parent dir's i_mutex for synchronization.
677 Description: loadgen improvements.
681 Description: Enable peer health detection in ksocklnd and ko2iblnd.
685 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
689 Description: Fix race on obd_devcie:md_stats
693 Description: Fix SLES 11 client builds.
694 Details : Resolved uninitalized pointer gcc warnings.
698 Description: Allow non-root access for "lfs check".
699 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
703 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
705 Severity : enhancement
707 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
708 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
712 Description: Sanity test_27p fails.
713 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
714 Sanity tests 27 were cleaned up and improved.
718 Description: Unify req format on client/servers
719 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
720 buffers and swabbers approach..
725 Description: drop join file stuff.
727 Severity : enhancement
729 Description: Add 2.6.27 and SLES11 (patchless client) support
733 Description: Fix in ptlrpc_expire_one_request() to print the signed time
736 Severity : enhancement
738 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
741 Severity : enhancement
743 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
745 Severity : enhancement
747 Description: correctly handle big reply message.
748 Details : send LNet event if reply is bigger than buffer and adjust this buffer
753 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
755 Severity : enhancement
757 Description: Add support for OEL5.
761 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
763 Severity : enhancement
765 Description: "weak-modules" support
766 Details : Implement "weak-modules" support which enables kernel modules
767 to be used with any kernel that implements the same kABI. In
768 order to achieve this modules are now installed in
769 /lib/modules/$(uname -r)/updates/kernel on all distributions.
773 Description: client could not reconnect to OST because of an active request.
774 Details : The client now retries to reconnect to the same server, if a
775 connect request failed with EBUSY or -EAGAIN. Test case 35b
776 conf-sanity.sh has been created.
780 Description: Don't allow make backward step on assignin osc next id.
781 Details : race between allocation next id and ll_sync thread can be cause
782 of set wrong osc next id and can be kill valid ost objects.
786 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
790 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
794 Description: Error handling in osc_statfs_interpret() has been improved.
795 Details : Check in osc_statfs_interpret() for EBADR.
799 Description: Do not send statfs() requests to OSTs disabled by administrator.
800 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
802 Severity : enhancement
804 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
806 Severity : enhancement
808 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
810 Severity : enhancement
812 Description: Enhancement for lfs(1) command to use numeric uid/gid.
816 Description: Improved cleanup procedure for conf-sanity test_46a.
820 Description: Do not start bulk transfer, if client tried to reconnect.
821 Details : Do not start bulk transfer, if client tried to reconnect,
822 but failed due to active rpcs. Check for exp_abort_active_req
823 has been implemented.
827 Description: Prevent inconsistences between linux and lustre mount structures.
828 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
829 Make the sleep interruptible.
831 Severity : enhancement
833 Description: provide server to client comms path
834 Details : server to client communications path via new LDLM_SET_INFO rpc
838 Description: truncate starts GFP_FS allocation under transaction causing deadlock
839 Details : ldiskfs_truncate calls grab_cache_page which may start page
840 allocation under an open transaction. This may lead to
841 calling prune_icache with consequent lustre reentrance.
845 Description: send correctly lsm on open replay
846 Details : MDS is trust to LSM size on replay open, but client can set wrong size
851 Description: enable client interface failover
852 Details : When a child reconnects from another NID, properly update export
853 nid hash position and ldlm reverse import.
855 Severity : enhancement
857 Description: Communicate OST degraded/readonly state via statfs to MDS
858 Details : Flags in the statfs returned from OSTs indicate whether the
859 OST is in a degraded RAID state, or if the filesystem has
860 turned read-only after a filesystem error is detected.
865 Description: don't panic if EPROTO was hit when reading symlink
866 Details : correctly handling request reference in error cases.
868 Severity : enhancement
870 Description: Add LustreNetLink, a kernel-userspace communcation path.
872 Severity : enhancement
874 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
879 Description: access to llog context before init.
880 Details : move handling CATALOGS file at osc layer and forbid access to llog
884 Frequency : rare, only when client is writing to OST on local node
886 Description: Avoid deadlock for local client writes
887 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
888 memory freeing context. This allows OST threads to set the
889 PF_MEMALLOC flag on task structures in order to allocate memory
890 from reserved pools and complete IO.
891 Use GFP_HIGHUSER for OST allocations for non-local client writes,
892 so that the OST threads generate memory pressure and allow
893 inactive pages to be reclaimed.
895 Severity : enhancement
897 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
902 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
903 Details : move ldlm namespace creation in setup phase to avoid grab
904 _lprocfs_lock with cli_sem held.
908 Description: Temporarily disable grant shrink.
909 Details : Disable the feature for debugging.
912 Frequency : only when formatting test filesystems
914 Description: Unable to run several mkfs.lustre on loop devices at the same time
915 Details : mkfs.lustre returns error 256 on the concurrent loop devices
916 formatting. The solution is to proper handle the error.
918 Severity : enhancement
920 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
922 Severity : enhancement
924 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
929 Description: resolve race between obd_disconnect and class_disconnect_exports
930 Details : if obd_disconnect will be called to already disconnected export he
931 forget release one reference and osc module can't unloaded.
933 Severity : enhancement
935 Description: Allow tuning service thread via /proc
936 Details : For each service a new
937 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
938 created that can be used to set min/max thread counts, and get the
939 current number of running threads.
944 Descriptoin: don't return error if have partially created objects for file.
945 Details : lov_update_create_set uses set->set_success as index for created
946 objects, so if some requests will be failed, they will have hole
947 at end of array and qos_shrink_lsm for allocate correct lsm.
949 Severity : enhancement
951 Description: Update OFED support to 1.4.1
956 Description: Reduce small size read RPC
957 Details : Set read-ahead limit for every file and only do read-ahead when
958 available read-ahead pages are bigger than 1M to avoid small size
961 Severity : enhancement
963 Description: move AT tunable parameters for more consistent usage
964 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
966 Severity : enhancement
968 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
969 Details : allow skip disconnected ost for send statfs request and hide error
975 Description: fix racy locking of mballoc block bitmaps causing BUG
976 Details : The locking of the mballoc buddy bitmap and the in-memory
977 block bitmap was using two different spin locks in some
978 cases. This made it possible to incorrectly access the
979 mballoc bitmap while another process was modifying it,
980 causing a sanity assertion to fail. While no on-disk corruption
981 was reported, there was some risk of this happening.
984 Frequency : rare, on llog test 6
986 Descriptoin: don't allow connect to already connected import
987 Details : allowing connect to already connected import is hide connecting problem.
992 Descriptoin: allow kill process which wait statahead result
993 Details : for some reasons 'ls' can stick in waiting result from statahead,
994 in this case need way for kill this process.
996 Severity : enhancement
998 Description: Add state history info file, enhance import info file
999 Details : Track import connection state changes in a new osc/mdc proc file;
1000 add overview-type data to the osc/mdc import proc file.
1002 Severity : enhancement
1004 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1006 Severity : enhancement
1008 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1013 Descriptoin: OST grant too much space to client even there are not enough space.
1014 Details : Client will shrink its grant cache to OST if there are no write
1015 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1016 this grant cache if there are already not enough avaible space
1017 (left_space < total_clients * 32M).
1020 Frequency : start MDS on uncleanly shutdowned MDS device
1022 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1023 Details : stay in waiting mds<>ost recovery finished produce random bugs
1024 due race between two ll_sync thread for one lov target. send
1025 ACTIVATE event only if connect realy finished and import have
1029 Frequency : rare, connect and disconnect target at same time
1031 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1032 Details : don't call obd_disconnect under lov_lock. this long time
1033 operation and can block ptlrpcd which answer to connect request.
1038 Descriptoin: don't lose wakeup for imp_recovery_waitq
1039 Details : recover_import_no_retry or invalidate_import and import_close can
1040 both sleep on imp_recovery_waitq, but we was send only one wakeup
1044 Frequency : always with long access acl
1046 Descriptoin: mds can't pack reply with long acl.
1047 Details : mds don't control size of acl but they limited by reint/getattr
1050 Severity : enhancement
1052 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1054 Severity : enhancement
1056 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1059 Frequency : start MDS on uncleanly shutdowned MDS device
1061 Descriptoin: aborting recovery hang on MDS
1062 Details : don't throttle destroy RPCs for the MDT.
1065 Frequency : on remount
1067 Description: external journal device not working after the remount
1068 Details : clear dev_rdonly flag for external journal devices in
1074 Description: shutdown vs evict race
1075 Details : client_disconnect_export vs connect request race.
1076 if client will evicted at this time - we start invalidate
1077 thread without referece to import and import can be freed
1081 Frequency : rare, need acl's on inode.
1083 Description: client can't handle ost additional correctly
1084 Details : if ost was added after client connected to mds client can have
1085 hit lnet_try_match_md ... to big messages to wide striped files.
1086 in this case need teach client to handle config events about add
1087 lov target and update client max ea size at that event.
1089 Severity : enhancement
1091 Description: Changelogs
1092 Details : Changelogs are a lightweight mechanism to track filesystem
1093 metadata and namespace changes. The changelog is recorded
1094 permanently on the MDTs, and is periodically "consumed" / purged
1095 when records are no longer needed.
1097 Severity : enhancement
1099 Description: compact fld format with extents
1100 Details : Store range of seq rather than every seq in FLD. Seq
1101 controller update FLD rather than clients. In Case of CMD, mdt0
1102 has FLD, all other metadata server act as non persistent proxy
1103 for FLD queries and cache fld entries in fld cache.
1108 Description: don't skip ost target if they assigned to file
1109 Details : Drop slow OSCs if we can, but not for requested start idx.
1110 This means "if OSC is slow and it is not the requested
1111 start OST, then it can be skipped, otherwise skip it only
1112 if it is inactive/recovering/out-of-space.
1116 Description: more cleanup in mds_lov
1117 Details : not send LOV EA under replay, we can't know about they size at this
1118 time. Don't allow client connect to mds before any ost connected,
1119 for avoid problems with LOV EA size and returning EIO to client.
1121 Severity : enhancement
1123 Description: Interoperability at server side (Disk interoperability)
1125 Severity : enhancement
1127 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1129 Severity : enhancement
1131 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1133 Severity : enhancement
1135 Description: New client IO stack (CLIO).
1137 Severity : enhancement
1139 Description: Commit on sharing. Eliminate inter-client dependencies between
1140 uncommitted transactions by doing transaction commits.
1141 Thereby clients may recovery independently.
1144 Frequency : Create a symlink file with a very long name
1146 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1147 Details : If there is no extra space in the request for early cancels,
1148 ldlm_req_handles_avail() returns 0 instead of a negative value.
1150 Severity : enhancement
1152 Description: Add /proc entry for import status
1153 Details : The mdc, osc, and mgc import directories now have
1154 an import directory that contains useful import data for debugging
1155 connection problems.
1157 Severity : enhancement
1159 Description: Re-disable certain /proc logging
1160 Details : Enable and disable client's offset_stats, extents_stats and
1161 extents_stats_per_process stats logging on the fly.
1164 Frequency : Only on FC kernels 2.6.22+
1166 Description: oops in statahead
1167 Details : Do not drop reference count for the dentry from VFS when lookup,
1168 VFS will do that by itself.
1170 Severity : enhancement
1172 Description: Generic /proc file permissions
1173 Details : Set /Proc file permissions in a more generic way to enable non-
1174 root users operate on some /proc files.
1178 Description: Hitting mdc_commit_close() ASSERTION
1179 Details : Properly handle request reference release in
1180 ll_release_openhandle().
1184 Description: quota recovery deadlock during mds failover
1185 Details : This patch includes att18982, att18236, att18237 in bz14840.
1187 1. fix osts hang when mds does failover with quotaon
1188 2. prevent watchdog storm when osts threads wait for the
1193 Frequency : only patchless client
1194 Description: add workaround for race between add/remove dentry from hash
1196 Severity : enhancement
1198 Description: Allow OST glimpses to return PW locks
1202 Description: LBUG when llog conf file is full
1203 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1207 Description: Prevent import from entering FULL state when server in recovery
1211 Description: service mount cannot take device name with ":"
1212 Details : Only when device name contains ":/" will mount treat it as
1218 Description: replace ptlrpcd with the statahead thread to interpret the async
1219 statahead RPC callback
1223 Frequency : on recovery
1224 Description: I/O failures after umount during fail back
1225 Details : if client reconnected to restarted server we need join to recovery
1226 instead of find server handler is changed and process self
1227 eviction with cancel all locks.
1229 Severity : enhancement
1231 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1233 Severity : enhancement
1235 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1237 Severity : enhancement
1239 Description: Upcall on Lustre log has been dumped
1240 Details : Allow for a user mode script to be called once a Lustre log has
1241 been dumped. It passes the filename of the dumped log to the
1242 script, the location of the script can be specified via
1243 /proc/sys/lnet/debug_log_upcall.
1248 Description: avoid idr_remove called for id which is not allocated.
1249 Details : Move assigment s_dev for clustered nfs to end of initialization,
1250 to avoid problem with error handling.
1255 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1256 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1257 into CONN_USED_HASH and this prodice warning when put connection
1258 again in unused hash.
1261 Severity : enhancement
1263 Description: File striping can now be set to use an arbitrary pool of OSTs.
1265 Severity : enhancement
1267 Description: Export bytes_read/bytes_write count on OSC/OST.
1271 Description: Early reply size mismatch, MGC loses connection
1272 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1273 the connect flags are properly negotiated.
1278 Description: quotas are not honored with O_DIRECT
1279 Details : all writes with the flag O_DIRECT will use grants which leads to
1280 this problem. Now using OBD_BRW_SYNC to guard this.
1284 Description: add quota statistics
1285 Details : 1. sort out quota proc entries and proc code.
1286 2. add quota statistics
1288 Severity : enhancement
1290 Description: enable quota support for HEAD.
1294 Description: Properly propagate oinfo flags from lov to osc for statfs
1295 Details : restore missing copy oi_flags to lov requests.
1297 Severity : enhancement
1299 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1300 : l_getgroups(8), lst(8), routerstat(8)
1301 Details : included man pages for llobdstat(8), llstat(8),
1302 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1304 Severity : enhancement
1306 Description: configure's --enable-quota should check the
1307 : kernel .config for CONFIG_QUOTA
1308 Details : configure is terminated if --enable-quota is passed but
1309 : no quota support is in kernel
1313 Description: Remove portals compatibility
1314 Details : Remove portals compatibility, not interoperable with releases
1319 Description: Resolve device initialization race
1320 Details : Prevent proc handler from accessing devices added to the
1321 obd_devs array but yet be intialized.
1323 Severity : enhancement
1325 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1327 Severity : enhancement
1329 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1334 Description: Using wrong pointer in osc_brw_prep_request
1335 Details : Access to array[-1] can produce panic if kernel compiled with
1336 CONFIG_PAGE_ALLOC enabled
1340 Description: Client runs out of low memory
1341 Details : Consider only lowmem when counting initial number of llap pages
1345 Description: *optional* service tags registration
1346 Details : if the "service tags" package is installed on a Lustre node
1347 When the filesystem is mounted, a local-node service tag will
1348 be created. See http://inventory.sun.com/ for more information
1349 about the Service Tags asset management system.
1353 Description: Kernel BUG tries to release flock
1354 Details : Lustre does not destroy flock lock before last reference goes
1355 away. So always drop flock locks when client is evicted and
1356 perform unlock regardless of successfulness of speaking to MDS.
1360 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1364 Description: sanity test 65a fails if stripecount of -1 is set
1365 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1370 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1371 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1374 Severity : enhancement
1376 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1380 Description: do not process already freed flock
1381 Details : flock can possibly be freed by another thread before it reaches
1382 to ldlm_flock_completion_ast.
1386 Description: LBUG during stress test
1387 Details : Need properly lock accesses the flock deadlock detection list.
1391 Description: oops in page fault handler
1392 Details : kernel page fault handler can return two special 'pages' in error
1393 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1397 Description: timeout with invalidate import.
1398 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1399 should be handled by ptlrpcd. This produce long age waiting and
1400 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1402 Severity : enhancement
1404 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1408 Description: enable MGS and MDT services start separately
1409 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1410 a co-located MGS without starting the MGS, which is a complement
1411 to 'nosvc' mount option.
1415 Frequency : after recovery
1416 Description: precreate to many object's after del orphan.
1417 Details : del orphan st in oscc last_id == next_id and this triger growing
1418 count of precreated objects. Set flag LOW to skip increase count
1419 of precreated objects.
1423 Frequency : rare, on clear nid stats
1424 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1425 Details : when clean nid stats sometimes try destroy live entry,
1426 and this produce panic in free.
1430 Description: Stack overflow during MDS log replay
1431 ease stack pressure by using a thread dealing llog_process.
1435 Description: wait until IO finished before start new when do lock cancel.
1436 Details : VM protocol want old IO finished before start new, in this case
1437 need wait until PG_writeback is cleared until check dirty flag and
1438 call writepages in lock cancel callback.
1440 Severity : enhancement
1442 Description: using special macro for print time and cleanup in includes.
1446 Description: mds_mfd_close() ASSERTION(rc == 0)
1447 Details : In mds_mfd_close(), we need protect inode's writecount change
1448 within its orphan write semaphore to prevent possible races.
1452 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1456 Frequency : rare, on shutdown ost
1457 Description: don't hit live lock with umount ost.
1458 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1459 use shrink_dcache_sb instead.
1463 Description: don't panic with use echo client
1464 Details : echo client pass NULL as client nid pointer and this produce null
1465 pointer dereference.
1469 Description: fix build on ppc32
1470 Details : compile code with -m64 flag produce wrong object file for ppc32.
1474 Description: add message levels for liblustreapi
1478 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1479 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1480 will result in an -EAGAIN and in the exit call path we call
1481 alloc_rr() which will with increasing aggressiveness attempt to
1482 aquire precreated objects on the minimum number of required OSCs.
1486 Description: Use old size assignment to avoid deadlock
1487 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1488 the scheduling while holding a spinlock. We do not need locking
1489 for size in ll_update_inode() because size is only updated from
1490 the MDS for directories or files without objects, so there is no
1491 other place to do the update, and concurrent access to such inodes
1492 are protected by the inode lock.
1496 Description: resolve "_IOWR redefined" build error on SLES10
1500 Description: dump the memory debugging after all modules are unloaded to
1501 suppress false negative in conf_sanity test 39
1503 Severity : enhancement
1505 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1510 Description: fix SLES kernel versioning
1511 Details : the kernel version for our SLES 10 kernel did not include a "-"
1512 before the "smp" at the end. while this was not a problem in
1513 general, it did mean that software trying to use the kernel
1514 version to try to detect a vendor specific kernel would fail.
1515 this was most evident by the OFED build scripts.
1519 Description: Don't update lov_desc members until making sure they are valid
1520 Details : When updating lov_desc members via proc fs, need fix their
1521 validities before doing the real update.
1525 Description: don't put request into delay list while invalidate in flight.
1526 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1527 import in flight. this produce timeout for invalidate and sometimes
1528 can cause stale data.
1532 Frequency : on ppc only
1533 Description: not convert ost objects for directory because it's not exist.
1534 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1538 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1540 Severity : enhancement
1542 Description: Update to support 2.6.22.14 vanilla kernel.
1546 Frequency : rare, on recovery
1547 Description: read procfs can produce deadlock in some situation
1548 Details : Holding lprocfs lock which send rpc can produce block for destroy
1549 obd objects and this also block reconnect with -EALREADY. This
1550 isn't fix all lprocfs bugs - but make it rare.
1552 Severity : enhancement
1554 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1557 Frequency : frequent on X2 node
1559 Description: mdc_set_open_replay_data LBUG
1560 Details : Set replay data for requests that are eligible for replay.
1564 Description: lustre_mgs: operation 101 on unconnected MGS
1565 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1566 MGC, and late on MGC cannot successfully connect to MGS and a lot
1567 of the error messages complaining that MGS is not connected.
1571 Frequency : on network error
1572 Description: panic with double free request if network error
1573 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1574 but it's true only for synchronus enqueue, for async enqueue
1575 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1578 Severity : enhancement
1580 Description: client-side metadata stat-ahead during readdir(directory readahead)
1581 Details : perform client-side metadata stat-ahead when the client detects
1582 readdir and sequential stat of dir entries therein
1585 Frequency : on start mds
1587 Description: Implement get_info(last_id) in obdfilter.
1590 Frequency : occasional
1592 Description: Correctly check stale fid, not start epoch if ost not support SOM
1593 Details : open with flag O_CREATE need set old fid in op_fid3 because
1594 op_fid2 was overwritten with new generated fid, but mds can answer
1595 with one of these two fids and both is not stale. Setattr
1596 incorrectly started an epoch and assume will be called
1597 done_writing, but without SOM done_writing ever being called.
1600 Frequency : rare, depends on device drivers and load
1602 Description: MDS or OSS nodes crash due to stack overflow
1603 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1604 In some cases, in conjunction with device drivers that use a lot
1605 of stack the MDS (or possibly OSS) service threads could overflow
1606 the stack. One change which was identified to consume additional
1607 stack has been reworked to avoid the extra stack usage.
1610 Frequency : occasional
1612 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1613 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1614 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1615 oscc_has_objects and see OSC has no objests and call
1616 oscc_internal_create to resend create request.
1618 Severity : enhancement
1620 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1622 Severity : enhancement
1624 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1627 Frequency : very rare
1629 Description: Fix replay if there is an un-replied request and open
1630 Details : In some cases, older replay request will revert the
1631 mcd->mcd_last_xid on MDS which is used to record the client's
1632 latest sent request.
1634 Severity : enhancement
1636 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1638 Severity : enhancement
1640 Description: Add rhel5 support to HEAD.
1642 Serverity : enhancement
1644 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1649 Description: Don't allow skipping OSTs if index has been specified.
1650 Details : Don't allow skipping OSTs if index has been specified, make
1651 locking in internal create lots better.
1655 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1656 Details : access to bitfield in structure is always rounded to long
1657 and this produce problem with not atomic change any bit.
1661 Description: Lustre make rpms failed.
1662 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1663 builds Lustre rpms from tarball.
