3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.21.3.el7 (RHEL7.6)
9 3.10.0-1062.1.1.el7 (RHEL7.7)
10 4.4.120-92.70 (SLES12 SP2)
11 4.4.180-94.97 (SLES12 SP3)
12 4.4.0-142 (Ubuntu 16.04)
13 4.15.0-45 (Ubuntu 18.04)
14 vanilla linux 4.6.7 (ZFS only)
15 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
16 * Client known to build on unpatched kernels (others may also work):
17 2.6.32-696.18.7.el6 (RHEL6.9)
18 2.6.32-754.6.3.el6 (RHEL6.10)
19 3.10.0-862.14.4.el7 (RHEL7.5)
20 3.10.0-957.21.3.el7 (RHEL7.6)
21 3.10.0-1062.1.1.el7 (RHEL7.7)
22 4.14.0-49.13.1.el7a (RHEL7.5)
23 4.14.0-115.2.2.el7a (RHEL7.6)
24 4.18.0-80.4.2.el8 (RHEL8.0)
25 4.4.120-92.70 (SLES12 SP2)
26 4.4.180-94.97 (SLES12 SP3)
27 4.12.14-95.29 (SLES12 SP4)
28 4.4.0-142 (Ubuntu 16.04)
29 4.15.0-45 (Ubuntu 18.04)
30 vanilla linux 4.15.0 (ZFS only)
31 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
32 * Recommended ZFS / SPL version: 0.7.13
33 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
34 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
35 stack overflow. For more information, please refer to bugzilla 17630.
36 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
37 the re-exporting nodes to be mounted with "32bitapi" mount option
38 * Note for Ubuntu 14.04.5:
39 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
40 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
44 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
45 for currently supported client and server kernel versions.
46 * Server known to build on the following kernels (others may also work):
47 2.6.32-696.18.7.el6 (RHEL6.9)
48 3.10.0-862.14.4.el7 (RHEL7.5)
49 3.10.0-957.10.1.el7 (RHEL7.6)
50 4.4.120-92.70 (SLES12 SP2)
51 4.4.162-94.69 (SLES12 SP3)
52 4.4.0-142 (Ubuntu 16.04)
53 4.15.0-45 (Ubuntu 18.04)
54 vanilla linux 4.6.7 (ZFS only)
55 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
56 * Client known to build on unpatched kernels (others may also work):
57 2.6.32-696.18.7.el6 (RHEL6.9)
58 2.6.32-754.6.3.el6 (RHEL6.10)
59 3.10.0-862.14.4.el7 (RHEL7.5)
60 3.10.0-957.10.1.el7 (RHEL7.6)
61 4.14.0-49.13.1.el7a (RHEL7.5)
62 4.14.0-115.2.2.el7a (RHEL7.6)
63 4.4.120-92.70 (SLES12 SP2)
64 4.4.162-94.69 (SLES12 SP3)
65 4.12.14-95.13 (SLES12 SP4)
66 4.4.0-142 (Ubuntu 16.04)
67 4.15.0-45 (Ubuntu 18.04)
68 vanilla linux 4.15.0 (ZFS only)
69 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
70 * Recommended ZFS / SPL version: 0.7.13
71 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
72 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
73 stack overflow. For more information, please refer to bugzilla 17630.
74 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
75 the re-exporting nodes to be mounted with "32bitapi" mount option
76 * Note for Ubuntu 14.04.5:
77 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
78 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
83 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
84 for currently supported client and server kernel versions.
85 * Server known to build on the following kernels (others may also work):
86 2.6.32-696.18.7.el6 (RHEL6.9)
87 3.10.0-862.14.4.el7 (RHEL7.5)
88 3.10.0-957.10.1.el7 (RHEL7.6)
89 4.4.120-92.70 (SLES12 SP2)
90 4.4.162-94.69 (SLES12 SP3)
91 4.4.0-142 (Ubuntu 16.04)
92 4.15.0-45 (Ubuntu 18.04)
93 vanilla linux 4.6.7 (ZFS only)
94 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
95 * Client known to build on unpatched kernels (others may also work):
96 2.6.32-696.18.7.el6 (RHEL6.9)
97 2.6.32-754.6.3.el6 (RHEL6.10)
98 3.10.0-862.14.4.el7 (RHEL7.5)
99 3.10.0-957.10.1.el7 (RHEL7.6)
100 4.14.0-49.13.1.el7a (RHEL7.5)
101 4.14.0-115.2.2.el7a (RHEL7.6)
102 4.4.120-92.70 (SLES12 SP2)
103 4.4.162-94.69 (SLES12 SP3)
104 4.12.14-95.13 (SLES12 SP4)
105 4.4.0-142 (Ubuntu 16.04)
106 4.15.0-45 (Ubuntu 18.04)
107 vanilla linux 4.15.0 (ZFS only)
108 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
109 * Recommended ZFS / SPL version: 0.7.13
110 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
111 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
112 stack overflow. For more information, please refer to bugzilla 17630.
113 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
114 the re-exporting nodes to be mounted with "32bitapi" mount option
115 * Note for Ubuntu 14.04.5:
116 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
117 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
121 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
122 for currently supported client and server kernel versions.
123 * Server known to build on the following kernels (others may also work):
124 2.6.32-696.18.7.el6 (RHEL6.9)
125 