1666 Frequency : rare on shutdown ost
1668 Description: If llog cancel was not send before clean_exports phase, this can
1669 produce deadlock in llog code.
1670 Details : If llog thread has last reference to obd and call class_import_put
1671 this produce deadlock because llog_cleanup_commit_master wait when
1672 last llog_commit_thread exited, but this never success because was
1673 called from llog_commit_thread.
1677 Description: allow userland application know is lost one of stripes.
1678 Details : fill lvb_blocks with error code on ost and return it to
1679 application if error flag found.
1683 Description: NULL lov_tgts causing MDS oops
1684 Details : more safe checks for NULL lov_tgts for avoid oops.
1686 Severity : enhancement
1688 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1692 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1693 Details : If an OST is down the MDS will hang indefinitely in
1694 obd_statfs_rqset() waiting for the statfs data. While for
1695 MDS QOS usage of statfs, it should not stuck in waiting.
1697 Severity : enhancement
1699 Description: remote_acl support
1700 Details : Support ACL-based permission check for remote user.
1701 Support setfacl/getfacl for remote user with the utils
1702 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1703 the system "{s,g}etfacl" utils.
1705 Severity : enhancement
1707 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1709 Severity : enhancement
1711 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1715 Description: make mgs_setparam() handle fsname containing dash
1716 Details : fsname containing a dash does not work with lctl conf_param
1718 Severity : enhancement
1720 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1722 Severity : enhancement
1724 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1728 Frequency : rare, at shutdown
1729 Description: access already free / zero obd_namespace.
1730 Details : if client_disconnect_export was called without force flag set,
1731 and exist connect request in flight, this can produce access to
1732 NULL pointer (or already free pointer) when connect_interpret
1733 store ocd flags in obd_namespace.
1737 Frequency : only at startup
1738 Description: not alloc memory with spinlock held.
1739 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1740 if any spinlock held.
1742 Severity : enhancement
1744 Description: make lustre randomly fail allocating memory
1745 Details : Make lustre randomly failed allocating memory for testing purpose.
1747 Severity : enhancement
1749 Description: lost problems with lov objid file
1750 Details : Fixes some scability and access to not inited memory problems
1751 in work with lov objdid file.
1756 Description: lfs find does not continue on file error
1757 Details : Continue other files processing when a file/dir is absent.
1761 Description: Inconsistent usage of lustre_pack_reply()
1762 Details : Standardize the usage of lustre_pack_reply() such that it
1763 always generate a CERROR on failure.
1766 Frequency : occasional
1768 Description: MDS hang or stay in waiting lock
1769 Details : If client receive lock with CBPENDING flag ldlm need send lock
1770 cancel as separate rpc, to avoid situation when cancel request
1771 can't processed due all i/o threads stay in wait lock.
1775 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1779 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1783 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1787 Description: Update to RHEL4 latest kernel.
1789 Severity : enhancement
1791 Description: Build SLES10 patchless client fails
1792 Details : The configure was broken by run ./configure with
1793 --with-linux-obj=.... argument for patchless client. When the
1794 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1795 can't search header adequately. Use absolute path such as
1796 -I($LINUX)/include instead.
1800 Description: interrupt oig_wait produce painc on resend.
1801 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1802 private set, and this produce situation when rq_ptlrpcd_data not
1803 copyed to new allocated request and triggered LBUG on assert
1804 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1805 ptlrpcd set if request is changed and can be safety changed to use
1810 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1811 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1812 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1813 these macros and replace them with their difinitions instead.
1817 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1818 Details : Update which_patch & target file for SLES10 latest kernel.
1821 Description: Remove client patches from SLES 10 kernel.
1822 Details : This causes SLES 10 clients to behave as patchless clients
1823 even on a Lustre-patched (server) kernel.
1825 Severity : enhancement
1827 Description: self-adjustable client's lru lists
1828 Details : use adaptive algorithm for managing client cached locks lru
1829 lists according to current server load, other client's work
1830 pattern, memory activities, etc. Both, server and client
1831 side namespaces provide number of proc tunables for controlling
1834 Severity : enhancement
1836 Description: light-weight GSS support
1837 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1838 and reduce performance overhead.
1840 Severity : enhancement
1842 Description: Linux keyring support
1843 Details : Support using service of Linux keyring for Lustre GSS internal
1844 context refresh/cache mechanism.
1848 Description: Fix errors in lfs documentation
1849 Details : Fixes man pages
1853 Description: don't use GFP_* in generic Lustre code.
1854 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1858 Description: obdclass is limited by single OBD_ALLOC(idarray)
1859 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1863 Description: warnings with build patchless client with vanila 2.6.19 and up
1864 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1865 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1869 Frequency : Only for SLES
1870 Description: sanity_quota fail test_1
1871 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1872 file, which makes TSTID[2] inunique.
1876 Frequency : Only for SLES9
1877 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1878 Details : Update target/ChangeLog/which_patch .
1882 Frequency : Only for test_10 in sanity-quota.sh
1883 Description: a bug in quota test code
1884 Details : checking a test flag in a wrong place
1888 Frequency : Only for test_13 in sanity-quota.sh
1889 Description: a bug in quota test script
1890 Details : shouldn't check overfull allocated quota
1894 Description: jbd statistics
1895 Details : Port older jbd statistics patch for sles10
1897 Severity : enhancement
1899 Description: Add printing inode info into message about error in writepage.
1904 Description: reply_lock_interpret crash due to race with it and lock cancel.
1905 Details : Do not replay locks that are being cancelled. Do not reference
1906 locks by their address during replay, just by their handle.
1910 Frequency : When flocks are used.
1911 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1912 Details : Flock locks might destroy just granted lock if it could be merged
1913 with another existing flock, this is done in completion handler,
1914 so teach ldlm_cli_enquque_fini that this is a valid case for
1920 Description: Oops in read and write path when failing to allocate lock.
1921 Details : Check if lock allocation failed and return error back.
1925 Description: lstripe command fails for valid OST index
1926 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1927 instead of lov->desc.ld_active_tgt_count.
1931 Description: sanity.sh failed test 103
1932 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1937 Description: df doesn't work properly if diskfs blocksize != 4K
1938 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1942 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1943 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1945 Severity : enhancement
1947 Description: Adaptive timeouts
1948 Details : RPC timeouts adapt to changing server load and network
1949 conditions to reduce resend attempts and improve recovery time.
1953 Description: llapi_file_create() does not allow some changes
1954 Details : add llapi_file_open() that allows specifying the mode and
1955 open flags, and also returns an open file handle.
1959 Description: merge and cleanup kernel patches.
1960 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1964 Description: Add journal checksum support.(Kernel part)
1965 Details : The journal checksum feature adds two new flags i.e
1966 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1967 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1968 indicates that the commit block contains the checksum for
1969 the blocks described by the descriptor blocks. Now commit
1970 record can be sent to disk without waiting for descriptor
1971 blocks to be written to disk. This behavior is controlled
1972 using JBD2_FEATURE_ASYNC_COMMIT flag.
1976 Description: OSS needs mutliple precreate threads
1977 Details : Add ability to start more than one create thread per OSS.
1981 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1982 Details : execute lfs setstripe on client
1986 Description: mds_obd_create error creating tmp object
1987 Details : When the user sets quota on root, llog will be affected and can't
1988 create files and write files.
1991 Frequency : Always on ia64 patchless client, and possibly others.
1993 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1994 Details : This allows the patchless client to be loaded on architectures
1995 without this export.
1999 Description: RedHat Update kernel for RHEL5
2000 Details : Add the kernel config file for RHEL5.
2004 Description: RedHat Update kernel for RHEL5
2005 Details : Modify the kernel config file more closer RHEL5.
2009 Description: Build failure against Centos5 (RHEL5)
2010 Details : Define PAGE_SIZE when it isn't present.
2014 Description: test_7 on sanity-quota.sh can't be used on separate machines
2015 Details : add facet to handle it
2019 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2020 Details : with NFS, the anon dentry's parent was set to itself in
2021 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2022 corresponding dentry other than use rec->ur_name.
2024 Severity : enhancement
2026 Description: lfs setstripe enhancement
2027 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2031 Description: enable data checksumming by default
2032 Details : enable checksum by default, allow --disable-checksum
2033 configure option and "-o nochecksum" mount option. Checksums
2034 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2038 Description: lustre support for RHEL5
2039 Details : Add support for RHEL5.
2043 Description: Client eviction due to failover config
2044 Details : after a connection loss, the lustre client should attempt to
2045 reconnect to the last active server first before trying the
2046 other potential connections.
2050 Description: when mds and osts use different quota unit(32bit and 64bit),
2051 quota will be released repeatly.
2052 Details : void sending multiple quota reqs to mds, which will keep the
2053 status between the reqs.
2057 Description: osts not allocated evenly to files
2058 Details : change the condition to increase offset_idx
2061 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2063 Description: Data corruption for OSTs that are formatted larger than 2TB
2065 Details : When generating the bio request for lustre file writes the
2066 sector number would overflow a temporary variable before being
2067 used for the IO. The data reads correctly from Lustre (which
2068 will overflow in a similar manner) but other file data or
2069 filesystem metadata may be corrupted in some cases.
2073 Description: Tune the kernel for good SCSI performance.
2074 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2075 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2080 Description: rewrite ext2-derived code in obdclass/uuid.c
2081 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2082 from scratch preserving functionality.
2086 Description: "lfs find -obd UUID" prints directories
2087 Details : "lfs find -obd UUID" will return all directory names instead
2088 of just file names. It is incorrect because the directories
2089 do not reside on the OSTs.
2093 Description: use i_size_read and i_size_write in 2.6 port
2094 Details : replace inode->i_size access with i_size_read/write()
2097 Frequency : when removing large files
2099 Description: scheduling issue during removal of large Lustre files
2100 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2101 It causes scheduling issues when removing large files (17TB in the
2105 Frequency : only with liblustre clients on XT3
2107 Description: evictions taking too long
2108 Details : allow llrd to evict clients directly on OSTs
2111 Frequency : only on ppc
2113 Description: /proc/fs/lustre/devices broken on ppc
2114 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2115 We should make sure the type of 'index' is loff_t and then cast
2116 explicitly as needed below. Do not assign an explicitly cast
2120 Frequency : only for rhel5
2122 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2123 Details : Modify the target file & which_kernel.
2125 Severity : enhancement
2127 Description: omit set fsid for export NFS
2128 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2129 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2133 Frequency : Always, for kernels after 2.6.16
2134 Description: Fix warning idr_remove called for id=.. which is not allocated.
2135 Details : Last kernels save old s_dev before kill super and not allow
2136 to restore from callback - restore it before call kill_anon_super.
2140 Description: buffer overruns could theoretically occur
2141 Details : llapi_semantic_traverse() modifies the "path" argument by
2142 appending values to the end of the origin string, and a
2143 overrun may occur. Adding buffer overrun check in liblustreapi.
2147 Description: Fix error on 'ls .' at the top of the Lustre mount.
2148 Details : Don't revalidate dentry if it is a root dentry.
2152 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2153 Details : Modify vm-tunables-rhel4.patch.
2157 Description: Kernel config for 2.6.18-vanilla.
2158 Details : Modify targets/2.6-vanilla.target.in.
2159 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2160 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2161 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2162 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2166 Description: improve handling recoverable errors
2167 Details : if request processig with error which can be recoverable on server
2168 request should be resend, otherwise page released from cache and
2173 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2174 Details : Modify target file & which_patch.
2175 A flaw was found in the IA32 system call emulation provided
2176 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2177 value could be stored in the %RAX register, which could trigger an
2178 out-of-bounds system call table access. An untrusted local user
2179 could exploit this flaw to run code in the kernel
2180 (ie a root privilege escalation). (CVE-2007-4573).
2184 Description: O_DIRECT bypasses client statistics.
2185 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2186 read_ahead_stats were updated. copied the stats section from
2187 osc_send_oap_rpc() into async_internal().
2191 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2195 Description: change order of libsysio includes
2196 Details : '#include sysio.h' should always come before '#include xtio.h'
2200 Description: handle "serious error: objid * already exists" more gracefully
2201 Details : If LAST_ID value on disk is smaller than the objects existing in
2202 the O/0/d* directories, it indicates disk corruption and causes an
2203 LBUG(). If the object is 0-length, then we should use the existing
2204 object. This will help to avoid a full fsck in most cases.
2206 Severity : enhancement
2208 Description: adapt the lustre_config script to support the upgrade case
2209 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2210 server targets to 1.6 in parallel.
2214 Description: To avoid grant space > avaible space when the disk is almost
2215 full. Without this patch you might see the error "grant XXXX >
2216 available" or some LBUG about grant, when the disk is almost
2218 Details : In filter_check_grant, for non_grant cache write, we should
2219 check the left space by if (*left > ungranted + bytes), instead
2220 of (*left > ungranted), because only we are sure the left space
2221 is enough for another "bytes", then the ungrant space should be
2222 increase. In client, we should update cl_avail_grant only there
2223 is OBD_MD_FLGRANT in the reply.
2227 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2228 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2229 described in CVE-2007-4573. This problem could allow untrusted
2230 local users to gain root access.
2233 Frequency : when using O_DIRECT and quotas
2235 Description: Incorrect file ownership on O_DIRECT output files
2236 Details : block usage reported by 'lfs quota' does not take into account
2237 files that have been written with O_DIRECT.
2242 Description: touch file failed when fs is not full
2243 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2244 otherwise we can get ENOSP while fs is not full.
2248 Description: parallel lock callbacks
2249 Details : Instead of sending blocking and completion callbacks as separated
2250 requests, adding them to a set and sending in parallel.
2253 Frequency : only for Cray XT3
2254 Bugzilla : 12829/13455
2255 Description: Changing primary group doesn't change the group lustre assigns to
2257 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2258 overridden with the primary group provided by the group upcall,
2259 whereas the supplied fsgid can be trusted if it is in the list of
2260 supplementary groups returned by the group upcall.
2262 Severity : enhancement
2264 Description: Allow masking D_WARNING, D_ERROR messages from console
2265 Details : Console messages can now be disabled via lnet.printk.
2269 Description: User code with malformed file open parameter crashes client node
2270 Details : Before packing join_file req, all the related reference should be
2271 checked carefully in case some malformed flags cause fake
2272 join_file req on client.
2276 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2277 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2278 mode is equal to the granted mode. However, it should also check
2279 that LDLM_FL_CP_REQD is not set.
2283 Description: Heavy nfs access might result in deadlocks
2284 Details : After ELC code landed, it is now improper to enqueue any mds
2285 locks under och_sem, because enqueue might want to decide to
2286 cancel open locks for same inode we are holding och_sem for.
2290 Description: Client eviction while running blogbench
2291 Details : A lot of unlink operations with concurrent I/O can lead to a
2292 deadlock causing evictions. To address the problem, the number of
2293 oustanding OST_DESTROY requests is now throttled to
2294 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2295 callbacks are processed in priority.
2299 Description: enable ACLs on MDS by default
2300 Details : ACLs must be enabled on MDS by default.
2303 Frequency : PPC/PPC64 only
2305 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2306 Details : fix duplicated definitions between asm-ppc64/types.h and
2307 lustre_types.h on PPC.
2310 Frequency : PPC/PPC64 only
2312 Description: asm-ppc/segment.h does not exist
2313 Details : fix compile issue on PPC.
2317 Description: better handle error messages in extents code
2320 Frequency : RHEL4 only
2322 Description: mkfs is very slow on IA64/RHEL4
2323 Details : A performance regression has been discovered in the MPT Fusion
2324 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2325 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2326 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2328 Severity : enhancement
2330 Description: SNMP support enhancement
2331 Details : Adding total number of sampled request for an MDS node in snmp
2334 Severity : enhancement
2336 Description: Optimize ldlm waiting list processing for PR extent locks
2337 Details : When processing waiting list for read extent lock and meeting read
2338 lock that is same or wider to it that is not contended, skip
2339 processing rest of the list and immediatelly return current
2340 status of conflictness, since we are guaranteed there are no
2341 conflicting locks in the rest of the list.
2345 Description: Time out and refuse to reconnect
2346 Details : When the failover node is the primary node, it is possible
2347 to have two identical connections in imp_conn_list. We must
2348 compare not conn's pointers but NIDs, otherwise we can defeat
2349 connection throttling.
2353 Description: port llog fixes from b1_6 into HEAD
2354 Details : Port llog reference couting and some llog cleanups from b1_6
2355 (bug 10800) into HEAD, for protect from panic and access to already
2356 free llog structures.
2360 Description: Detect stride IO mode in read-ahead
2361 Details : When a client does stride read, read-ahead should detect that and
2362 read-ahead pages according to the detected stride pattern.
2366 Description: data checksumming impacts single node performance
2367 Details : add support for several checksum algorithm. Currently, only CRC32
2368 and Adler-32 are supported. The checksum type can be changed on
2369 the fly via /proc/fs/lustre/osc/*/checksum_type.
2373 Description: use adler32 for page checksums
2374 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2379 Description: build for x2 fails
2380 Details : fix compile issue on Cray systems.
2384 Description: Properly match for duplicate locks
2385 Details : Due to different lock order from skiplists code, we need to
2386 traverse entire list for now
2389 Frequency : only on PPC/SLES10
2391 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2392 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2393 be defined. Add a hack in mkfs_lustre.c to work around this compile
2398 Description: LASSERT on MDS when client holding flock lock dies
2399 Details : ldlm pool logic depends on number of granted locks equal to
2400 number of released locks which is not true for flock locks, so
2401 just exclude such locks from consideration.
2405 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2406 Details : Use fsfilt_sync() for both the whole filesystem sync and
2407 individual file sync to eliminate dangerous inode locking
2408 with I_LOCK that can lead to a deadlock.
2412 Description: performance in 1.6.3
2413 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2414 sure that 1MB requests are not fragmented by the block layer.
2416 Severity : enhancement
2418 Description: organize the server-side client stats on per-nid basis
2419 Details : Change the structure of stats under obdfilter and mds to
2427 The "uuid"s file would list the uuids of _active_ exports.
2428 And the clear entry is to clear all stats and stale nids.
2430 Severity : enhancement
2432 Description: eliminate client locks in face of contention
2433 Details : file contention detection and lockless i/o implementation
2434 for contended files.
2438 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2441 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2443 Description: MDS may lose file striping (and hence file data) in some cases
2444 Details : If there are additional extended attributes stored on the MDS,
2445 in particular ACLs, SELinux, or user attributes (if user_xattr
2446 is specified for the client mount options) then there is a risk
2447 of attribute loss. Additionally, the Lustre file striping
2448 needs to be larger than default (e.g. striped over all OSTs),
2449 and an additional attribute must be stored initially in the
2450 inode and then increase in size enough to be moved to the
2451 external attribute block (e.g. ACL growing in size) for file
2456 Description: skiplist implementation simplification
2457 Details : skiplists are used to group compatible locks on granted list
2458 that was implemented as tracking first and last lock of each
2459 lock group the patch changes that to using doubly linked lists
2463 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2464 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2465 because LASSERT on both the data supplied by a client, and the
2466 data on disk is dangerous and incorrect.
2468 Severity : enhancement
2470 Description: Slow truncate/writes to huge files at high offsets.
2471 Details : Directly associate cached pages to lock that protect those pages,
2472 this allows us to quickly find what pages to write and remove
2473 once lock callback is received.
2477 Description: more ldlm soft lockups
2478 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2479 starve other threads from the resource lock for a long time in
2480 case of long waiting queue, so change the debug level from
2481 D_OTHER to the less frequently used D_INFO.
2483 Severity : enhancement
2485 Description: add -gid, -group, -uid, -user options to lfs find
2489 Description: Hung threads in invalidate_inode_pages2_range
2490 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2491 once one is completed. As a result, some RPCs are stuck in the
2492 queue and are never sent.
2496 Description: filter threads hungs on waiting journal commit
2497 Details : Cleanup filter group llog code, then only filter group llog will
2498 be only created in the MDS/OST syncing process.
2502 Description: Procfs and llog threads access destoryed import sometimes.
2503 Details : Sync the import destoryed process with procfs and llog threads by
2504 the import refcount and semaphore.
2506 Severity : enhancement
2508 Description: openlock cache of b1_6 port to HEAD
2513 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2514 Details : If insertion of an extent fails, then discard the inode
2515 preallocation and free data blocks else it can lead to duplicate
2520 Description: don't always update ctime in ext3_xattr_set_handle()
2521 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2522 In some cases the ctime should not be updated, for example for
2523 2.0->1.8 compatibility it is necessary to delete an xattr and it
2524 should not update the ctime.
2528 Bugzilla : 15713/16362
2529 Description: Assertion in iopen_connect_dentry in 1.6.3
2530 Details : looking up an inode via iopen with the wrong generation number can
2531 populate the dcache with a disconneced dentry while the inode
2532 number is in the process of being reallocated. This causes an
2533 assertion failure in iopen since the inode's dentry list contains
2534 both a connected and disconnected dentry.
2538 Description: assertion failure in ldlm_handle2lock()
2539 Details : fix a race between class_handle_unhash() and class_handle2object()
2540 introduced in lustre 1.6.5 by bug 13622.
2545 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2546 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2551 Description: oops in sd_iostats_seq_show()
2552 Details : unloading/reloading the scsi low level driver triggers a kernel
2553 bug when trying to access the sd iostat file.
2558 Description: Kernel panics during QLogic driver reload
2559 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2560 patch, causing memory corruption.
2565 Description: journal_dev option does not work in b1_6
2566 Details : pass mount option during pre-mount.
2568 Severity : enhancement
2570 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2571 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2572 extent information of a file. It can be used to map logical blocks
2573 in a file to physical blocks in the block device.
2577 Description: LDLM soft lockups - improvement
2578 Details : It is be possible to send the lock handle along with each read
2579 or write request because the client is already doing a lock match
2580 itself so there isn't any reason the OST should have to re-do that
2584 Frequency : only on Cray X2
2586 Description: X2 build failures
2587 Details : fix build failures on Cray X2.
2591 Description: xid & resent requests
2592 Details : Initialize RPC XID from clock at startup (randomly if clock is
2595 Severity : enhancement
2597 Description: Add lustre_start utility to start or stop multiple Lustre servers
2602 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2603 Details : In case of memory pressure, list_del() can be called twice on
2604 req->rq_history_list, causing a kernel oops.
2608 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2609 Details : fix stack overflow in the distributed lock manager by defering
2610 export eviction after a failed AST to the elt thread instead of
2611 handling it in the dlm interpret routine.
2615 Description: Convert some comments to new format.
2616 Details : Update documenting comments to match doxygen conventions.
2620 Description: Grammar fixes.
2621 Details : A couple of trivial spelling fixes.
2625 Description: OSD_COUNTERS-mandatory
2626 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2627 to simplify the code.
2631 Description: simplify cmm_device freeing logic.
2632 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2636 Description: Add lockdep support to dt_object_operations locking interface.
2637 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2638 parameter indicating lock ordering. Update mdd code to use new
2643 Description: Introduce failloc constants for lockless IO tests.
2644 Details : Add two new failloc constants to test lockless IO. Only one of
2645 them in implemented---another is checked in yet to be landed
2650 Description: Add lockdep support for inode mutex.
2651 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2652 in the situations where Lustre has to lock more than one inode
2657 Description: Add optional invariants checking support.
2658 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2659 configure switch. This macro is to be used for consistency and
2660 sanity checks that are too expensive to be left in `production'
2665 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2666 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2671 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2672 Details : Use spin_lock_nested() in (the only) situation where more than
2673 one ldlm_lock is locked simultaneously. Also, fix possible
2674 dead-lock in ldlm_lock_change_resource() by enforcing particular
2679 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2680 Details : Instead of specifying each ldlm_lock call-back through separate
2681 parameter, wrap them into struct ldlm_callback_suite.
2685 Description: Kill join_lru obd method and its callers.
2686 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2687 in memory---a requirement implemented through ->o_join_lru() obd
2688 method in HEAD. Remove this method and its users.
2692 Description: Add asynchronous ldlm ENQUEUE completion handler.
2693 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2694 case that stock ldlm_completion_ast() cannot handle as it waits
2695 until lock is granted. Introduce new ldlm_completion_ast_async()
2696 for this. Also comment ldlm_completion_ast().
2700 Description: ldlm_error <-> errno conversion.
2701 Details : Add functions to map (rather arbitrary) between LDLM error codes
2702 and standard errno values. CLIO needs this to prevent LDLM specific
2703 constants from escaping ldlm and osc.
2707 Description: Kill unused ldlm_handle2lock_ns() function.
2708 Details : Kill unused ldlm_handle2lock_ns() function.
2712 Description: Add lu_ref support to ldlm_lock
2713 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2714 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2715 and ldlm_resource. LDLM interface has to be changed, because code that
2716 releases a reference on a lock, has to "know" what reference this is.
2717 In the most frequent case
2719 lock = ldlm_handle2lock(handle);
2721 LDLM_LOCK_PUT(lock);
2723 no changes are required. When any other reference (received _not_ from
2724 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2725 instead of LDLM_LOCK_PUT().
2727 Arguably, changes are pervasive, and interface requires some discipline
2728 for proper use. On the other hand, it was very instrumental in finding
2729 a few leaked lock references.
2733 Description: Add ldlm_lock_addref_try().
2734 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2735 attempts to addref a lock that might be being canceled
2740 Description: Add ldlm_weigh_callback().
2741 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2742 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2743 hard-coded `number of pages' logic.
2747 Description: Add lockdep annotations to llog code.
2748 Details : Use appropriately tagged _nested() locking calls in the places
2749 where llog takes more than one ->lgh_lock lock.
2753 Description: Add loi_kms_set().
2754 Details : Wrap kms updates into a helper function.
2758 Description: Constify instances of struct lsm_operations.
2759 Details : Constify instances of struct lsm_operations.
2763 Description: lu_conf support.
2764 Details : On a server, a file system object is uniquely identified
2765 by a fid, which is sufficient to locate and load all object
2766 state (inode). On a client, on the other hand, more data are
2767 necessary instantiate an object. Change lu_object_find() and
2768 friends to take additional `lu_conf' argument describing object.
2769 Typically this includes layout information.
2773 Description: lu_context fixes.
2774 Details : Introduce new lu_context functions that are needed on the client
2775 side, where some system threads (ptlrpcd) are shared by multiple
2776 modules, and so cannot be stopped during module shutdown.
2780 Description: Add start and stop methods to lu_device_type_operations.
2781 Details : Introduce two new methods in lu_device_type_operations, that are
2782 invoked when first instance of a given type is created and last one
2783 is destroyed respectively. This is need by CLIO.
2787 Description: Add lu_ref support to struct lu_device.
2788 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2789 track leaked references.
2793 Description: Introduce lu_kmem_descr.
2794 Details : lu_kmem_descr and its companion interface allow to create
2795 and destroy a number of kmem caches at once.
2799 Description: Fix lu_object finalization race.
2800 Details : Fix a race between lu_object_find() finding an object and its
2801 concurrent finalization. This race is (most likely) not possible
2802 on the server, but might happen on the client.
2806 Description: Introduce lu_ref interface.
2807 Details : lu_ref is a debugging module allowing to track references to
2808 a given object. It is quite cpu expensive, and has to be
2809 explicitly enabled with --enable-lu_ref. See usage description
2814 Description: Factor lu_site procfs stats into a separate function.
2815 Details : Separate lu_site stats printing code into a separate function
2816 that can be reused on a client.
2820 Description: Constify instances of struct {lu,dt,md}_device_operations.
2821 Details : Constify instances of struct {lu,dt,md}_device_operations.
2825 Description: Introduce struct md_site and move meta-data specific parts of
2826 struct lu_site here.
2827 Details : Move md-specific fields out of struct lu_site into special struct
2828 md_site, so that lu_site can be used on a client.
2832 Description: Kill mdd_lov_destroy().
2833 Details : Remove unused mdd code.
2837 Description: Add st_block checking to multistat.c.
2838 Details : Add st_block checking to multistat.c.
2842 Description: Add lu_ref support to struct obd_device.
2843 Details : Add lu_ref tracking to obd_device.
2847 Description: Kill obd_set_fail_loc().
2848 Details : Remove unused code.
2852 Description: Add special type for ptlrpc_request interpret functions.
2853 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2854 there. Actual usage will be in CLIO.
2858 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2859 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2860 doesn't work with lockdep.
2864 Description: Add rwv.c test program.
2865 Details : New testing program exercising readv(2) and writev(2) (Qian).
2869 Description: Add sendfile.c test program.
2870 Details : New testing program exercising sendfile(2) (Jay).
2874 Description: Ratelimit a message that can be very frequent.
2875 Details : Ratelimit a memory allocation failure message that can
2880 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2881 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2885 Description: Liblustre build fixes.
2886 Details : Liblustre build fixes.
2890 Description: libcfs: add cfs_{need,cond}_resched() interface.
2891 Details : libcfs: add cfs_{need,cond}_resched() definition and
2892 implementations for Linux, NT, and liblustre.
2894 Severity : enhancement
2896 Description: More exported tunables for mballoc
2897 Details : Add support for tunable preallocation window and new tunables for
2898 large/small requests
2902 Description: Detect corruption of block bitmap and checking for preallocations
2903 Details : Checks validity of on-disk block bitmap. Also it does better
2904 checking of number of applied preallocations. When corruption is
2905 found, it turns filesystem readonly to prevent further corruptions.
2909 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2910 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2911 window to 0. If it does hit cache miss, and read-pattern is still
2912 stride-io mode, does not reset the stride window, but also does not
2913 increase the stride window length in this case.
2917 Frequency : only for big-endian servers
2918 Description: Check if system is big-endian while mounting fs with extents feature
2919 Details : Mounting a filesystem with extents feature will fail on big-endian
2920 systems since ext3-based ldiskfs is not supported on big-endian
2921 systems. This can be over-riden with "bigendian_extents" mount option.
2923 Severity : enhancement
2925 Description: The root squash functionality
2926 Details : A security feature, which is to prevent users from being able
2927 to mount lustre on their desktop, run as root, and delete
2928 all of the files in the filesystem. The goal is accomplished by
2929 remapping user id (UID) and group id (GID) of the root user to
2930 a UID and GID specified by the system administartor via Lustre
2931 configuration management server (MGS). The functionality also
2932 allows to specify sets of clients for which the remapping does
2937 Description: Excessive recovery window
2938 Details : With AT enabled, the recovery window can be excessively long (6000+
2939 seconds). To address this problem, we no longer use
2940 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2941 timeout no longer depends on the service time, it is set to
2942 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2943 time via pb_service_time.
2947 Description: Watchdog triggered on MDS failover
2948 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2949 the OSTs know that the MDS "UUID" can be reused for the same export
2950 from a different NID, so we do not need to wait for the export to be
2954 Frequency : rare, only if using MMP with Linux RAID
2956 Description: MMP doesn't work with Linux RAID
2957 Details : While using HA for Lustre servers with Linux RAID, it is possible
2958 that MMP will not detect multiple mounts. To make this work we
2959 need to unplug the device queue in RAID when the MMP block is being
2960 written. Also while reading the MMP block, we should read it from
2961 disk and not the cached one.
2963 Severity : enhancement
2965 Description: open file using fid
2966 Details : A file can be opened using just its fid, like
2967 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2970 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2973 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2974 Details : add proper locking for creation of jbd_4k slab cache
2978 Description: MMP check in ext3_remount() fails without displaying any error
2979 Details : When multiple mount protection fails during remount, proper error
2982 Severity : enhancement
2984 Description: Allow stripe size to be up to 4G-64k
2985 Details : Fix math logic to allow large stripe sizes.
2989 Description: add check for >8TB ldiskfs filesystems
2990 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2991 Don't allow >8TB ldiskfs filesystems to be mounted without
2992 force_over_8tb mount option
2997 Descriptoin: OSS complains "too many blocks" and crashes
2998 Details : In filter_commitrw_write():
2999 1. lquota_chkquota();
3000 2. obdo_from_inode();
3001 3. lquota_pending_commit();
3002 "2." has changed uid and gid in this case, then caused
3003 lquota_chkquota() and lquota_pending_commit() to use
3004 different uid and gid. That is the root of the bug.
3008 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3009 Details : Detect and print out the 'stripe count = -1' case properly.
3013 Description: Out or order replies might be lost on replay
3014 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3015 request with tid smaller than one currently being added, add it
3016 to the start, not end of the list.
3018 --------------------------------------------------------------------------------
3020 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3022 * Support for kernels:
3023 2.6.5-7.283 (SLES 9),
3024 2.6.9-55.EL (RHEL 4),
3025 2.6.16.46-0.14 (SLES 10),
3026 2.6.18.8 vanilla (kernel.org)
3027 * Client support for unpatched kernels:
3028 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3029 2.6.16 - 2.6.22 vanilla (kernel.org)
3030 * Due to recently discovered recovery problems, we do not recommend
3031 using patchless RHEL 4 clients with this or any earlier release.
3032 * Recommended e2fsprogs version: 1.39.cfs8
3033 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3034 * Starting with this release, the ldiskfs backing filesystem required
3035 by Lustre is now in its own package, lustre-ldiskfs. This package
3036 should be installed. It is versioned separately from Lustre and
3037 may be released separately in future.
3041 Description: block reactivating mgc import until all deactivates complete
3042 Details : Fix race when failing back MDT/MGS to itself (testing)
3045 Frequency : at statup only
3047 Description: mds_lov_synchronize race leads to various problems
3048 Details : simultaneous MDT->OST connections at startup can cause the
3049 sync to abort, leaving the OSC in a bad state.
3051 Severity : enhancement
3053 Description: add optional extra BUILD_VERSION info
3054 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3055 to override the lustre version.
3058 Frequency : 2.6.18 servers only
3060 Description: ll_kern_mount() doesn't release the module reference
3061 Details : The ldiskfs module reference count never drops down to 0
3062 because ll_kern_mount() doesn't release the module reference.
3067 Description: server LBUG when using old ost_num_threads parameter
3068 Details : Accept the old ost_num_threads parameter but warn that it
3069 is deprecated, and fix an off-by-one error that caused an LBUG.
3074 Description: Transient SCSI error results in persistent IO issue
3075 Details : iobuf->dr_error is not reinitialized to 0 between two
3079 Frequency : sometimes when underlying device returns I/O errors
3081 Description: OSTs not going read-only during write failures
3082 Details : OSTs are not remounted read-only when the journal commit threads
3083 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3084 instead of the ext3 wrappers.
3088 Description: poor connect scaling with increasing client count
3089 Details : Don't run filter_grant_sanity_check for more than 100 exports
3090 to improve scaling for large numbers of clients.
3093 Frequency : SLES10 only
3095 Description: sanity-quota.sh quotacheck failed: rc = -22
3096 Details : Quotas cannot be enabled on SLES10.
3099 Frequency : liblustre clients only
3101 Description: getdirentries does not give error when run on compute nodes
3102 Details : getdirentries does not fail when the size specified as an argument
3103 is too small to contain at least one entry
3105 Severity : enhancement
3107 Description: Add LNET router traceability for debug purposes
3108 Details : If a checksum failure occurs with a router as part of the
3109 IO path, the NID of the last router that forwarded the bulk data
3110 is printed so it can be identified.
3115 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3116 Details : Due to a race condition, liblustre clients were occasionally
3117 evicted incorrectly.
3119 Severity : enhancement
3121 Description: lfs setstripe use optional parameters instead of postional
3124 Severity : enhancement
3126 Description: Nanosecond timestamp support for ldiskfs
3127 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3128 resolution timestamps. There is not yet support for this at
3129 the Lustre filesystem level.
3132 Frequency : during server recovery
3134 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3135 Details : request with rq_no_resend flag not awake l_wait_event if they get
3139 Frequency : nfs export on patchless client
3141 Description: connectathon hang when test nfs export over patchless client
3142 Details : Disconnected dentry cannot be found with lookup, so we do not need
3143 to unhash it or make it invalid
3146 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3147 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3148 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3153 Description: open req refcounting wrong on reconnect
3154 Details : If reconnect happened between getting open reply from server and
3155 call to mdc_set_replay_data in ll_file_open, we will schedule
3156 replay for unreferenced request that we are about to free.
3157 Subsequent close will crash in variety of ways.
3158 Check that request is still eligible for replay in
3159 mdc_set_replay_data().
3164 Description: disable writes to filesystem when reading health_check file
3165 Details : the default for reading the health_check proc file has changed
3166 to NOT do a journal transaction and write to disk, because this
3167 can cause reads of the /proc file to hang and block HA state
3168 checking on a healthy but otherwise heavily loaded system. It
3169 is possible to return to the previous behaviour during configure
3170 with --enable-health-write.
3172 Severity : enhancement
3174 Description: 64-bit inode version
3175 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3176 to the inode. This will be required for version-based recovery.
3181 Description: MDS fails to start if a duplicate client export is detected
3182 Details : in some rare cases it was possible for a client to connect to
3183 an MDS multiple times. Upon recovery the MDS would detect this
3184 and fail during startup. Handle this more gracefully.
3186 Severity : enhancement
3188 Description: Add -o localflock option to simulate old noflock
3190 Details : This will achieve local-only flock/fcntl locks
3196 Description: log_commit_thread vs filter_destroy race leads to crash
3197 Details : Take import reference before releasing llog record semaphore
3202 Description: Wrong request locking in request set processing
3203 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3204 imp_delayed_list, in this place should be used imp_lock.
3207 Frequency : when reconnection
3209 Description: Grant Leak when osc reconnect to OST
3210 Details : When osc reconnect ost, OST(filter) should check whether it
3211 should grant more space to client by comparing fed_grant and
3212 cl_avail_grant, and return the granted space to client instead
3213 of "new granted" space, because client will call osc_init_grant
3214 to update the client grant space info.
3217 Frequency : when client reconnect to OST
3219 Description: Grant Leak when osc do resend and replay bulk write
3220 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3221 bulk write request, because the grant info while be sync between
3222 OSC and OST when reconnect, and we should ignore the grant info
3223 these of resend/replay write req.
3228 Description: Grant space more than available left space sometimes.
3229 Details : When then OST is about to be full, if two bulk writing from
3230 different clients came to OST. Accord the avaliable space of the
3231 OST, the first req should be permitted, and the second one
3232 should be denied by ENOSPC. But if the seconde arrived before
3233 the first one is committed. The OST might wrongly permit second
3234 writing, which will cause grant space > available space.
3237 Frequency : when client is evicted
3239 Description: Grant might be wrongly erased when osc is evicted by OST
3240 Details : when the import is evicted by server, it will fork another
3241 thread ptlrpc_invalidate_import_thread to invalidate the
3242 import, where the grant will be set to 0. While the original
3243 thread will update the grant it got when connecting. So if
3244 the former happened latter, the grant will be wrongly errased
3245 because of this race.
3250 Description: Checking Stale with correct fid
3251 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3252 to check whether it is stale, because sometimes, we want the
3253 enqueue happened anyway, and op_data.fid2 will not be initialized.
3255 Severity : enhancement
3257 Description: update patchless client
3258 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3261 Frequency : only with 2.4 kernel
3263 Description: random memory corruption
3264 Details : size of struct ll_inode_info is to big for union inode.u and this
3265 can be cause of random memory corruption.
3270 Description: Memory leak in recovery
3271 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3272 It should also check obd_fail before fsfilt_start, otherwise if
3273 fsfilt_start return -EROFS,(failover mds during mds recovery).
3274 then the req will return with repmsg->transno = 0 and rc = EROFS.
3275 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3276 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3277 be freed no matter whether fsfilt_commit success or not.
3280 Frequency : only with huge count clients
3282 Description: Prevents from taking the superblock lock in llap_from_page for
3284 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3285 for llap_from_page call in ll_removepage prevents from taking the
3286 superblock lock for a soon died page.
3291 Description: Not check open intent error before release open handle
3292 Details : in some rare cases, the open intent error is not checked before
3293 release open handle, which may cause
3294 ASSERTION(open_req->rq_transno != 0), because it tries to release
3295 the failed open handle.
3300 Description: Set cat log bitmap only after create log success.
3301 Details : in some rare cases, the cat log bitmap is set too early. and it
3302 should be set only after create log success.
3306 Description: Accessing a block bevice can re-enable I/O when Lustre is
3307 tearing down a device.
3308 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3313 Description: service threads may hog cpus when there are a lot of requests
3314 Details : Insert cond_resched to give other threads a chance to use some CPU
3319 Description: the cat log was not initialized in recovery
3320 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3321 unlink log on mds will not be initialized until mds post
3322 recovery. And also in mds post recovery, the unlink log will
3323 initialization will be done asynchronausly, so there will be race
3324 between add unlink log and unlink log initialization.
3328 Description: brw_stats were being printed incorrectly
3329 Details : brw_stats were being printed as log2 but all of them were not
3330 recorded as log2. Also remove some code duplication arising from
3331 filter_tally_{read,write}.
3335 Frequency : rare, only in recovery.
3336 Description: ASSERTION(req->rq_type != LI_POISON) failed
3337 Details : imp_lock should be held while iterating over imp_sending_list for
3338 prevent destroy request after get timeout in ptlrpc_queue_wait.
3342 Description: replay-single.sh test 52 fails
3343 Details : A lock's skiplist need to be cleanup when it being unlinked
3344 from its resource list.
3348 Description: Short directio read returns full requested size rather than
3350 Details : Direct I/O operations should return actual amount of bytes
3351 transferred rather than requested size.
3353 Severity : enhancement
3355 Description: metadata RPC reduction (e.g. for rm performance)
3356 Details : decrease the amount of synchronous RPC between clients and servers
3357 by canceling conflicing lock before the operation on the client
3358 and packing thier handles into the main operation RPC to server.
3360 Severity : enhancement
3362 Description: Async OSC create to avoid the blocking unnecessarily.
3363 Details : If an OST has no remaining object, system will block on the
3364 creation when it needs to create a new object on this OST. Now,
3365 ways use pre-created objects when available, instead of blocking on
3366 an empty osc while others are not empty. If we must block, we
3367 block for the shortest possible period of time.
3371 Description: improve handling recoverable errors
3372 Details : if request processig with error which can be recoverable on server
3373 request should be resend, otherwise page released from cache and
3376 Severity : enhancement
3378 Description: refine locking for avoid write wrong info into lov_objid file
3379 Details : fix possible races with add new target and write/update data in
3382 --------------------------------------------------------------------------------
3384 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3389 Frequency : on some architectures
3391 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3392 Details : Uninitialized flags sometimes cause configuration commands to
3396 Frequency : patchless clients only
3398 Description: missing __iget() symbol export
3399 Details : The __iget() symbol export is missing. To avoid the need for
3400 this on patchless clients the deathrow inode reaper is turned
3401 off, and we depend on the VM to clean up old inodes. This
3402 dependency was during via the fix for bug 12181.
3404 --------------------------------------------------------------------------------
3406 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3408 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3409 INTEROPERATE with older versions automatically. In many cases a
3410 special upgrade step is needed. Please read the
3411 user documentation before upgrading any part of a live system.
3412 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3413 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3414 * WARNING: Lustre configuration and startup changes are required with
3415 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3419 Severity : enhancement
3421 Description: MountConf
3422 Details : Lustre configuration is now managed via mkfs and mount
3423 commands instead of lmc and lconf. New obd types (MGS, MGC)
3424 are added for dynamic configuration management. See
3425 https://mail.clusterfs.com/wikis/lustre/MountConf for
3428 Severity : enhancement
3430 Description: dynamic OST addition
3431 Details : OSTs can now be added to a live filesystem
3433 Severity : enhancement
3435 Description: startup order invariance
3436 Details : MDTs and OSTs can be started in any order. Clients only
3437 require the MDT to complete startup.
3439 Severity : enhancement
3441 Description: parallel, asynchronous orphan cleanup
3442 Details : orphan cleanup is now performed in separate threads for each
3443 OST, allowing parallel non-blocking operation.
3445 Severity : enhancement
3447 Description: optimized stripe assignment
3448 Details : stripe assignments are now made based on ost space available,
3449 ost previous usage, and OSS previous usage, in order to try
3450 to optimize storage space and networking resources.
3452 Severity : enhancement
3454 Description: Permanently set tunables
3455 Details : All writable /proc/fs/lustre tunables can now be permanently
3456 set on a per-server basis, at mkfs time or on a live system.
3458 Severity : enhancement
3460 Description: Lustre message v2
3461 Details : Add lustre message format v2.
3463 Severity : enhancement
3465 Description: client OST exclusion list
3466 Details : Clients can be started with a list of OSTs that should be
3467 declared "inactive" for known non-responsive OSTs.
3471 Description: SPEC SFS validation failure on NFS v2 over lustre.
3472 Details : Changes the blocksize for regular files to be 2x RPC size,
3473 and not depend on stripe size.
3475 Severity : enhancement
3477 Description: Multiple MD RPCs in flight.
3478 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3479 To avoid overly-overloading MDS, introduce a limit on number of
3480 MDS RPCs in flight for a single client and add /proc controls
3481 to adjust this limit.
3483 Severity : enhancement
3485 Description: client read/write statistics
3486 Details : Add client read/write call usage stats for performance
3487 analysis of user processes.
3488 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3489 file access. extents_stats shows chunk size distribution.
3490 extents_stats_per_process show chunk size distribution per
3493 Severity : enhancement
3495 Description: mds statistics
3496 Details : Add detailed mds operations statistics in
3497 /proc/fs/lustre/mds/*/stats.
3501 Description: Failure of copying files with lustre special EAs.
3502 Details : Client side always return success for setxattr call for lustre
3503 special xattr (currently only "trusted.lov").
3508 Description: Refcount LNET uuids
3509 Details : The global LNET uuid list grew linearly with every startup;
3510 refcount repeated list entries instead of always adding to
3513 Severity : enhancement
3515 Description: Dynamic service threads
3516 Details : Within a small range, start extra service threads
3517 automatically when the request queue builds up.
3520 Frequency : mixed-endian client/server environments
3522 Description: mixed-endian crashes
3523 Details : The new msg_v2 system had some failures in mixed-endian
3526 Severity : enhancement
3528 Description: Easy OST removal
3529 Details : OSTs can be permanently deactivated with e.g. 'lctl
3530 conf_param lustre-OST0001.osc.active=0'
3532 Severity : enhancement
3534 Description: MGS proc entries
3535 Details : Added basic proc entries for the MGS showing what filesystems
3538 Severity : enhancement
3540 Description: provide MGS failover
3541 Details : Added config lock reacquisition after MGS server failover.
3543 Severity : enhancement
3545 Description: add Linux 2.4 support
3546 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3550 Description: a large application tries to do I/O to the same resource and dies
3551 in the middle of it.
3552 Details : Check the req->rq_arrival time after the call to
3553 ost_brw_lock_get(), but before we do anything about
3554 processing it & sending the BULK transfer request. This
3555 should help move old stale pending locks off the queue as
3556 quickly as obd_timeout.
3559 Frequency : when an incorrect nid is specified during startup
3561 Description: ptlrpc connect to non-existant node causes kernel crash
3562 Details : LNET can't be re-entered from an event callback, which
3563 happened when we expire a message after the export has been
3564 cleaned up. Instead, hand the zombie cleanup off to another
3567 Severity : enhancement
3569 Description: plain/inodebits lock performance improvement
3570 Details : Grouping plain/inodebits in granted list by their request modes
3571 and bits policy, thus improving the performance of search through
3575 Frequency : only if OST filesystem is corrupted
3577 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3578 Details : for a short time RPCs with bulk IO are in the replay list,
3579 but replay of bulk IOs is unimplemented. If the OST filesystem
3580 is corrupted due to disk cache incoherency and then replay is
3581 started it is possible to trip an assertion. Avoid putting
3582 committed RPCs into the replay list at all to avoid this issue.
3585 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3587 Description: System hang on startup
3588 Details : This bug allowed the liblustre (e.g. catamount) client to
3589 return to the app before handling all startup RPCs. This
3590 could leave the node unresponsive to lustre network traffic
3591 and manifested as a server ptllnd timeout.
3593 Severity : enhancement
3595 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3596 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3597 boolean to control whether to print peer debug info when a
3598 client's RPC times out.
3601 Frequency : only for kernels with patches from Lustre below 1.4.3
3603 Description: Remove old rdonly API
3604 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3607 Frequency : only for devices with external journals
3609 Description: Set external device read-only also
3610 Details : During a commanded failover stop, we set the disk device
3611 read-only while the server shuts down. We now also set any
3612 external journal device read-only at the same time.
3615 Frequency : when upgrading from 1.4 while trying to change parameters
3617 Description: The wrong (new) MDC name was used when setting parameters for
3618 upgraded MDT's. Also allows changing of OSC (and MDC)
3619 parameters if --writeconf is specified at tunefs upgrade time.
3622 Frequency : when setting specific ost indicies
3624 Description: QOS code breaks on skipped indicies
3625 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3626 created with --index need not be sequential.
3628 Severity : enhancement
3630 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3631 Details : The uninit_groups feature works in conjunction with the kernel
3632 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3633 up the pass1 processing of e2fsck. This is a read-only feature
3634 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3635 mount filesystems that have had this feature enabled.
3637 Severity : enhancement
3639 Description: Improve multi-block allocation algorithm to avoid fragmentation
3640 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3641 allocation locality and avoid filesystem fragmentation.
3643 ------------------------------------------------------------------------------
3645 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3647 * Support for kernels:
3648 2.4.21-47.0.1.EL (RHEL 3)
3649 2.6.5-7.283 (SLES 9)
3650 2.6.9-42.0.10.EL (RHEL 4)
3651 2.6.12.6 vanilla (kernel.org)
3652 2.6.16.27-0.9 (SLES 10)
3653 * Recommended e2fsprogs version: 1.39.cfs5
3655 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3659 Frequency : occasional, depends on client load and configuration
3660 Bugzilla : 12181, 12203
3661 Description: data loss for recently-modified files
3663 Details : In some cases it is possible that recently written or created
3664 files may not be written to disk in a timely manner (this should
3665 normally be within 30s unless client IO load is very high).
3666 The problem appears as zero-length files or files that are a
3667 multiple of 1MB in size after a client crash or client eviction
3668 that are missing data at the end of the file.
3670 This problem is more likely to be hit on clients where files are
3671 repeatedly created and unlinked in the same directory, clients
3672 have a large amount of RAM, have many CPUs, the filesystem has
3673 many OSTs, the clients are rebooted frequently, and/or the files
3674 are not accessed by other nodes after being written.
3676 The presence of the problem can be detected by looking at
3677 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3678 smaller than the second (nr_unused) then dirty files will not
3679 be flushed automatically to disk. "sync; sleep 10" should be
3680 run several times on the node before unmounting it to update
3681 Lustre (this is also safe to run on nodes without this problem).
3683 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3684 that can cause this same problem, so customers using that kernel
3685 also need to update the kernel in addition to Lustre. In order
3686 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3688 It is normal that files written just before a client crash (less
3689 than 30s) may not yet have been flushed to disk, even for local
3693 Frequency : frequent on thin XT3 nodes
3695 Description: UUID collision on thin XT3 Linux nodes
3696 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3697 randomly, since we relied on an insufficiently-seeded PRNG.
3702 Description: OSS hangs after "All ost request buffers busy"
3703 Details : A deadlock between quota and journal operations caused OSS
3704 hangs after printing "All ost request buffers busy."
3707 Frequency : always on liblustre builds
3709 Description: Cleanup compiler warnings on liblustre
3712 Frequency : always on liblustre builds on XT3
3714 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3718 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3719 > 32000 subdirectories
3720 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3721 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3722 subdirectory count crosses 32000. This will aid e2fsck to
3723 correctly handle more than 32000 subdirectories.
3726 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3728 Description: System hang on startup
3729 Details : This bug allowed the liblustre (e.g. catamount) client to
3730 return to the app before handling all startup RPCs. This
3731 could leave the node unresponsive to lustre network traffic
3732 and manifested as a server ptllnd timeout.
3734 Severity : enhancement
3736 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3737 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3738 boolean to control whether to print peer debug info when a
3739 client's RPC times out.
3744 Description: make O_SYNC working on 2.6 kernels
3745 Details : 2.6 kernels use different method for mark pages for write,
3746 so need add a code to lustre for O_SYNC work.
3751 Description: Failure to close file and release space on NFS
3752 Details : Put inode details into lock acquired in ll_intent_file_open.
3753 Use mdc_intent_lock in ll_intent_open to properly
3754 detect all kind of errors unhandled by mdc_enqueue.
3759 Description: proc file read during shutdown sometimes raced obd removal,
3761 Details : Add lock to prevent obd access after proc file removal.
3764 Frequency : Only for files larger than 4GB on 32-bit clients.
3766 Description: improperly doing page alignment of locks
3767 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3768 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3773 Description: RPCs being resent when they shouldn't be
3774 Details : Some RPCs that should not be resent are being resent. This
3775 can cause inconsistencies in the RPC state machine. Do not
3776 resend such requests.
3779 Frequency : rare, only with NFS export
3781 Description: Crash on NFS re-export node
3782 Details : under very unusual load conditions an assertion is hit in
3783 ll_intent_file_open()
3786 Frequency : only if OST filesystem is corrupted
3788 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3789 Details : for a short time RPCs with bulk IO are in the replay list,
3790 but replay of bulk IOs is unimplemented. If the OST filesystem
3791 is corrupted due to disk cache incoherency and then replay is
3792 started it is possible to trip an assertion. Avoid putting
3793 committed RPCs into the replay list at all to avoid this issue.
3798 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3799 Details : Large single O_DIRECT read and write calls can fail to allocate
3800 a sufficiently large buffer to process the request. In case of
3801 allocation failure the allocation is retried with a smaller
3802 buffer and broken into smaller requests.
3804 Severity : enhancement
3806 Description: Add -o localflock option to simulate old noflock behaviour.
3807 Details : This will achieve local-only flock/fcntl locks coherentness.
3812 Description: versioning check is incomplete
3813 Details : Checking the version difference of client vs. server, report
3814 error if the gap is too big.
3819 Description: add support PG_writeback bit
3820 Details : add support for PG_writeback bit for Lustre, for more carefull
3821 work with page cache in 2.6 kernel. This also fix some deadlocks
3822 and remove hack for work O_SYNC with 2.6 kernel.
3824 Severity : enhancement
3826 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3827 Details : The uninit_groups feature works in conjunction with the kernel
3828 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3829 up the pass1 processing of e2fsck. This is a read-only feature
3830 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3831 mount filesystems that have had this feature enabled.
3833 Severity : enhancement
3835 Description: Improve multi-block allocation algorithm to avoid fragmentation
3836 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3837 allocation locality and avoid filesystem fragmentation.
3839 ------------------------------------------------------------------------------
3841 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3843 * Support for kernels:
3844 2.6.9-42.0.3.EL (RHEL 4)
3845 2.6.5-7.276 (SLES 9)
3846 2.4.21-47.0.1.EL (RHEL 3)
3847 2.6.12.6 vanilla (kernel.org)
3848 2.6.16.21-0.8 (SLES10)
3849 * Recommended e2fsprogs version: 1.39.cfs2-0
3851 * The backwards-compatible /proc/sys/portals symlink has been removed
3852 in this release. Before upgrading, please ensure that you change
3853 any configuration scripts or /etc/sysctl.conf files that access
3854 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3855 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3856 in advance of the upgrade on any system running Lustre 1.4.6 or
3857 newer, since /proc/sys/lnet was added in that version.
3858 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3862 Frequency : only when quota is used
3864 Description: avoid scanning export list for quota master
3865 Details : Change the algorithms to avoid scanning export list in order
3866 to improve the efficiency.
3869 Frequency : MDS failover only, very rarely
3871 Description: "went back in time" messages on mds failover
3872 Details : The greatest transno may be lost when the current operation
3873 finishes with an error (transno==0) and the client's last_rcvd
3874 record is over-written. Save the greatest transno in the
3875 mds_last_transno for this case.
3878 Frequency : always for specific kernels and striping counts
3880 Description: client may get "Matching packet too big" without ACL support
3881 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3882 when trying to access files in certain configurations. The
3883 clients should in fact be denied when mounting because they do
3884 not understand ACLs.
3887 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3889 Description: many clients connecting with IO in progress causes connect timeouts
3890 Details : Avoid synchronous journal commits to avoid delays caused by many
3891 clients connecting/disconnecting when bulk IO is in progress.
3892 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3893 OST_IO_PORTAL to avoid delays behind potentially many pending
3897 Frequency : occasionally with multiple writers to a single file
3899 Description: shared writes to file may result in wrong size reported by stat()
3900 Details : Allow growing of kms when extent lock is cancelled
3903 Frequency : always with random mmap IO to multi-striped file
3905 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3906 Details : Only if the hole is at the end of OST object so that kms is too
3907 small. Fix is to increase kms accordingly in ll_nopage.
3910 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3912 Description: writes to a missing object would leak memory on the OST
3913 Details : If there is an inconsistency between the MDS and OST filesystems,
3914 such that the MDS references an object that doesn't exist, writes
3915 to that object will leak memory due to incorrect cleanup in the
3916 error handling path, eventually running out of memory on the OST.
3921 Description: Creating too long symlink causes lustre errors
3922 Details : Check symlink and name lengths before sending requests to MDS.
3925 Frequency : only if flock is enabled (not on by default)
3927 Description: posix locks not released on fd closure on 2.6.9+
3928 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3929 kernels, this caused such locks not to be released on fd close and
3930 then assertions on fs unmount about still used locks.
3933 Frequency : MDS failover only, very rarely
3935 Description: clients may get ASSERTION(granted_lock != NULL)
3936 Details : When request was taking a long time, and a client was resending
3937 a getattr by name lock request. The were multiple lock requests
3938 with the same client lock handle and
3939 mds_getattr_name->fixup_handle_for_resent_request found one of the
3940 lock handles but later failed with ASSERTION(granted_lock != NULL).
3945 Description: handle->h_buffer_credits > 0, assertion failure
3946 Details : h_buffer_credits is zero after truncate, causing assertion
3947 failure. This patch extends the transaction or creates a new
3951 Frequency : NFS re-export or patchless client
3952 Bugzilla : 11179, 10796
3953 Description: Crash on NFS re-export node (__d_move)
3954 Details : We do not want to hash the dentry if we don't have a lock.
3955 But if this dentry is later used in d_move, we'd hit uninitialised
3956 list head d_hash, so we just do this to init d_hash field but
3957 leave dentry unhashed.
3960 Frequency : NFS re-export or patchless client
3962 Description: NFS exports has problem with symbolic link
3963 Details : lustre client didn't properly install dentry when re-exported
3964 to NFS or running patchless client.
3967 Frequency : NFS re-export or patchless client
3969 Description: Various nfs/patchless fixes.
3970 Details : fixes reuse disconected alias for lookup process - this fixes
3971 warning "find_exported_dentry: npd != pd",
3972 fix permission error with open files at nfs.
3973 fix apply umask when do revalidate.
3976 Frequency : occasional
3978 Description: Crash on NFS re-export node
3979 Details : calling clear_page() on the wrong pointer triggered oops in
3980 generic_mapping_read().
3983 Frequency : rarely, using O_DIRECT IO
3985 Description: unaligned directio crashes client with LASSERT
3986 Details : check for unaligned buffers before trying any requests.
3989 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3991 Description: stale data returned from RAID cache
3992 Details : If only a small amount of IO is done to the RAID device before
3993 reading it again it is possible to get stale data from the RAID
3994 cache instead of reading it from disk.
3997 Frequency : always for sles10 kernel
3999 Description: sles10 support
4000 Details : ll_follow_link: compile fixes and using of nd_set_link
4001 under newer kernels.
4004 Frequency : depends on arch, kernel and compiler version, always on sles10
4007 Description: recursive or deep enough symlinks cause stack overflow
4008 Details : getting rid of large stack-allocated variable in
4012 Frequency : depends on hardware
4014 Description: lustre write performance loss in the SLES10 kernel
4015 Details : the performance loss is caused by using of write barriers in the
4016 ext3 code. The SLES10 kernel turns barrier support on by
4017 default. The fix is to undo that change for ldiskfs.
4019 ------------------------------------------------------------------------------
4021 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4023 * Support for kernels:
4024 2.6.9-42.0.3EL (RHEL 4)
4025 2.6.5-7.276 (SLES 9)
4026 2.4.21-47.0.1.EL (RHEL 3)
4027 2.6.12.6 vanilla (kernel.org)
4031 Frequency : quota enabled and large files being deleted
4033 Description: releasing more than 4GB of quota at once hangs OST
4034 Details : If a user deletes more than 4GB of files on a single OST it
4035 will cause the OST to spin in an infinite loop. Release
4036 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4041 Description: statfs data retrieved from /proc may be stale or zero
4042 Details : When reading per-device statfs data from /proc, in the
4043 {kbytes,files}_{total,free,avail} files, it may appear
4044 as zero or be out of date.
4047 Frequency : systems with MD RAID1 external journal devices
4049 Description: lconf's call to blkid is confused by RAID1 journal devices
4050 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4051 of returning all block devices that match the journal UUID.
4054 Frequency : always, for aggregate stripe size over 4GB
4056 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4057 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4058 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4061 Frequency : always if "lfs find" used on a local file/directory
4063 Description: "lfs find" segfaults if used on a local file/directory
4064 Details : The case where a directory component was not specified wasn't
4065 handled correctly. Handle this properly.
4068 Frequency : always on ppc64
4070 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4071 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4072 perform endianness conversion.
4075 Frequency : rarely (truncate to non-zero file size after write under load)
4076 Bugzilla : 10730, 10687
4077 Description: Files padded with zeros to next 4K multiple
4078 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4079 it is possible that files that are truncated to a non-zero size
4080 immediately after being written are filled with zero bytes beyond
4081 the truncated size. No file data is lost.
4083 Severity : enhancement
4085 Description: Allow recovery/failover for liblustre clients.
4086 Details : liblustre clients were unaware of failover configurations until
4089 Severity : enhancement
4091 Description: user file locks should fail when not mounting with flock option
4092 Details : Set up an error-returning stub in ll_file_operations.lock field
4093 to prevent incorrect behaviour when client is mounted without
4094 flock option. Also, set up properly f_op->flock field for
4098 Frequency : always on ia64
4100 Description: "lfs df" loops on printing out MDS statfs information
4101 Details : The obd_ioctl_data was not initialized and in some systems
4102 this caused a failure during the ioctl that did not return
4103 an error. Initialize the struct and return an error on failure.
4106 Frequency : SLES 9 only
4108 Description: Error of copying files with lustre special EAs as root
4109 Details : Client side always return success for setxattr call for lustre
4110 special xattr (currently only "trusted.lov").
4113 Frequency : rarely on clusters with both ia64+i386 clients
4115 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4116 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4117 clients are doing concurrent writes to the same file it is
4118 possible that the ia64 clients may LASSERT because the OST
4119 extent locks are not PAGE_SIZE aligned. Ensure that grown
4120 locks are always aligned on the request boundary.
4123 Frequency : specific use, occasional
4125 Description: Overwriting in use executable truncates on-disk binary image
4126 Details : If one node attempts to overwrite an executable in use by
4127 another node, we now correctly return ETXTBSY instead of
4128 truncating the file.
4130 Severity : enhancement
4132 Description: Async OSC create to avoid the blocking unnecessarily.
4133 Details : If an OST has no remaining object, system will block on the creating
4134 when need to create a new object on this OST. Now, ways use
4135 pre-created objects when available, instead of blocking on an
4136 empty osc while others are not empty. If we must block, we block
4137 for the shortest possible period of time.
4142 Description: chmod on Lustre root is propagated to other clients
4143 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4144 invalid by the follow_mount time.
4147 Frequency : liblustre clients only
4149 Description: Race in 'instant cancel' lock handling could lead to such locks
4150 never to be granted in case of SMP MDS
4151 Details : Do not destroy not yet granted but cbpending locks in
4155 Frequency : replay/resend of open
4157 Description: non null lock assetion failure in mds_intent_policy
4158 Details : Trying to replay/resend lockless open requests resulted in
4159 mds_open() returning 0 with no lock. Now it sets a flag if
4160 a lock is going to be returned.
4162 Severity : enhancement
4164 Description: Checksum enhancements
4165 Details : New checksum enhancements allow for resending RPCs that failed
4168 Severity : enhancement
4170 Description: Tunables on number of dirty pages in cacche
4171 Details : Allow to set limit on number of dirty pages cached.
4176 Description: client crash on unmount - lock still has references
4177 Details : In some error handling cases it was possible to leak a lock
4178 reference on a client while accessing a file. This was not
4179 harmful to the client during operation, but would cause the
4180 client to crash when the filesystem is unmounted.
4183 Frequency : specific case, rare
4185 Description: ETXTBSY on mds though file not in use by client
4186 Details : ETXTBSY is no longer incorrectly returned when attempting to
4187 chmod or chown a directory that the user previously tried to
4188 execute or a currently-executing binary.
4191 Frequency : extremely rare except on liblustre-based clients
4193 Description: Lustre space not freed when files are deleted
4194 Details : Clean up open-unlinked files after client eviction. Previously
4195 the unlink was skipped and the files remained as orphans.
4200 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4201 Details : In some cases it was possible to send a blocking callback to
4202 a client doing a glimpse, even though that client didn't get
4203 a lock granted. When the glimpse lock is cancelled on the OST
4204 the freed lock is left on the waiting list and corrupted the list.
4207 Frequency : all core dumps
4209 Description: Broke core dumps to lustre
4210 Details : Negative dentry may be unhashed if parent does not have UPDATE
4211 lock, but some callers, e.g. do_coredump, expect dentry to be
4212 hashed after successful create, hash it in ll_create_it.
4214 ------------------------------------------------------------------------------
4216 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4218 * Support for kernels:
4219 2.6.9-42.0.2.EL (RHEL 4)
4220 2.6.5-7.276 (SLES 9)
4221 2.4.21-40.EL (RHEL 3)
4222 2.6.12.6 vanilla (kernel.org)
4226 Frequency : always on RHEL 3
4228 Description: Number of open files grows over time
4229 Details : The number of open files grows over time, whether or not
4230 Lustre is started. This was due to a filp leak introduced
4231 by one of our kernel patches.
4233 ------------------------------------------------------------------------------
4235 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4237 * Support for kernels:
4238 2.6.9-42.EL (RHEL 4)
4239 2.6.5-7.276 (SLES 9)
4240 2.4.21-40.EL (RHEL 3)
4241 2.6.12.6 vanilla (kernel.org)
4246 Bugzilla : 5719, 9635, 9792, 9684
4247 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4248 Details : If a server is under heavy load and cannot reply to new
4249 connection requests before the client resends the (re)connect,
4250 the connection handling code can behave badly if two service
4251 threads are concurrently handing separate (re)connections from
4252 the same client. Add better locking to the connection handling
4253 code, and ensure that only a single connection will be processed
4254 for a given client UUID, even if the lock is dropped.
4256 Severity : enhancement
4258 Description: add TCP zero-copy support to kernel
4259 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4260 sends if the hardware supports scatter-gather and checksumming.
4261 This allows socklnd to do client-write and server-read more
4262 efficiently and reduce CPU utilization from skbuf copying.
4265 Frequency : only if NFS exporting from client
4267 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4268 Details : If setting attributes on a file created under NFS that had
4269 never been opened it would be possible to oops the client
4270 if the file had no objects.
4273 Frequency : always for liblustre
4275 Description: liblustre client does MDS+OSTs setattr RPC for each write
4276 Details : When doing a write from a liblustre client, the client
4277 incorrectly issued an RPC to the MDS and each OST the file was
4278 striped over in order to update the timestamps. When writing
4279 with small chunks and many clients this could overwhelm the MDS
4280 with RPCs. In all cases it would slow down the write because
4281 these RPCs are unnecessary.
4283 Severity : enhancement
4285 Description: allow number of MDS service threads to be changed at module load
4286 Details : It is now possible to change the number of MDS service threads
4287 running. Adding "options mds mds_num_threads={N}" to the MDS's
4288 /etc/modprobe.conf will set the number of threads for the next
4289 time Lustre is restarted (assuming the "mds" module is also
4290 reloaded at that time). The default number of threads will
4291 stay the same, 32 for most systems.
4296 Description: OST crash if filesystem is unformatted or corrupt
4297 Details : If an OST is started on a device that has never been formatted
4298 or if the filesystem is corrupt and cannot even mount then the
4299 error handling cleanup routines would dereference a NULL pointer.
4304 Description: NULL pointer deref in llap_from_page.
4305 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4306 incorrect) mapping if the page was truncated/reclaimed while it was
4307 searched for. Check for this condition and skip such pages when
4308 doing readahead. Introduce extra check to llap_from_page() to
4309 verify page->mapping->host is non-NULL (so page is not anonymous).
4312 Frequency : Sometimes when using sys_sendfile
4314 Description: "page not covered by a lock" warnings from ll_readpage
4315 Details : sendfile called ll_readpage without right page locks present.
4316 Now we introduced ll_file_sendfile that does necessary locking
4317 around call to generic_file_sendfile() much like we do in
4321 Frequency : with certain MDS communication failures at client mount time
4323 Description: NULL pointer deref after failed client mount
4324 Details : a client connection request may delayed by the network layer
4325 and not be sent until after the PTLRPC layer has timed out the
4326 request. If the client fails the mount immediately it will try
4327 to clean up before the network times out the request. Add a
4328 reference from the request import to the obd device and delay
4329 the cleanup until the network drops the request.
4332 Frequency : occasionally during client (re)connect
4334 Description: assertion failure during client (re)connect
4335 Details : processing a client connection request may be delayed by the
4336 client or server longer than the client connect timeout. This
4337 causes the client to resend the connection request. If the
4338 original connection request is replied in this interval, the
4339 client may trip an assertion failure in ptlrpc_connect_interpret()
4340 which thought it would be the only running connect process.
4343 Frequency : only with obd_echo servers and clients that are rebooted
4345 Description: kernel BUG accessing uninitialized data structure
4346 Details : When running an obd_echo server it did not start the ping_evictor
4347 thread, and when a client was evicted an uninitialized data
4348 structure was accessed. Start the ping_evictor in the RPC
4349 service startup instead of the OBD startup.
4351 Severity : enhancement
4352 Bugzilla : 10193 (patchless)
4353 Description: Remove dependency on various unexported kernel interfaces.
4354 Details : No longer need reparent_to_init, exit_mm, exit_files,
4355 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4358 Frequency : rare (only users of deprecated and unsupported LDAP config)
4360 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4361 Details : LDAP apparently contains 'lustreName' attributes instead of
4362 'name'. A simple remapping of the name is sufficient.
4365 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4367 Description: waiting_locks_callback trips kernel BUG if client is evicted
4368 Details : Running with the dump_on_timeout debug flag turned on makes
4369 it possible that the waiting_locks_callback() can try to dump
4370 the Lustre kernel debug logs from an interrupt handler. Defer
4371 this log dumping to the expired_lock_main() thread.
4373 Severity : enhancement
4375 Description: Support NFS exporting on 2.6 kernels.
4376 Details : Implement non-rawops metadata methods for NFS server to use without
4377 changing NFS server code.
4380 Frequency : very rare (synthetic metadata workload only)
4382 Description: two racing renames might cause an MDS thread to deadlock
4383 Details : Running the "racer" program may cause one MDS thread to rename
4384 a file from being the source of a rename to being the target of
4385 a rename at exactly the same time that another thread is doing
4386 so, and the second thread has already enqueued these locks after
4387 doing a lookup of the target and is trying to relock them in
4388 order. Ensure that we don't try to re-lock the same resource.
4391 Frequency : only very large systems with liblustre clients
4393 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4394 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4395 all OSTs in parallel. This allows the eviction of stale liblustre
4396 clients to proceed much faster than if they were done in series,
4397 and also offers similar improvements for other set_info RPCs.
4402 Description: excessive CPU usage during initial read phase on client
4403 Details : During the initial read phase on a client, it would agressively
4404 retry readahead on the file, consuming too much CPU and impacting
4405 performance (since 1.4.5.8). Improve the readahead algorithm
4406 to avoid this, and also improve some other common cases (read
4407 of small files in particular, where "small" is files smaller than
4408 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4413 Description: MDS crash when receiving packet with unknown intent.
4414 Details : Do not LBUG in unknown intent case, just return -EFAULT
4416 Severity : enhancement
4417 Bugzilla : 9293, 9385
4418 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4419 Details : Do not serialize getattr (non-intent version) and statfs.
4422 Frequency : occasional, when OST network is overloaded/intermittent
4424 Description: client evicted by OST after bulk IO timeout
4425 Details : If a client sends a bulk IO request (read or write) the OST
4426 may evict the client if it is unresposive to its data GET/PUT
4427 request. This is incorrect if the network is overloaded (takes
4428 too long to transfer the RPC data) or dropped the OST GET/PUT
4429 request. There is no need to evict the client at all, since
4430 the pinger and/or lock callbacks will handle this, and the
4431 client can restart the bulk request.
4434 Frequency : Always when mmapping file with no objects
4436 Description: client crashes when mmapping file with no objects
4437 Details : Check that we actually have objects in a file before doing any
4438 operations on objects in ll_vm_open, ll_vm_close and
4444 Description: Request leak when working with deleted CWD
4445 Details : Introduce advanced request refcount tracking for requests
4446 referenced from lustre intent.
4448 Severity : Enhancement
4450 Description: Cache open file handles on client.
4451 Details : MDS now will return special lock along with openhandle, if
4452 requested and client is allowed to hold openhandle, even if unused,
4453 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4454 closing files for every read/write openration.
4456 Severity : Enhancement
4458 Description: Cache open negative dentries on client when possible.
4459 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4460 negative dentries on lock revocation.
4465 Description: Remounting a client read-only wasn't possible with a zconf mount
4466 Details : It wasn't possible to remount a client read-only with llmount.
4468 Severity : enhancement
4469 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4470 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4471 MPICH is included to add Lustre-specific ADIO interfaces.
4472 This is based closely on the UFS ADIO layer and only differs
4473 in file creation, in order to allow the OST striping to be set.
4474 This is user-contributed code and not supported by CFS.
4479 Description: extended inode attributes (immutable, append-only) work improperly
4480 when 2.4 and 2.6 kernels are used on client/server or vice versa
4481 Details : Introduce kernel-independent values for these flags.
4483 Severity : enhancement
4486 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4487 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4488 granularity. Now, Fractional MB granularity is be supported,
4489 this is very useful for low memory system.
4491 Severity : enhancement
4493 Description: Getattr by fid
4494 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4495 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4496 have locks and avoids getattr rpc requests for it on every stat.
4499 Frequency : Always, for filesystems larger than 2TB
4501 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4502 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4503 filesystems larger than 2TB to be mounted with mballoc enabled.
4506 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4508 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4509 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4510 filesystems larger than 2TB is mounted then the kernel will
4511 silently corrupt the start of the filesystem. CONFIG_LBD is
4512 enabled for all CFS-supported kernels, but the possibility of
4513 this happening with a modified kernel config exists.
4515 Severity : enhancement
4517 Description: add client O_DIRECT support for 2.6 kernels
4518 Details : It is now possible to do O_DIRECT reads and writes to files
4519 in the Lustre client mountpoint on 2.6 kernel clients.
4521 Severity : enhancement
4523 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4524 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4525 parallel, not waiting for response from every OST before sending
4526 a rpc to the next OST.
4531 Description: setattr vs write race when updating file timestamps
4532 Details : Client processes that update a file timestamp into the past
4533 right after writing to the file (e.g. tar) it is possible that
4534 the updated file modification time can be reset to the current
4535 time due to a race between processing the setattr and write RPC.
4537 Severity : enhancement
4539 Description: Bring 'lfs find' closer in line with regular Linux find.
4540 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4541 -print0 options and obtains all the needed info through the lustre
4544 Severity : enhancement
4546 Description: support up to 1024 configured devices on one node
4547 Details : change obd_dev array from statically allocated to dynamically
4548 allocated structs as they are first used to reduce memory usage
4553 Description: Flush dirty partially truncated pages during truncate
4554 Details : Immediatelly flush partially truncated pages in filter_setattr,
4555 this way we completely avoid having any pages in page cache on OST
4556 and can retire ugly workarounds during writes to flush such pages.
4561 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4562 Details : protect inode from truncation within vfs_unlink() context
4563 just take a reference before calling vfs_unlink() and release it
4564 when parent's i_sem is free.
4569 Description: last_id value checked outside lock on OST caused LASSERT failure
4570 Details : If there were multiple MDS->OST object precreate requests in
4571 flight, it was possible that the OST's last object id was checked
4572 outside a lock and incorrectly tripped an assertion. Move checks
4573 inside locks, and discard old precreate requests.
4576 Frequency : always, if extents are used on OSTs
4578 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4579 Details : OSTs using the extents format would not zero the high 16 bits of
4580 the index physical block number. This is not a problem for any
4581 OST filesystems smaller than 16TB, and no kernels support ext3
4582 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4583 new/modified files) and can be fixed for existing filesystems
4584 with e2fsprogs-1.39-cfs1.
4589 Description: import connection selection may be incorrect if timer wraps
4590 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4591 import connections to be ignored if the 32-bit jiffies counter
4592 wraps. Use a 64-bit jiffies counter.
4595 Frequency : very large clusters immediately after boot
4597 Description: LNET request buffers exhausted under heavy short-term load
4598 Details : If a large number of client requests are generated on a service
4599 that has previously never seen so many requests it is possible
4600 that the request buffer growth cannot keep up with the spike in
4601 demand. Instead of dropping incoming requests, they are held in
4602 the LND until the RPC service can accept more requests.
4605 Frequency : Sometimes during replay
4607 Description: Assertion failure in ll_local_open after replay.
4608 Details : If replay happened on an open request reply before we were able
4609 to set replay handler, reply will become not swabbed tripping the
4610 assertion in ll_local_open. Now we set the handler right after
4611 recognising of open request
4614 Frequency : very rare
4616 Description: kernel reports "badness in vsnprintf"
4617 Details : Reading from the "recovery_status" /proc file in small chunks
4618 may cause a negative length in lprocfs_obd_rd_recovery_status()
4619 call to vsnprintf() (which is otherwise harmless). Exit early
4620 if there is no more space in the output buffer.
4622 Severity : enhancement
4624 Description: clear OBD RPC statistics by writing to them
4625 Details : It is now possible to clear the OBD RPC statistics by writing
4626 to the "stats" file.
4631 Description: Client mtime is not the same on different clients after utimes
4632 Details : In some cases, the client was using the utimes() syscall on
4633 a file cached on another node. The clients now validate the
4634 ctime from the MDS + OSTs to determine which one is right.
4639 Description: Inability to activate failout mode
4640 Details : lconf script incorrectly assumed that in pythong string's numeric
4641 value is used in comparisons.
4644 Frequency : always with multiple stripes per file
4646 Description: Inefficient object allocation for mutli-stripe files
4647 Details : When selecting which OSTs to stripe files over, for files with
4648 a stripe count that divides evenly into the number of OSTs,
4649 the MDS is always picking the same starting OST for each file.
4650 Return the OST selection heuristic to the original design.
4655 Description: mount failures may take full timeout to return an error
4656 Details : Under some heavy load conditions it is possible that a
4657 failed mount can wait for the full obd_timeout interval,
4658 possibly several minutes, before reporting an error.
4659 Instead return an error as soon as the status is known.
4661 Frequency : quota enabled and large files being deleted
4663 Description: releasing more than 4GB of quota at once hangs OST
4664 Details : If a user deletes more than 4GB of files on a single OST it
4665 will cause the OST to spin in an infinite loop. Release
4666 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4671 Description: statfs data retrieved from /proc may be stale or zero
4672 Details : When reading per-device statfs data from /proc, in the
4673 {kbytes,files}_{total,free,avail} files, it may appear
4674 as zero or be out of date.
4677 Frequency : systems with MD RAID1 external journal devices
4679 Description: lconf's call to blkid is confused by RAID1 journal devices
4680 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4681 of returning all block devices that match the journal UUID.
4684 Frequency : always, for aggregate stripe size over 4GB
4686 Description: assertion fails when trying to use 4GB stripe size
4687 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4688 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4691 Frequency : always on ppc64
4693 Description: the first write on an ext3 filesystem with mballoc got stuck
4694 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4695 perform endianness conversion.
4697 ------------------------------------------------------------------------------
4699 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4701 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4702 INTEROPERATE with older versions automatically. Please read the
4703 user documentation before upgrading any part of a live system.
4704 * WARNING: Lustre networking configuration changes are required with
4705 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4708 * Support for kernels:
4709 2.6.9-22.0.2.EL (RHEL 4)
4710 2.6.5-7.244 (SLES 9)
4711 2.6.12.6 vanilla (kernel.org)
4714 Severity : enhancement
4715 Bugzilla : 7981/8208
4716 Description: Introduced Lustre Networking (LNET)
4717 Details : LNET is new networking infrastructure for Lustre, it includes
4718 a reorganized network configuration mode (see the user
4719 documentation for full details) as well as support for routing
4720 between different network fabrics. Lustre Networking Devices
4721 (LNDs) for the supported network fabrics have also been
4722 created for this new infrastructure.
4724 Severity : enhancement
4725 Description: Introduced Access control lists
4726 Details : clients can set ACLs on files and directories in order to have
4727 more fine-grained permissions than the standard Unix UGO+RWX.
4728 The MDS must be started with the "-o acl" mount option.
4730 Severity : enhancement
4731 Description: Introduced filesystem quotas
4732 Details : Administrators may now establish per-user quotas on the
4735 Severity : enhancement
4737 Description: Configuration change for the XT3
4738 The PTLLND is now used to run Lustre over Portals on the XT3
4739 The configure option(s) --with-cray-portals are no longer used.
4740 Rather --with-portals=<path-to-portals-includes> is used to
4741 enable building on the XT3. In addition to enable XT3 specific
4742 features the option --enable-cray-xt3 must be used.
4747 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4748 Details : A race between allocating a new llcd and re-getting the llcd_lock
4749 allowed another thread to grab newly-allocated llcd.
4751 Severity : enhancement
4753 Description: 2.6 OST async journal commit and locking fix to improve performance
4754 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4755 2.6 kernels are now async as they already were in 2.4 kernels so
4756 that they can commit concurrently with the network bulk transfer.
4757 For block-allocated files the filter allocation semaphore is held
4758 to avoid filesystem fragmentation during allocation. BKL lock
4759 removed for 2.6 xattr operations where it is no longer needed.
4764 Description: lconf incorrectly determined whether two IP networks could talk
4765 Details : In some more complicated routing and multiple-network
4766 configurations, lconf will avoid trying to make a network
4767 connection to a disjoint part of the IP space. It was doing the
4768 math incorrectly for one set of cases.
4773 Description: Fix for potential infinite loop processing records in an llog.
4774 Details : If an llog record is corrupted/zeroed, it is possible to loop
4775 forever in llog_process(). Validate the llog record length
4776 and skip the remainder of the block on error.
4779 Frequency : occasional (liblustre only)
4781 Description: liblustre could not open files whose last component is a symlink
4782 Details : sysio_path_walk() would incorrectly pass the open intent to
4783 intermediate path components.
4786 Frequency : rare (liblustre only with non-standard tuning)
4787 Bugzilla : 7201 (7350)
4788 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4789 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4790 after it has been released, only do this for referenced locks
4792 Severity : enhancement
4794 Description: specifying an (invalid) directory default stripe_size of -1
4795 would reset the directory default striping
4796 Details : stripe_size -1 was used internally to signal directory stripe
4797 removal, now use "all default" to signal dir stripe removal
4798 as a directory striping of "all default" is not useful
4801 Frequency : common for large clusters running liblustre clients
4803 Description: doing an ls when liblustre clients are running is slow
4804 Details : sending a glimpse AST to a liblustre client waits for every AST
4805 to time out, as liblustre clients will not respond. Since they
4806 cannot cache data we refresh the OST lock LVB from disk instead.
4808 Severity : enhancement
4810 Description: doing an ls at the same time as file IO can be slow
4811 Details : enqueue and other "small" requests can be blocked behind many
4812 large IO requests. Create a new OST IO portal for non-IO
4813 requests so they can be processed faster.
4816 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4818 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4819 created file with mode 000
4820 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4821 allow an ftruncate() as a non-root user to a file with mode 000.
4822 Lustre now respects this flag to disable mode checks when
4823 truncating a file owned by the user
4826 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4829 Description: Revoking locks from clients that went dead or catatonic might take
4831 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4832 cancellation of such locks instant on servers without waiting for
4833 any reply from clients. Clients drops these locks when cancel
4834 notification from server is received without replying.
4837 Frequency : liblustre-only, when liblustre client dies or becomes busy
4839 Description: Doing ls on Linux clients can take a long time with active
4841 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4842 granting such locks to it in the first place if possible. Locks
4843 are taken by proxy on the OST during the read or write and
4844 dropped immediately afterward. Add connect flags handling, do
4845 not grant locks to liblustre clients for glimpse ASTs.
4847 Severity : enhancement
4849 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4850 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4851 sequential reads and behaves badly with applications doing random
4852 reads. Improve it by reading ahead at least read region, and
4853 avoiding excessive large RPC for small reads.
4855 Severity : enhancement
4857 Description: Creating more than 1000 files for a single job may cause a load
4858 imbalance on the OSTs if there are also a large number of OSTs.
4859 Details : qos_prep_create() uses an OST index reseed value that is an
4860 even multiple of the number of available OSTs so that if the
4861 reseed happens in the middle of the object allocation it will
4862 still utilize the OSTs as uniformly as possible.
4867 Description: OST or MDS may oops in ping_evictor_main()
4868 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4869 but doesn't restart at beginning of obd_exports_timed list
4872 Severity : enhancement
4874 Description: improve by-nid export eviction on the MDS and OST
4875 Details : allow multiple exports with the same NID to be evicted at one
4876 time without re-searching the exports list.
4879 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4881 Description: MDS may oops in groups_free()
4882 Details : in rare race conditions a newly allocated group_info struct is
4883 freed again, and this can be NULL. The 2.4 compatibility code
4884 for groups_free() checked for a NULL pointer, but 2.6 did not.
4887 Frequency : common for liblustre clients doing little filesystem IO
4888 Bugzilla : 9352, 7313
4889 Description: server may evict liblustre clients accessing contended locks
4890 Details : if a client is granted a lock or receives a completion AST
4891 with a blocking AST already set it would not reply to the AST
4892 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4895 Frequency : lfs setstripe, only systems with more than 160 OSTs
4897 Description: unable to set striping with a starting offset beyond OST 160
4898 Details : llapi_create_file() incorrectly limited the starting stripe
4899 index to the maximum single-file stripe count.
4902 Frequency : LDAP users only
4904 Description: lconf did not handle in-kernel recovery with LDAP properly
4905 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4907 Severity : enhancement
4909 Description: bind OST threads to NUMA nodes to improve performance
4910 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4911 node and do their allocations there to localize memory access
4913 Severity : enhancement
4915 Description: llmount can determine client NID directly from Myrinet (GM)
4916 Details : the client NID code from gmnalnid was moved directly into
4917 llmount, removing the need to use this or specifying the
4918 client NID explicitly when mounting GM clients with zeroconf
4921 Frequency : if client is started with down MDS
4923 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4924 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4925 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4926 ptlrpc_import_delayed_req() because we didn't check if the
4927 request was interrupted, and we also didn't break out of the
4928 event loop if there was no timeout
4933 Description: data loss during non-page-aligned writes to a single file from
4934 both multiple nodes and multiple threads on one node at same time
4935 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4936 inconsistency led to false short-reads, that were cached and later
4937 used by ->prepare_write() to fill in partially written page,
4938 leading to data loss.
4941 Frequency : always, if lconf --abort_recovery used
4943 Description: lconf --abort_recovery fails with 'Operation not supported'
4944 Details : lconf was attempting to abort recovery on the MDT device and not
4947 Severity : enhancement
4949 Description: remove cleanup logs
4950 Details : replace lconf-generated cleanup logs with lustre internal
4951 cleanup routines. Eliminates the need for client-cleanup and
4954 Severity : enhancement
4956 Description: add support for EAs (user and system) on lustre filesystems
4957 Details : it is now possible to store extended attributes in the Lustre
4958 client filesystem, and with the user_xattr mount option it
4959 is possible to allow users to store EAs on their files also
4961 Severity : enhancement
4963 Description: Add possibility (config option) to show minimal available OST free
4965 Details : When compiled with --enable-mindf configure option, statfs(2)
4966 (and so, df) will return least minimal free space available from
4967 all OSTs as amount of free space on FS, instead of summary of
4968 free spaces of all OSTs.
4970 Severity : enhancement
4972 Description: do not expand extent locks acquired on OST-side
4973 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4974 by server: they are not cached anyway.
4977 Frequency : when mmap is used/binaries executed from Lustre
4979 Description: Unmmap pages before throwing them away from read cache.
4980 Details : llap_shrink cache now attempts to unmap pages before discarding
4981 them (if unmapping failed - do not discard). SLES9 kernel has
4982 extra checks that trigger if this unmapping is not done first.
4987 Description: lconf didn't resolve symlinks before checking to see whether a
4988 given mountpoint was already in use
4991 Frequency : when migrating failover services
4992 Bugzilla : 6395, 9514
4993 Description: When migrating a subset of services from a node (e.g. failback
4994 from a failover service node) the remaining services would
4995 time out and evict clients.
4996 Details : lconf --force (implied by --failover) sets the global obd_timeout
4997 to 5 seconds in order to quickly disconnect, but this caused
4998 other RPCs to time out too quickly. Do not change the global
4999 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5001 Severity : enhancement
5002 Frequency : if MDS is started with down OST
5003 Bugzilla : 9439,5706
5004 Description: Allow startup/shutdown of an MDS without depending on the
5005 availability of the OSTs.
5006 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5007 Add appropriate locking and lov-osc refcounts for safe
5008 cleaning. Add osc abort_inflight calls in case the
5009 synchronize never started.
5012 Frequency : occasional (Cray XT3 only)
5014 Description: root not authorized to access files in CRAY_PORTALS environment
5015 Details : The client process capabilities were not honoured on the MDS in
5016 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5017 been accessed by an authorized user then root was able to access
5018 the file on the local client also. The root user capabilities
5019 are now allowed on the MDS, as this environment has secure UID.
5022 Frequency : occasional
5024 Description: ldiskfs "too long searching" message happens too often
5025 Details : A debugging message (otherwise harmless) prints too often on
5026 the OST console. This has been reduced to only happen when
5027 there are fragmentation problems on the filesystem.
5032 Description: Division by zero in statfs when all OSCs are inactive
5033 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5034 lov_statfs divides by value returned by lov_get_stripecnt().
5038 Bugzilla : 9489, 3273
5039 Description: First write from each client to each OST was only 4kB in size,
5040 to initialize client writeback cache, which caused sub-optimal
5041 RPCs and poor layout on disk for the first written file.
5042 Details : Clients now request an initial cache grant at (re)connect time
5043 and so that they can start streaming writes to the cache right
5044 away and always do full-sized RPCs if there is enough data.
5045 If the OST is rebooted the client also re-establishes its grant
5046 so that client cached writes will be honoured under the grant.
5051 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5052 Details : Now I/O RPCs go to different portal number and (presumably) fast
5053 lock requests (and glimses) and other RPCs get their own service
5054 threads pool that should be able to service those RPCs
5057 Severity : enhancement
5059 Description: Ability to exchange lustre version between client and servers and
5060 issue warnings at client side if client is too old. Also for
5061 liblustre clients there is ability to refuse connection of too old
5063 Details : New 'version' field is added to connect data structure that is
5064 filled with version info. That info is later checked by server and
5068 Frequency : rare, liblustre only.
5069 Bugzilla : 9296, 9581
5070 Description: Two simultaneous writes from liblustre at offset within same page
5071 might proceed at the same time overwriting eachother with stale
5073 Details : I/O lock withing llu_file_prwv was released too early, before data
5074 actually was hitting the wire. Extended lock-holding time until
5075 server acknowledges receiving data.
5078 Frequency : extremely rare. Never observed in practice.
5080 Description: avoid generating lustre_handle cookie of 0.
5081 Details : class_handle_hash() generates handle cookies by incrementing
5082 global counter, and can hit 0 occasionaly (this is unlikely, but
5083 not impossible, because initial value of cookie counter is
5084 selected randonly). Value of 0 is used as a sentinel meaning
5085 "unassigned handle" --- avoid it. Also coalesce two critical
5086 sections in this function into one.
5088 Severity : enhancement
5090 Description: allow liblustre clients to delegate truncate locking to OST
5091 Details : To avoid overhead of locking, liblustre client instructs OST to
5092 take extent lock in ost_punch() on client's behalf. New connection
5093 flag is added to handle backward compatibility.
5095 Severity : enhancement
5096 Bugzilla : 4928, 7341, 9758
5097 Description: allow number of OST service threads to be specified
5098 Details : a module parameter allows the number of OST service threads
5099 to be specified via "options ost ost_num_threads={N}" in the
5100 OSS's /etc/modules.conf or /etc/modprobe.conf.
5104 Bugzilla : 6146, 9635, 9895
5105 Description: servers crash with bad pointer in target_handle_connect()
5106 Details : In rare cases when a client is reconnecting it was possible that
5107 the connection request was the last reference for that export.
5108 We would temporarily drop the export reference and get a new
5109 one, but this may have been the last reference and the export
5110 was just destroyed. Get new reference before dropping old one.
5112 Severity : enhancement
5113 Frequency : if client is started with failover MDS
5115 Description: Allow multiple MDS hostnames in the mount command
5116 Details : Try to read the configuration from all specified MDS
5117 hostnames during a client mount in case the "primary"
5120 Severity : enhancement
5122 Description: Stop sending data to evicted clients as soon as possible.
5123 Details : Check if the client we are about to send or are sending data to
5124 was evicted already. (Check is done every second of waiting,
5125 for which l_wait_event interface was extended to allow checking
5126 of exit condition at specified intervals).
5129 Frequency : rare, normally only when NFS exporting is done from client
5131 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5133 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5136 Frequency : always, if srand() is called before liblustre initialization
5138 Description: Liblustre uses system PRNG disturbing its usage by user application
5139 Details : Introduce internal to lustre fast and high-quality PRNG for
5140 lustre usage and make liblustre and some other places in generic
5141 lustre code to use it.
5143 Severity : enhancement
5144 Bugzilla : 9477, 9557, 9870
5145 Description: Verify that the MDS configuration logs are updated when xml is
5146 Details : Check if the .xml configuration logs are newer than the config
5147 logs stored on the MDS and report an error if this is the case.
5148 Request --write-conf, or allow starting with --old_conf.
5150 Severity : enhancement
5152 Description: Handle symlinks in the path when checking if Lustre is mounted.
5153 Details : Resolve intermediate symlinks when checking if a client has
5154 mounted a filesystem to avoid duplicate client mounts.
5159 Description: lconf can hit an error exception but still return success.
5160 Details : The lconf command catches the Command error exception at the top
5161 level script context and will exit with the associated exit
5162 status, but doesn't ensure that this exit status is non-zero.
5167 Description: failure of ptlrpc thread startup can cause oops
5168 Details : Starting a ptlrpc service thread can fail if there are a large
5169 number of threads or the server memory is very fragmented.
5170 Handle this without oopsing.
5173 Frequency : always, only if liblustre and non-default acceptor port was used
5175 Description: liblustre cannot connect to servers with non-default acceptor port
5176 Details : tcpnal_set_default_params() was not called and was therefore
5177 ignoring the environment varaible TCPNAL_PORT, as well as other
5178 TCPNAL_ environment variables
5183 Description: two objects could be created on the same OST for a single file
5184 Details : If an OST is down, in some cases it was possible to create two
5185 objects on a single OST for a single file. No problems other
5186 than potential performance impact and spurious error messages.
5190 Bugzilla : 5681, 9562
5191 Description: Client may oops in ll_unhash_aliases
5192 Details : Client dcache may become inconsistent in race condition.
5193 In some cases "getcwd" can fail if the current directory is
5199 Description: Inode refcounting problems in NFS export code
5200 Details : link_raw functions used to call d_instantiate without obtaining
5201 extra inode reference first.
5205 Bugzilla : 9942, 9903
5206 Description: Referencing freed requests leading to crash, memleaks with NFS.
5207 Details : We used to require that call to ll_revalidate_it was always
5208 followed by ll_lookup_it. Also with revalidate_special() it is
5209 possible to call ll_revalidate_it() twice for the same dentry
5210 even if first occurence returned success. This fix changes semantic
5211 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5212 reference on a request referred from the intent.
5213 ll_intent_release() then releases such a request.
5216 Frequency : rare, normally benchmark loads only
5218 Description: unlinked inodes were kept in memory on the client
5219 Details : If a client is repeatedly creating and unlinking files it
5220 can accumulate a lot of stale inodes in the inode slab cache.
5221 If there is no other client load running this can cause the
5222 client node to run out of memory. Instead flush old inodes
5223 from client cache that have the same inode number as a new inode.
5226 Frequency : SLES9 2.6.5 kernel and long filenames only
5227 Bugzilla : 9969, 10379
5228 Description: utime reports stale NFS file handle
5229 Details : SLES9 uses out-of-dentry names in some cases, which confused
5230 the lustre dentry revalidation. Change it to always use the
5234 Frequency : rare, unless heavy write-truncate concurrency is continuous
5235 Bugzilla : 4180, 6984, 7171, 9963, 9331
5236 Description: OST becomes very slow and/or deadlocked during object unlink
5237 Details : filter_destroy() was holding onto the parent directory lock
5238 while truncating+unlinking objects. For very large objects this
5239 may block other threads for a long time and slow overall OST
5240 responsiveness. It may also be possible to get a lock ordering
5241 deadlock in this case, or run out of journal credits because of
5242 the combined truncate+unlink. Solution is to do object truncate
5243 first in one transaction without parent lock, and then do the
5244 final unlink in a new transaction with the parent lock. This
5245 reduces the lock hold time dramatically.
5248 Frequency : rare, 2.4 kernels only
5250 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5251 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5252 before unmounting their filesystem in order to avoid deadlock.
5253 The kernel_locked() function in 2.4 kernels only checks whether
5254 the kernel lock is held, not whether it is this process that is
5255 holding it as 2.6 kernels do.
5260 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5261 Details : The client ptlrpc code may be trying to reconnect to a down
5262 server before a previous connection attempt has timed out.
5263 Increase the reconnect interval to be longer than the connection
5264 timeout interval to avoid sending duplicate connections to
5268 Frequency : echo_client brw_test command
5270 Description: fix echo_client to work with OST preallocated code
5271 Details : OST preallocation code (5137) didn't take echo_client IO path
5272 into account: echo_client calls filter methods outside of any
5273 OST thread and, hence, there is no per-thread preallocated
5274 pages and buffers to use. Solution: hijack pga pages for IO. As
5275 a byproduct, this avoids unnecessary data copying.
5279 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5280 Description: Client can oops in mdc_commit_close() after open replay
5281 Details : It was possible for the MDS to return an open request with no
5282 transaction number in mds_finish_transno() if the client was
5283 evicted, but without actually returning an error. Clients
5284 would later try to replay that open and may trip an assertion
5285 Simplify the client close codepath, and always return an error
5286 from the MDS in case the open is not successful.
5289 Frequency : rare, 2.6 OSTs only
5291 Description: OST may deadlock under high load on fragmented files
5292 Details : If there was a heavy load and highly-fragmented OST filesystems
5293 it was possible to have all the OST threads deadlock waiting on
5294 allocation of biovecs, because the biovecs were not released
5295 until the entire RPC IO was completed. Instead, release biovecs
5296 as soon as they are complete to ensure forward IO progress.
5298 Severity : enhancement
5300 Description: Support for specifying external journal device at mount
5301 Details : If an OST or MDS device is formatted with an external journal
5302 device, this device major/minor is stored in the ext3 superblock
5303 and may not be valid for failover. Allow detecting and
5304 specifying the external journal at mount time.
5309 Description: Mounting an MDS with pending unlinked files may cause oops
5310 Details : target_finish_recovery() calls mds_postrecov() which returned
5311 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5312 considers this an error and immediately begins cleaning up the
5313 lov, just after starting the mds_lov process
5315 Severity : enhancement
5317 Description: Implement 'lfs df' to report actual free space on per-OST basis
5318 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5319 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5320 usage of objects; '-h' to report in human readable format.
5322 ------------------------------------------------------------------------------
5324 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5331 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5332 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5333 at startup during orphan unlinking. Instead, skip buddy update
5334 before setup, as it will be regenerated after recovery is complete.
5339 Description: If an OST is inactive, its locks might reference stale inodes.
5340 Details : lov_change_cbdata() must iterate over all namespaces, even if
5341 they are inactive to clear inode references from the lock.
5343 Severity : enhancement
5344 Frequency : occasional, if non-standard max_dirty_mb used
5346 Description: Client will block write RPCs if not enough grant
5347 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5348 then the client will block writes while waiting for another RPC
5349 to complete instead of consuming its dirty limit. With change
5350 we get improved performance when max_dirty_mb is small.
5352 Severity : enhancement
5353 Bugzilla : 3389, 6253
5354 Description: Add support for supplementary groups on the MDS.
5355 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5356 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5357 lookups for user supplementary groups into a cache.
5361 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5362 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5363 for liblustre O_EXCL is enforced by MDS.
5367 Description: readdir never returns NULL in liblustre.
5368 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5369 dentry in struct dirent.
5373 Frequency : liblustre only, depends on application IO pattern
5374 Description: liblustre clients evicted if not contacting servers
5375 Details : Don't put liblustre clients into the ping_evictor list, so
5376 they will not be evicted by the pinger ever.
5378 Severity : enhancement
5380 Description: Add ability to evict clients by NID from MDS.
5381 Details : By echoing "nid:$NID" string into
5382 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5383 $NID would be instantly evicted from this MDS and from all active
5384 OSTs connected to it.
5388 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5389 Details : lookup_it_finish() used to query file size from OSTs that was not
5394 Description: service threads change working directory to that of init
5395 Details : Starting lustre service threads may pin the working directory
5396 of the parent thread, making that filesystem busy. Threads
5397 now change to the working directory of init to avoid this.
5401 Frequency : during shutdown only
5402 Description: shutdown with a failed MDS or OST can cause unmount to hang
5403 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5408 Frequency : 2.6 only
5409 Description: chmod/chown may include an extra supplementary group
5410 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5411 supplementary group and if none is specified this is used.
5414 Bugzilla : 5479 (6816)
5416 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5417 Details : If lookup is in progress on a file that is unlinked we might try
5418 to revalidate the inode and fail in revalidate after lookup is
5419 complete and ll_file_open() enqueues the open again but
5420 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5423 Frequency : always, if lconf --abort_recovery used
5425 Description: lconf --abort_recovery fails with 'Operation not supported'
5426 Details : lconf was attempting to abort recovery on the MDT device and not
5429 ------------------------------------------------------------------------------
5431 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5436 Frequency : rare (only unsupported configurations with a node running as an
5438 Bugzilla : 6514, 5137
5439 Description: Mounting a Lustre file system on a node running as an OST could
5441 Details : OSTs now preallocates memory needed to write out data at
5442 startup, instead of when needed, to avoid having to
5443 allocate memory in possibly low memory situations.
5444 Specifically, if the file system is mounted on on OST,
5445 memory pressure could force it to try to write out data,
5446 which it needed to allocate memory to do. Due to the low
5447 memory, it would be unable to do so and the node would
5448 become unresponsive.
5450 Severity : enhancement
5452 Description: Addition of lconf --service command line option
5453 Details : lconf now accepts a '--service <arg>' option, which is
5454 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5456 Severity : enhancement
5458 Description: Failover mode is now the default for OSTs.
5459 Details : By default, OSTs will now run in failover mode. To return to
5460 the old behaviour, add '--failout' to the lmc line for OSTs.
5462 Severity : enhancement
5464 Description: Health checks are now provided for MDS and OSTs
5465 Details : Additional detailed health check information on MSD and OSTs
5466 is now provided through the procfs health_check value.
5469 Frequency : occasional, depends on IO load
5471 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5472 after numerous create/delete cycles
5473 Details : The ext3 inode allocation policy would not allocate new inodes
5474 very well on the OSTs because there are no new directories
5475 being created. Instead we look for groups with free space if
5476 the parent directories are nearly full.
5481 Description: Network or server problems during mount may cause partially
5482 mounted clients instead of returning an error.
5483 Details : The config llog parsing code may overwrite the error return
5484 code during mount error handling, returning success instead
5490 Description: MDS can fail to allocate large reply buffers
5491 Details : After long uptimes the MDS can fail to allocate large reply
5492 buffers (e.g. zconf client mount config records) due to memory
5493 fragmentation or consumption by the buffer cache. Preallocate
5494 some large reply buffers so that these replies can be sent even
5495 under memory pressure.
5499 Frequency : rare (liblustre)
5500 Description: fsx running with liblustre complained that using truncate() to
5501 extend the file doesn't work. This patch corrects that issue.
5502 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5503 ATTR_SIZE and lsm use in llu_setattr_raw.
5507 Frequency : rare, only 2.6 kernels
5508 Description: Unusual file access patterns on the MDS may result in inode
5509 data being lost in very rare circumstances.
5510 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5511 code in the 2.6 kernel caused the inode and/or EA data not to
5512 be read from disk, causing single-file corruption.
5516 Frequency : rare, only 2.6 filesystems using extents
5517 Description: Heavy concurrent write and delete load may cause data corruption.
5518 Details : It was possible under high-load situations to have an extent
5519 metadata block in the block device cache from a just-unlinked
5520 file overwrite a newly-allocated data block. We now unmap any
5521 metadata buffers that alias just-allocated data blocks.
5525 Frequency : filesystems with default stripe_count larger than 77
5526 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5527 Details : lconf specifies an inode size of 4096 bytes when the default
5528 stripe_count is larger than 77. This conflicts with the default
5529 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5530 this case to avoid pinning too much memory for large EAs.
5532 ------------------------------------------------------------------------------
5534 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5539 Frequency : rare (extremely heavy IO load with hundreds of clients)
5541 Description: Client is evicted, gets IO error writing to file
5542 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5543 caused clients to be evicted for AST timeouts. The fixes in
5544 bug 5192 mean we no longer need to have such short AST timeouts
5545 so ldlm_timeout has been increased.
5548 Frequency : occasional during --force or --failover shutdown under load
5549 Bugzilla : 5949, 4834
5550 Description: Server oops/LBUG if stopped with --force or --failover under load
5551 Details : a collection of import/export refcount and cleanup ordering
5552 issues fixed for safer force cleanup
5555 Frequency : only filesystems larger than 120 OSTs
5556 Bugzilla : 5990, 6223
5557 Description: lfs getstripe would oops on a very large filesystem
5558 Details : lov_getconfig used kfree on vmalloc'd memory
5561 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5562 Bugzilla : 6242, 6243
5563 Description: reading from files that had been truncated to a non-zero size
5564 but never opened returned no data
5565 Details : ll_file_read() reads zeros from no-object files to EOF
5570 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5571 Details : There's little harm in aborting MDS/OSS recovery and letting it
5572 try again, so I removed the LASSERT and return an error instead.
5574 Severity : enhancement
5576 Description: New debugging infrastructure for tracking down data corruption
5577 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5578 (b) cover more of the client-side code path, and (c) try to narrow
5579 down where problems occurred
5583 Bugzilla : 3819, 4364, 4397, 6313
5584 Description: Racing close and eviction MDS could cause assertion in mds_close
5585 Details : It was possible to get multiple mfd references during close and
5586 client eviction, leading to one thread referencing a freed mfd.
5588 Severity: : enhancement
5589 Bugzilla : 3262, 6359
5590 Description: Attempts to reconnect to servers are now more aggressive.
5591 Details : This builds on the enhanced upcall-less recovery that was added
5592 in 1.4.2. When trying to reconnect to servers, clients will
5593 now try each server in the failover group every 10 seconds. By
5594 default, clients would previously try one server every 25 seconds.
5599 Description: After recovery, certain operations trigger a failed
5600 assertion on a client.
5601 Details : Failing over an mds, using lconf -d --failover, while a
5602 client was doing a readdir() call would cause the client to
5603 LBUG after recovery completed and the readdir() was resent.
5605 Severity : enhancement
5607 Description: Default groups are now added by lconf
5608 Details : You can now run lconf --group <servicename> without having to
5609 manually add groups with lmc.
5612 Frequency : occasional
5614 Description: Nodes with an elan id of 0 trigger a failed assertion
5617 Frequency : always when accessing e.g. tty/console device nodes
5619 Description: tty and some other devices nodes cannot be used on lustre
5620 Details : file's private_data field is used by device data and lustre
5621 values in there got lost. New field was added to struct file to
5622 store fs-specific private data.
5625 Frequency : when exporting Lustre via NFS
5627 Description: NFSD failed occasionally when looking up a path component
5628 Details : NFSD is looking up ".." which was broken in ext3 directories
5629 that had grown large enough to become hashed.
5632 Frequency : Clusters with multiple interfaces not on the same subnet
5634 Description: Nodes will repeatedly try to reconnect to an interface which it
5635 cannot reach and report an error to the log.
5636 Details : Extra peer list entries will be created by lconf with some peers
5637 unreachable. lconf now validates the peer before adding it.
5640 Frequency : Only if a default stripe is set on the filesystem root.
5642 Description: Setting a default stripe on the filesystem root prevented the
5643 filesystem from being remounted.
5644 Details : The client was sending extra request flags in the root getattr
5645 request and did not allocate a reply buffer for the dir EA.
5648 Frequency : occasional, higher if lots of files are accessed by one client
5649 Bugzilla : 6159, 6097
5650 Description: Client trips assertion regarding lsm mismatch/magic
5651 Details : While revalidating inodes the VFS looks up inodes with ifind()
5652 and in rare cases can find an inode that is being freed.
5653 The ll_test_inode() code will free the lsm during ifind()
5654 when it finds an existing inode and then the VFS later attaches
5655 this free lsm to a new inode.
5659 Bugzilla : 6422, 7030
5660 Description: MDS deadlock between mkdir and client eviction
5661 Details : Creating a new file via mkdir or mknod (starting a transaction
5662 and getting the ns lock) can deadlock with client eviction
5663 (gets ns lock and trying to finish a synchronous transaction).
5666 Frequency : occasional
5667 Description: While starting a server, the fsfilt_ext3 module could not be
5669 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5670 kernels. Previously, lconf would still use the ext3 name
5671 when trying to load modules. Now, it will correctly use
5672 ext3 on 2.4 and ldiskfs on 2.6.
5674 Severity : enhancement
5675 Description: The default stripe count has been changed to 1
5676 Details : The interpretation of the default stripe count (0, to lfs
5677 or lmc) has been changed to mean striping across a single
5678 OST, rather than all available. For general usage we have
5679 found a stripe count of 1 or 2 works best.
5681 Severity : enhancement
5682 Description: Add support for compiling against Cray portals.
5683 Details : Conditional compiling for some areas that are different
5687 Frequency : occasional
5688 Bugzilla : 6409, 6834
5689 Description: Creating files with an explicit stripe count may lead to
5690 a failed assertion on the MDS
5691 Details : If some OSTs are full or unavailable, creating files may
5692 trigger a failed assertion on the MDS. Now, Lustre will
5693 try to use other servers or return an error to the
5697 Frequency : occasional
5699 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5700 caused lock timeout during readahead (since 1.4.2).
5701 Details : Processes doing readahead might match a lock that hasn't been
5702 granted yet if there are overlapping and conflicting lock
5703 requests. The readahead process waits on ungranted lock
5704 (original lock is CBPENDING), while OST waits for that process
5705 to cancel CBPENDING read lock and eventually evicts client.
5707 Severity : enhancement
5709 Description: Initial enabling of flock support for clients
5710 Details : Implements fcntl advisory locking and file status functions.
5711 This feature is provided as an optional mount flag (default
5712 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5713 locking are implemented yet, and those that are are not guaranteed
5714 to be completely correct in production environments.
5715 mount -t lustre -o [flock|noflock] ...
5718 Frequency : occasional
5720 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5721 assertion in the ext3 JBD (journaling) layer.
5722 Details : The b_committed_data struct is protected by the big kernel lock
5723 in 2.4 kernels, serializing journal_commit_transaction() and
5724 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5725 access to this struct. In 2.6 kernels there is finer grained
5726 locking to improve SMP performance of the JBD layer.
5730 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5735 Description: merge and cleanup kernel patches.
5738 -----------------------------------------------------------------------------
5740 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5742 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5743 versions of Lustre. Please update all servers and clients to
5744 version 1.4.2 or later at the same time. You must also run
5745 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5746 to update the configuration logs.
5748 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5749 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5750 - swab llog records of type '0' so we get proper header size/idx (5861)
5751 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5752 - fix rename of one directory over another leaking an inode (5953)
5753 - avoid SetPageDirty on 2.6 (5981)
5754 - don't re-add just-being-destroyed locks to the waiting list (5653)
5755 - when creating new directories, inherit the parent's custom
5756 striping settings if present parent (3048)
5757 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5758 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5759 - don't hold client locks on temporary worklist from l_lru (5666)
5760 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5761 - automatically evict dead clients (5921)
5762 - Update file size properly in create+truncate+fstat case (6196)
5763 - Do not unhash mountpoint dentries, do not allow removal of
5765 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5766 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5767 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5768 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5769 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5770 - Correctly handle reads of files with no objects (6243)
5771 - lctl recover will also mark a device active if deactivate used (5933)
5773 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5774 - allow --write-conf on an MDS with different nettype than client (5619)
5775 - don't write config llogs to MDS for mounts not from that MDS (5617)
5776 - lconf should create multiple TCP connections from a client (5201)
5777 - init scripts are now turned off by default; run chkconfig --on
5778 lustre and chkconfig --on lustrefs to use them
5779 - upcalls are no longer needed for clients to recover to failover
5781 - add --abort-recovery option to lconf to abort recovery on device
5783 - add support for an arbitrary number of OSTs (3026)
5784 - Quota support protocol changes.
5785 - forward compatibility changes to wire structs (6007)
5786 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5787 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5788 - improved "lustre status" script
5789 - initialize blocksize for non-regular files (6062)
5790 - added --disable-server and --disable-client configure options (5782)
5791 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5792 - Vanilla 2.4.29 support
5793 - increase maximum number of obd devices to 520 (6242)
5794 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5795 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5796 SLES 9 2.6.5 kernels
5797 - limit stripes per file to 160 (the maximum EA size) (6093)
5799 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5802 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5803 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5804 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5805 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5806 - small ext3 extents cleanups and fixes (5733)
5807 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5808 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5809 - avoid needless client->OST connect, fix handle mismatch (5317)
5810 - fix DLM error path that led to out-of-sync client, long delays (5779)
5811 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5812 - fix several locking issues related to i_size (5492,5624,5654,5672)
5813 - don't move pending lock onto export if it is already evicted (5683)
5814 - fix kernel oops when creating .foo in unlinked directory (5548)
5815 - fix deadlock in obdfilter statistics vs. object create (5811)
5816 - use time_{before,after} to avoid timer jiffies wrap (5882)
5817 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5818 - Do not leak request if server was not able to process it (5154)
5819 - If mds_open unable to find parent dir, make that negative lookup(5154)
5820 - don't create new directories with extent-mapping (5909, 5936)
5822 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5823 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5824 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5825 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5826 - support for RHEL4 kernel on i686 (5773)
5827 - provide error messages when incompatible logs are encountered (5898)
5829 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5830 * version 1.4.0.10 (1.4.1 release candidate 1)
5832 - don't keep a lock reference when lock is not granted (4238)
5833 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5834 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5835 - drop import inflight refcount on signal_completed_replay error (5255)
5836 - unlock page after async write error during send (3677)
5837 - handle missing objects in filter_preprw_read properly (5265)
5838 - no transno return for symlink open, don't save no-trasno open (3440)
5839 - don't try to complete elan receive that already failed (4012)
5840 - free RPC server reply state on error (5406)
5841 - clean up thread from ptlrpc_start_thread() on error (5160)
5842 - readahead could read extra page into cache that wasn't ejected (5388)
5843 - prevent races in class_attach/setup/cleanup/detach (5260)
5844 - don't dereference de->d_inode after l_dput of de (5458)
5845 - use "int" for stripe value returned from lock_to_stripe (5544)
5846 - mballoc allocation and error-checking fixes in 2.6 (5504)
5847 - block device patches to fix I/O request sizes in 2.6 (5482)
5848 - look up hostnames for IB nals (5602)
5849 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5850 - don't start multiple acceptors for the same port (5277)
5851 - fix incorrect LASSERT in mds_getattr_name (5635)
5852 - export a proc file for general "ping" checking (5628)
5853 - fix "lfs check" to not block when the MDS is down (5628)
5855 - service request history (4965)
5856 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5857 - create an "evict_client" /proc entry on OSTs, like the MDS has
5858 - fix mount usage message, return errors per mount(8) (5168)
5859 - change grep [] to grep "[]" in tests so they work in more UMLs
5860 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5861 - remove ancient LOV_MAGIC_V0 EA support (5047)
5862 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5863 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5864 - allow readahead allocations to fail when low on memory (5383)
5865 - mmap locking landed again, after considerable improvement (2828)
5866 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5868 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5871 - send OST transaction number in read/write reply to free req (4966)
5872 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5873 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5874 - the watchdog thread now runs as interruptible (5246)
5875 - flock/lockf fixes (but it's still disabled, pending 5135)
5876 - don't use EXT3 constants in llite code (5094)
5877 - memory shortage at startup could cause assertion (5176)
5879 - reorganization of lov code
5880 - single portals codebase
5882 - add extents/mballoc support (5025)
5883 - direct I/O reads in the obdfilter (4048)
5884 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5886 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5889 - send OST transaction number in read/write reply to free req (4966)
5890 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5891 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5892 - the watchdog thread now runs as interruptible (5246)
5893 - handle missing objects in filter_preprw_read properly (5265)
5894 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5895 - drop import inflight refcount on signal_completed_replay error (5255)
5896 - unlock page after async write error during send (3677)
5897 - return original error code on reconstructed replies (3761)
5898 - no transno return for symlink open, don't save no-trasno open (3440)
5900 - add pid to ldlm debugging output (4922)
5901 - bump the watchdog timeouts -- we can't handle 30sec yet
5902 - extra debugging for orphan dentry/inode bug (5259)
5904 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5907 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5908 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5909 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5910 - fix NULL dereference in /proc/sys/portals/routes (4827)
5911 - allow failed mdc_close() operations to be interrupted (4561)
5912 - stop precreate on OST before MDS would time out on it (4778)
5913 - don't send partial-page writes before EOF from client (4410)
5914 - discard client grant for sub-page writes on large-page clients (4520)
5915 - don't free dentries not owned by NFS code, check generation (4806)
5916 - fix lsm leak if mds_create_objects() fails (4801)
5917 - limit debug_daemon file size, always print CERROR messages (4789)
5918 - use transno after validating reply (3892)
5919 - process timed out requests if import state changes (3754)
5920 - update mtime on OST during writes, return in glimpse (4829)
5921 - add mkfsoptions to LDAP (4679)
5922 - use ->max_readahead method instead of zapping global ra (5039)
5923 - don't interrupt __l_wait_event() during strace
5925 - add software watchdogs to catch hung threads quickly (4941)
5926 - make lustrefs init script start after nfs is mounted
5927 - fix CWARN/ERROR duplication (4930)
5928 - return async write errors to application if possible (2248)
5929 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5930 - print NAL number in %x format (4645)
5931 - update barely-supported suse-2.4.21-171 series (4842)
5932 - support for sles 9 %post scripts
5933 - support for building 2.6 kernel-source packages
5934 - support for sles km_* packages
5936 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5939 - ignore -ENOENT errors in osc_destroy (3639)
5940 - notify osc create thread that OSC is being cleaned up (4600)
5941 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5942 - reconstruct ost_handle() like mds_handle() (4657)
5943 - create a new thread to do import eviction to avoid deadlock (3969)
5944 - let lconf resolve symlinked-to devices (4629)
5945 - don't unlink "objects" from directory with default EA (4554)
5946 - hold socknal file ref over connect in case target is down (4394)
5947 - allow more than 32000 subdirectories in a single directory (3244)
5948 - fix blocks count for O_DIRECT writes (3751)
5949 - OST returns ENOSPC from object create when no space left (4539)
5950 - don't send truncate RPC if file size isn't changing (4410)
5951 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5952 - bind to privileged port in socknal and tcpnal (3689)
5954 - rate limit CERROR/CWARN console message to avoid overload (4519)
5955 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5956 - basic mmap support (3918)
5957 - kernel patch series update from b1_4 (4711)
5959 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5962 - avoid crash during MDS cleanup with OST shut down (2775)
5963 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5964 - don't use bad inodes on the MDS (3744)
5965 - dynamic object preallocation to improve recovery speed (4236)
5966 - don't hold spinlock over lock dumping or change debug flags (4401)
5967 - don't zero obd_dev when it is force cleaned (3651)
5968 - print grants to console if they go negative (4431)
5969 - "lctl deactivate" will stop automatic recovery attempts (3406)
5970 - look for existing locks in ldlm_handle_enqueue() (3764)
5971 - don't resolve lock handle twice in recovery avoiding race (4401)
5972 - revalidate should check working dir is a directory (4134)
5974 - don't always mark "slow" obdfilter messages as errors (4418)
5976 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5979 - don't close LustreDB during write_conf until it is done (3860)
5980 - fix typo in lconf for_each_profile (3821)
5981 - allow dumping logs from multiple threads at one time (3820)
5982 - don't allow multiple threads in OSC recovery (3812)
5983 - fix debug_size parameters (3864)
5984 - fix mds_postrecov to initialize import for llog ctxt (3121)
5985 - replace config semaphore with spinlock (3306)
5986 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5987 - don't allow enqueue to complete on a destroyed export (3822)
5988 - down write_lock before checking llog header bitmap (3825)
5989 - recover from lock replay timeout (3764)
5990 - up llog sem before sending rpc (3652)
5991 - reduce ns lock hold times when setting kms (3267)
5992 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5993 - fix NULL deref and obd_dev leak on setup error (3312)
5994 - replace some LBUG about llog ops with error handling (3841)
5995 - don't match INVALID dentries from d_lookup and spin (3784)
5996 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5997 - fix invalid assertion in ptlrpc_set_wait (3880)
5999 - add libwrap support for the TCP acceptor (3996)
6000 - add /proc/sys/portals/routes for non-root route listing (3994)
6001 - allow setting MDS UUID in .xml (2580)
6002 - print the stack of a process that LBUGs (4228)
6004 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6007 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6008 - make sure to unset replay flag from failed open requests (3440)
6009 - if default stripe count is 0, use OST count for inode size (3636)
6010 - update parent mtime/ctime on client for create/unlink (2611)
6011 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6012 - free recovery state on server during a forced cleanup (3571)
6013 - unregister_reply for resent reqs (3063)
6014 - loop back devices mounting and status check on 2.6 (3563)
6015 - fix resource-creation race that can provoke i_size == 0 (3513)
6016 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6017 - more debugging for page-accounting assertion (3746)
6018 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6019 - avoid deadlock after precreation failure (3758)
6020 - fix race and lock order deadlock in orphan handling (3450, 3750)
6021 - add validity checks when grabbing inodes from l_ast_data (3599)
6023 - add /proc/.../recovery_status to obdfilter (3428)
6024 - lightweight CDEBUG infrastructure, debug daemon (3668)
6025 - change default OSC RPC parameters to be better on small clusters
6026 - turn off OST read cache for files smaller than 32MB
6027 - install man pages and include them in rpms (3100)
6028 - add new init script for (un)mounting lustre filesystems (2593)
6029 - run chkconfig in %post for init scripts (3701)
6030 - drop scimac NAL (unmaintained)
6032 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6035 - clean kiobufs before and after use (3485)
6036 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6037 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6038 - change init script to fail more clearly if not run as root (1528)
6039 - allow clients to reconnect during replay (1742)
6040 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6041 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6042 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6043 - fix cond_resched (3554)
6044 - only evict unfinished clients after recovery (3515)
6045 - allow bulk resend, prevent data loss (3570)
6046 - dynamic ptlrpc request buffer allocation (2102)
6047 - don't allow unlinking open directory if it isn't empty (2904)
6048 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6049 - remove extraneous obd dereference causing LASSERT failure (3334)
6050 - don't use get_cycles() when creating temp. files on the mds (3156)
6051 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6052 - handle EEXIST for set-stripe, set proper directory name (3336)
6054 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6055 - fix ksocknal_fmb_callback() error messages (2918)
6057 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6060 - don't copy lvb into (possibly NULL) reply on error (2983)
6061 - don't deref dentry after dput, don't free lvb on error (2922)
6062 - use the kms to determine writeback rpc length (2947)
6063 - increment oti_logcookies when osc is inactive (2948)
6064 - update client's i_blocks count via lvb messages (2543)
6065 - handle intent open/close of special files properly (1557)
6066 - mount MDS with errors=remount-ro, like obdfilter (2009)
6067 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6068 - don't use cancelling-locks' kms values (2947)
6069 - use highest lock extent for kms, not last one (2925)
6070 - don't dereference ERR_PTR() dentry in error handling path (3107)
6071 - fix thread race in portals_debug_dumplog() (3122)
6072 - create lprocfs device entries at setup instead of at attach (1519)
6073 - common AST error handler, don't evict client on completion race (3145)
6074 - zero nameidata in detach_mnt in 2.6 (3118)
6075 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6076 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6077 - initialize RPC timeout timer earlier for 2.6 (3219)
6078 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6079 - print nal/nid for unknown nid (3258)
6080 - additional checks for oscc recovery before doing precreate (3284)
6081 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6082 - don't crash in mdc_close for bad permissions on open (3285)
6083 - zero i_rdev for non-device files (3147)
6084 - clear page->private before handing to FS, better assertion (3119)
6085 - tune the read pipeline (3236)
6086 - fix incorrect decref of invalidated dentry (2350)
6087 - provide read-ahead stats and refine rpc in flight stats (3328)
6088 - don't hold journal transaction open across create RPC (3313)
6089 - update atime on MDS at close time (3265)
6090 - close LDAP connection when recovering to avoid server load (3315)
6091 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6092 - don't leak open file on MDS after open resend (3325)
6093 - serialize filter_precreate and filter_destroy_precreated (3329)
6094 - loop device shouldn't call sync_dev() for nul device (3092)
6095 - clear page cache after eviction (2766)
6096 - resynchronize MDS->OST in background (2824)
6097 - refuse to mount the same filesystem twice on same mountpoint (3394)
6098 - allow llmount to create routes for mounting behind routers (3320)
6099 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6100 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6101 - fix rare problem with rename on htree directories (3417)
6103 - allow default OST striping configuration per directory (1414)
6104 - fix compilation for qswnal for 2.6 kernels (3125)
6105 - increase maximum number of MDS request buffers for large systems
6106 - change liblustreapi to be useful for external progs like lfsck (3098)
6107 - increase local configuration timeout for slow disks (3353)
6108 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6110 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6113 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6114 - don't overwrite extent policy data in reply if lock was blocked (2901)
6115 - drop filter export grants atomically with removal from device (2663)
6116 - del obd_self_export from work_list in class_disconnect_exports (2908)
6117 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6118 - swab reply message in mdc_close, other PPC fixes (2464)
6119 - fix destroying of named logs (2325)
6120 - overwrite old logs when running lconf --write_conf (2264)
6121 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6122 - fix race in target_handle_connect (2898)
6123 - mds_reint_create() should take same inode create lock (2926)
6124 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6125 - don't close files for self_export to avoid uninitialized obd (2936)
6126 - allow MDS with the same name as client node (2939)
6127 - hold dentry reference for closed log files for unlink (2325)
6128 - reserve space for all logs during transactions (2059)
6129 - don't evict page beyond end of stripe extent (2925)
6130 - don't oops on a deleted current working directory (2399)
6131 - handle hard links to targets without a parent properly (2517)
6132 - don't dereference NULL lock when racing during eviction (2867)
6133 - don't grow lock extents when lots of conflicting locks (2919)
6135 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6138 - account for cache space usage on clients to avoid data loss (974)
6139 - lfsck support in lustre kernel code (2349)
6140 - reduce journal credits needed for BRW writes (2370)
6141 - orphan handling to avoid losing space on client/server crashes
6142 - ptlrpcd can be blocked, stopping ALL progress (2477)
6143 - use lock value blocks to assist in proper KMS, faster stat (1021)
6144 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6145 - recovery for initial connections (2355)
6146 - fixes for mds_cleanup_orphans (1934)
6147 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6148 - block all file creations until orphan recovery completes (1901)
6149 - client remove rq_connection from request struct (2423)
6150 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6151 - recovery timer race (2670)
6152 - mdc_close recovey bug (2532)
6153 - ptlrpc cleanup bug (2710)
6154 - mds timeout on local locks (2588)
6155 - namespace lock held during RPCs (2431)
6156 - handle interrupted sync write properly (2503)
6157 - don't try to handle a message that hasn't been replied to (2699)
6158 - client assert failure during cleanup after abort recovery (2701)
6159 - leak mdc device after failed mount (2712)
6160 - ptlrpc_check_set allows timedout requests to complete (2714)
6161 - wait for inflight reqs when ptlrpcd finishes (2710)
6162 - make sure unregistered services are removed from the srv_list
6163 - reset bulk XID's when resending them (caught by 1138 test)
6164 - unregister_bulk after timeout
6165 - fix lconf error (2694)
6166 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6167 - readahead locks pages, leaves pending causing memory pressure (2673)
6168 - increase OST request buffers to 4096 on large machines (2729)
6169 - fix up permission of existing directories in simple_mkdir (2661)
6170 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6171 - don't assign transno to errored transactions (2742)
6172 - don't delete objects on OST if given a bogus objid from MDS (2751)
6173 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6174 - if rq_no_resend, then timeout request after recovery (2432)
6175 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6176 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6177 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6178 - MDS recovery shouldn't delete objects that it has given out (2730)
6179 - if enqueue arrives after completion, don't clobber LVB (2819)
6180 - don't unlock pages twice when trigger_group_io returns error (2814)
6181 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6182 - don't write pages to disk if there was an error (1450)
6183 - don't ping imports that have recovery disabled (2676)
6184 - take buffered bytes into account when balancing socknal conn (2817)
6185 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6186 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6187 - remove little-endian swabbing of llog records (1987)
6188 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6189 - retry reposting request buffers if they fail (1191)
6190 - grow extent at grant time to avoid granting a revoked lock (2809)
6191 - lock revoke doesn't evict page if covered by a second lock (2765)
6192 - disable VM readahead to avoid reading outside lock extents (2805)
6194 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6195 - updated kernel patches for hp-2.4.20 kernel (2681)
6197 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6200 - fix truncated write corruption (2366)
6201 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6203 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6204 - ensure len(uuid) < 37 in lmc (1171)
6205 - fix ia64 OOPS in llog_test (2255)
6206 - zero end of page at obdfilter for partial page writes (2648)
6207 - don't leave stale dentries around after renames (bug 2428)
6208 - fix timeouts when evicting a client with a single lock held (2642)
6209 - set deadline for the initial HELLO message to drain (2634)
6210 - print out dotted-quad IP addresses in the socknal (2302)
6212 - additional debugging for MDS client eviction problem (2443)
6213 - fix mkfsoptions support for osts (2603, 2604)
6215 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6218 - add series for the vanilla 2.6.0 kernel
6219 - add series for the vanilla 2.4.24 kernel
6220 - add series for a cray x86/64 UL kernel drop
6221 - fix xattr patches for the vanilla 2.4.19 series
6223 - generate true UUIDs in lmc (1171)
6224 - have portals stack dumping break in UML (2466)
6225 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6226 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6227 - wake callback waiting threads on client eviction (2460)
6228 - Add --ptldebug and --subsystem to lmc (1719)
6229 - update assertion to allow safe interrupt allocation
6230 - set rq_no_resend for cancel requests (2432)
6231 - recalculate ptlrpcd timeout after resend (2494)
6232 - call vfs_rmdir when removing pending directories (2368)
6233 - fix renaming a file to itself (2429)
6234 - lmc creates a default one-stripe lov (2454)
6235 - expand procfs space to handle large clusters (2326)
6236 - increase UML stack to avoid overflow
6237 - update lconf's list of debug and subsystem masks
6238 - fix lfs find --obd (2510)
6239 - /proc tunable for disabling filter read caching (2591)
6240 - stop rpm packages from altering slapd.conf (2301)
6241 - disable nagle in the socknal under 0conf (2578)
6242 - choose mds inode size based on stripe count (2572)
6243 - fix kernel-source rpm problems (2516)
6245 - add --disable-doc to avoid pdf generation (2421)
6246 - update documentation, tests, type-os, comments
6247 - avoid format warnings on ia64
6248 - remove the TOE NAL
6249 - tiny code cleanups by removing unused fields
6251 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6254 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6255 - remove the most blatant lies from BUILDING file (2371)
6256 - change default debug level to reasonable production setting
6257 - reduce client side cache size to reduce cache flush time
6258 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6259 - make TCP zerocopy and pinger support enabled by default (2476)
6260 - sync writes completed after process exits caused crashes (2319)
6261 - maintain correct mount count on the MDS (2356)
6262 - backout 1557, because 2316 wasn't really fixed
6263 - better file I/O statistics gathering in /proc
6264 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6265 - another kernel patch to fix zero-copy TCP function export
6266 - don't take duplicate lock when processing re-sent getattr (2420)
6267 - lctl uses obd_self_export instead of creating new conn (2353)
6268 - MDS/OST recovery case which requires object creation asserted (2425)
6269 - move lfs from /usr/sbin to /usr/bin in packages
6270 - fix race between mds_client_add and mds_client_free (2417)
6271 - use kmalloc instead of slabs in portals (2430)
6272 - don't create duplicate records when a failover MDS is present (2442)
6273 - remove unnecessary mount age check (2332)
6274 - don't remove directory inodes from locks prematurely (2451)
6275 - don't break if MDS service name is the same as hostname (2103)
6276 - fix races in client write RPC generation when cache full (2482)
6278 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6281 - remove now-unused request->rq_obd (278)
6282 - if an allocation fails, print out how much memory we've used (1933)
6283 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6284 - add the "configurable stack size" patch to most series files (1256)
6285 - ability to write large log records, for 100+ OST configs (2306)
6286 - fix NULL deref when filter_prep fails (2314)
6287 - fix operator precedence error in filter_sync
6288 - dynamic allocation of socknal TX descriptors (2315)
6289 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6290 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6291 - ensure that CWARN messages in Portals always get to the syslog
6292 - __init/__exit are not for prototype decls (ldlm_init/exit)
6293 - x86-64 compile warning fixes
6294 - fix gateway LMC keyword conflict (2318)
6295 - fix MDS lock inversions in getattr/reint paths (1844)
6296 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6297 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6298 - DLM race condition prevented some lock evictions (2328)
6299 - ENOMEM detection and retry on socknal sends (2230)
6300 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6301 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6302 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6303 - better error messages when a client is rejected during recovery (1505)
6304 - avoid cancelling locks which were never granted, after failure (2330)
6305 - fix i_sem/journal inversion in mds_client_add (2333)
6306 - fix truncate/getattr lock cycle deadlock (2334)
6307 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6308 - fix two rare exit paths which could leak an l_lock() ref (2321)
6309 - fencepost error in MDS/OST orphan recovery (2226)
6310 - make log record alignment 8 bytes (1988)
6311 - lstripe now fails when requested offset > ost_count (2237)
6312 - ensure that all kernel series have a complete list.h (1607)
6313 - fix crashes in special-file operations (2316)
6314 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6316 - allow configurable automake binary, for testing new versions
6317 - small update to the lfs documentation
6319 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6321 * fix negative export reference count in fsfilt_sync (2312)
6323 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6324 * release candidate 0.9.1
6326 - orphans are moved into the PENDING directory for possible recovery
6327 - replayed opens now open by fid for orphan/rename safety (1042)
6328 - last close of an orphan inode generates a transno (683)
6329 - chdir() and mount() now pin the directory entry (1020)
6330 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6331 - discard very old requests without processing them (1502)
6332 - remove some common, well-understood CERRORs (1505)
6333 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6334 - clear "grant" flags in OST replies until OST grant code lands (1644)
6335 - fix read performance by not clobbering i_blksize on client (1598)
6336 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6337 - make LRU size a /proc tunable, clears locks when reduced (707)
6338 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6339 - server locks take references on exports now (1558)
6340 - build fixes for 2.4.20-rh trees (1663)
6341 - return an error from lov_create if all OSCs are inactive (1751)
6342 - fix import levels when a reconnect happens without a timeout (1597)
6343 - exit early from mds_open if we get a lookup error (1749)
6344 - partial page read at EOF wouldn't wait for disk before sending (1642)
6345 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6346 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6347 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6348 - write/truncate lock inversion (1639)
6349 - Don't auto-load obdclass, portals modules during cleanup (1495)
6350 - fix timestamps from jumping to "now" (1763)
6351 - extra journal assertions (1648)
6352 - add an extra multiunlink test (1771)
6353 - fix read_record/write_record API (1776)
6354 - fix leak of offset_extent, possible incorrect i_size later (1772)
6355 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6356 - Debugging for the kqswnal_get_idle_tx problems (1820)
6357 - Allow recovery to be attempted multiple times (1536)
6358 - Write out MDS last_rcvd file after it is first created (1600)
6359 - Fix tx_descriptor leak in failed transmit situations (1827)
6360 - ext3 journaling fixes for assertion failure after IO error (1871)
6361 - class_export_put() on freed export after completion AST error (1896)
6362 - Fix revalidate looping in VFS (1322)
6363 - Don't access a freed export during MDS_REINT timeout (1521)
6364 - Add open-unlink recovery support on the MDS (1673,1764)
6365 - Return an error if no MDS data was read from last_rcvd (1946)
6366 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6367 - Don't setup a disk device that doesn't match exported UUID (317)
6368 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6369 - avoid committing NULL handle in force close
6370 - local.sh is now a one-stripe LOV configuration
6371 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6372 - fix ext3 htree duplicate directory entry corruption (1516)
6373 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6374 - update to new LOV EA format (2097)
6375 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6377 2003-06-15 Phil Schwan <phil@clusterfs.com>
6380 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6381 - per-import/export recovery handling (958, 931, 959)
6382 - multiple last-rcvd slots, for serving multiple FSes (949)
6383 - connections are again shared between multiple imp/exports (963, 964)
6384 - "umount -f" would hang if any requests needed to be sent (393, 978)
6385 - avoid pinning large req buffer by copying for queued messages (989)
6386 - add "uuid" to "lctl device" command to help upcalls (991)
6387 - "open" RPCs with transnos would confuse recovery counters (1037)
6388 - do proper endian conversion of all wire messages (288, 340, 891)
6389 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6390 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6391 - fix ptlrpc_connection leak in target_handle_connect (1174)
6392 - fix import refcounting bug in OST and MDS cleanup (1134)
6393 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6394 - fix very unlikely obd_types race condition (501)
6395 - remove osc_open hack for echo_client (1187)
6396 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6397 - a failure in read_inode2 leads to deadlock (1139)
6398 - cancel ack-locks as soon as transaction is committed (1072)
6399 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6400 - make sure to commitrw after any preprw to avoid deadlock (1162)
6401 - failing to execute a file in a lustre FS would lock inode (1203)
6402 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6403 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6404 - obd_uuid2tgt would walk off the end of the list (1255)
6405 - on IA64 the osc would give portals incorrect bulk size (1258)
6406 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6407 - fix lock inversion caused by new llite matching code (1282)
6408 - limit the number of dirty pages on a client to 10MB (1286)
6409 - timed out locks were not being corrected cancelled (1289)
6410 - fix O_DIRECT above 4GB on IA-32 (1292)
6411 * major user-visible changes
6412 - fail out/fail over policy now controlled by the upcall (993)
6414 - add OBD_PING to check server availability and failure (954)
6415 - lustre messages are now sent in sending host order (288, 340, 891)
6416 - add eadatalen to MDS getattr reply (340)
6417 - OST read replies may contain second buffer, with per-page status (593)
6419 2003-03-11 Phil Schwan <phil@clusterfs.com>
6422 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6423 - failed open()s could cause deadlock; fixed (867, 869)
6424 - stop cancelling OST locks when files are closed (481)
6425 - overlapping XID spaces caused network corruption (851, 853)
6426 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6427 - setattr_raw added, to do single-RPC, server-side setattrs
6428 - lmc/lconf syntax change for OST UUIDs
6429 - fix crashy race condition between ptlrpc_free_req and osc_close
6430 - don't use request in mdc_enqueue if we hit a timeout (889)
6431 - don't set the inode i_size for regular files from the MDS (896)
6432 - handle out of order completion AST (842)
6433 - don't LBUG if a lock request times out after receiving AST (913)
6434 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6435 - if a bad lock AST arrives, send an error instead of dropping entirely
6436 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6437 - fix leak in bulk IO when only partially completed (899, 900, 926)
6438 - fix O_DIRECT for ia64 (55)
6439 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6440 - C-z after timeout could hang a process forever; fixed (977)
6442 - client-side I/O cache (678, 924, 929, 941, 970)
6444 - READPAGE and SETATTRs which don't take server-side locks get
6447 2003-02-11 Phil Schwan <phil@clusterfs.com>
6450 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6451 - Fix fsfilt_extN_readpage() to read a full page of directory
6452 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6453 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6454 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6455 - Most intent operations were removed, in favour of a new RPC mode
6456 that does a single RPC to the server and bypasses most of the VFS
6457 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6458 - Aggressively cancel local locks on DLM servers
6459 - mds_reint_unlink sends EA to the client if it's the last nlink.
6460 client uses that EA to unlink OST objects.
6461 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6462 - recursive symlinks were fixed (439)
6463 - fixed NULL deref in DEBUG_REQ
6464 - filter_update_lastobjid no longer calls sync, which annoyed extN
6465 - fixed multi-client small-writes to a single file problem (445)
6466 - fixed mtime updates during file writes (607)
6467 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6468 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6469 - fixed Linux OST size reporting problem (444, 656)
6470 - OST now updates object mtime with writes or setattr (607, 619)
6471 - client verifies file size before zeroing page past EOF (445)
6472 - OST now writes last allocated objid to disk with allocation (108)
6473 - LOV on echo now works (409)
6475 - mds_reint_unlink sends a new buffer, with the EA included. this
6476 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6477 set if a regular file was being unlinked, and it was the last link
6478 - use PtlGet from the target for bulk writes (315)
6479 - OST now updates object mtime with writes or setattr (607, 619)
6480 - LDLM now has a grant-time callback to revalidate locked items, if
6482 - Many MDS operations were reorganized to combat race conditions
6484 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6485 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6487 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6490 - Fully reactivate OST imports after reconnection (512, others)
6491 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6492 - More graceful error handling for truncating on dead OST (515)
6493 - Don't error out unless we're actually accessing dead stripes (474)
6494 - Fix garbage sizes when stripes are missing (410)
6495 - LRU counters were broken, causing constant lock purge (433, 432)
6496 - garbage on read from stripes with failed OSTs (441)
6497 - mark OSCs as active before reconnecting during recovery (438)
6498 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6499 - lfind did not preserve OST order in output (443)
6500 - symlinks cause hung clients, incorrect data (439)
6501 - stop dereferencing request after dropping refcount (457)
6502 - don't LASSERT(spin_is_locked) on non-SMP (455)
6503 - fixes for many rename() bugs
6504 - fstat didn't correctly synchronize attributes (399)
6505 - server must handle lock cancellation during blocking AST prep (487)
6506 - bulk descriptors were free()d too soon (511)
6507 - fix paths in lconf, which would load incorrect modules (451, 507)
6508 - fix confusing lconf 'host not found' error message (386)
6509 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6510 - fix race condition in mdc_blocking_ast() for inode access (526)
6511 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6512 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6513 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6514 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6515 - fix lstripe in several ways (kernel side) (527)
6516 - fix request leak in ldlm_cli_enqueue (262)
6517 - incorrect OSC was marked inactive after OST failure
6518 - call mds_fs_cleanup before unmounting filesystem (524)
6519 - fix races between taking ns_lock and ldlm_lock_change_resource
6520 - fix races updating LOV export open file list
6521 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6522 - fix recovery NULL deref in ldlm_cli_cancel_unused
6523 - fix some DLM races by using new hash table for lock handles (419)
6524 - permit the client to specify desired inodes, at replay
6525 - duplicate requests when we queue them for replay reintegration
6526 - fix last_rcvd offset calculation
6527 - sync after each recovered transaction, so we always make progress
6528 - never, not always, ERESTART requests without transnos
6529 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6531 - skip replay if the MDS says that the client is already connected
6532 - don't check for a recovery-enabled export to match lctl's UUID
6533 - don't INC_USE_COUNT for phantom exports
6534 - don't crash when cleaning up phantom exports (567)
6535 - don't double-finish or set replay data for errored mdc_open requests
6536 - abort requests when they time out, so we don't get old replies
6537 - send/receive replies for AST messages again
6538 - if the client says that it doesn't have the lock, cancel it on the
6540 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6541 mark it as destroyed, it will all work out when decref is called
6542 - fix module use counts (22, 581)
6544 - ASTs now expect a reply (server cancels lock on error reply)
6546 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6549 - fix many simultaneous client startup (392)
6550 - fix dentry->d_it clobbering
6551 - credentials weren't being shipped for readdir/getattr operations
6552 - remove invalid assertions triggered during some concurrent MD
6554 - proper Lustre versions added (336, 389)
6555 - fix memory leak for create error case (398)
6556 - fix LOV locking bug that would get cli/srv out of sync
6557 - fix echo client over LOV (409)
6558 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6559 - fix double-O_EXCL intent crash (424)
6560 - avoid sending multiple lock CANCELs (352)
6562 - MDS can do multi-client recovery (modulo bugs in new code)
6564 - many updates, edits, cleanups
6566 2002-11-18 Phil Schwan <phil@clusterfs.com>
6569 - fix null d_it dereference (346)
6570 - fix full OST/dbench hang (333)
6571 - fix permission problem with file removal (286)
6572 - fix removal of OSCs from LOV when they fail
6573 - fix NULL deref during bulk timeout (214)
6574 - fix problems related to multiple filesystems on one MDS (241)
6575 - fixed serious subtle metadata locking bugs
6576 - free locks on clients when inodes are removed due to memory
6578 - fix inode pointer in lock data (285)
6579 - partial support for multiple MDS on a single host (241)
6580 - data locks weren't cancelled at clear_inode time (290, 311)
6581 - intent locks could lead to unbounded lock growth (205)
6582 - added a maximum lock count, an LRU list, and a flusher
6583 - fix multiple rename (365)
6584 - properly abstracted the echo client
6585 - OSC locked 1 byte too many; fixed
6586 - rewrote brw callback code:
6587 - fixed recovery bugs related to LOVs (306)
6588 - fixed too-many-pages-in-one-write crash (191)
6589 - fixed (again) crash in sync_io_timeout (214)
6590 - probably fixed callback-related race (385)
6592 - Add capability to MDS protocol
6593 - LDLM cancellations and callbacks on different portals
6595 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6598 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6599 - defer open object destruction to close time (601981, 138)
6600 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6601 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6602 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6603 - don't call obd_setattr in ll_file_release for destroyed objects
6604 * protocol change to lustre_msg: move |version| and add |flags|
6605 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6606 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6607 * added replay of create, unlink, link and rename operations during
6608 MDS failover; recovery should be much more robust now
6609 * remove failed OSCs from LOVs (only lov_create uses this so far)
6610 * the lustre-HOWTO was brought (more) up to date (582544)
6612 2002-10-23 Phil Schwan <phil@clusterfs.com>
6615 - in-use dentries weren't being reused properly (617851)
6616 - prevent multiple LDLM setup (599178)
6617 - fix LOV size calculations for truncate (617853)
6618 - fix client handling of MDS intent errors (POSIX)
6619 - fix permission bug in lovstripe.c test (624321)
6620 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6621 - truncate past end of file could corrupt data
6622 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6623 - a race in recovery could return ETIMEDOUT to apps (623947)
6624 - building outside the source directory was fixed
6625 * the lustre-HOWTO was brought (more) up to date (582544)
6626 * major progress was made on recovery functionality
6628 2002-10-10 Phil Schwan <phil@clusterfs.com>
6631 - recovery deadlock fix
6632 - rm -rf causes LBUG fix (617817)
6633 - file open by multiple tasks fix (618962)
6634 - directory permissions bugs (602707 and 620007)
6635 - journal_stop fixed with locking (611313)
6636 - O_APPEND failures resolved (618273, perhaps 614459)
6637 - lconf PATH fix (619770)
6638 - IA64 build fix (621450)
6639 - RPC buffer sizes scale with amount of memory
6641 2002-10-01 Phil Schwan <phil@clusterfs.com>
6644 - locks would be cancelled without throwing away data pages,
6645 resulting in inconsistent data (605627)
6646 - inode attributes were not always being refreshed (605627, 612449)
6647 - lconf now continues to cleanup after lctl reports an error
6648 - MDS now enforces user permissions (602707)
6649 - lprocfs cleanup fixed, but not yet enabled (614157)
6650 - fixed infinite server hang, should a client not respond to an AST
6651 - avoid going into recovery if user calls readlink() with a buffer
6652 that's too small (613941)
6653 - AST RPCs no longer require replies (614867) -- this may be changed
6654 - don't crash server if client sends an IOV that's too big (611336)
6655 - fixed lock conversion deadlock (611892)
6656 - fixed the following of symlinks (614622)
6657 * recovery: the server can remove locks from a client that dies, other
6658 clients can make progress
6659 * more extN patch fixes
6660 * compile-time configurable ptlrpc buffer allocations
6662 - collaborative read cache document
6663 - Lustre Lite Performance CDR document-in-progress
6665 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6668 - fix typo in patch-2.4.18
6670 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6673 - clear ptlrpc request each time in handle_incoming_request()
6674 - unlink of files now destroys the object on the OST
6676 2002-09-19 Peter Braam <braam@clusterfs.com>
6678 * add hard link support
6679 * change obdfile creation method
6680 * kernel patch changed
6682 2002-09-19 Peter Braam <braam@clusterfs.com>
6685 - stack overflow bug in extN fixed
6687 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6689 * documentation updates
6690 - add man pages for config tools
6691 - update tests/README to describe testing with new config tools
6692 - finish metadata API descriptions
6693 * bug fixes and cleanups
6694 - statfs workaround for 16TB limit
6695 - LOV stripe allocation improved, can stripe on subset of OSTs
6696 - LOV file size/IO offset was wrong for files > 4GB in size
6697 - object EA data was being dropped, caused files to be unreadable
6698 - memory overflow with non-LOV OST caused memory corruption
6699 - fixed regression tests to work with new config tools, obdfilter
6700 - fixed bug when directory size became larger than 1 block
6701 - fixed bug (for single client case) when PWD was deleted
6702 - invalidate local directory pages when doing intent-based ops
6703 - avoid LDLM oops when lock callback contained bad data
6705 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6707 * documentation updates
6708 * bug fixes and cleanups
6709 - configuration tools
6712 - 64-bit compile warnings
6713 - 64-bit internal statfs data
6715 * test_brw on persistent OST devices
6717 * lprocfs (disabled)
6719 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6721 * documentation updates
6722 * bug fixes and cleanups
6723 * configuration tools
6725 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6728 * many small fixes to 0_5_4
6729 * io/network handling
6730 * thinkos in MDS operations
6732 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6735 * crucial basic fixes to 0.5.3
6736 * IOR, Iozone work over Elan
6739 2002-08-07 Phil Schwan <phil@clusterfs.com>
6740 * version 0_5_3, our first alpha
6741 * we use the new Portals iovs
6742 * documentation updates
6743 * bug fixes and cleanups
6744 * small changes in the DLM wire protocol
6746 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6747 * version 0_5_1 with some initial stability,
6748 * locking on MD and file I/O.
6749 * documentation updates
6750 * several bug fixes since 0.5.0
6751 * small changes in wire protocol
6753 2002-07-18 Phil Schwan <phil@clusterfs.com>
6755 * delivered as Lustre Light Alpha
6756 * fixed a crash after handling invalid MDS requests
6757 * fixed directory pages for architectures with non-4k pages sizes
6759 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6760 * release version v0_4_4
6761 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6763 * Fixes a number of interruption problems with OST operations.
6764 * Update documentation for portals header changes
6765 * Move all wire protocol structs/defines to lustre_idl.h
6766 * Fixes symlink length bug.
6767 * Add tcpdump to repository.
6769 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6770 * release version v0_4_3
6771 * Fixes statfs for inodes on extN.
6772 * Fixes bug in runtests which would delete /etc/hosts.
6773 * Use 64-bit object IDs wherever possible (not into VFS though)
6774 Remove ost_get_info, which is unused by lustre, and out of date.
6776 2002-07-03 Peter Braam <braam@clusterfs.com>
6777 * release version v0_4_2 Fixes a lookup error (type not passed)
6778 * move forward to head of Portals
6779 * move forward to latest Lustre kernel
6781 2002-06-25 Peter Braam <braam@clusterfs.com>
6782 * release version v0_4_1. Hopefully stable on single node use.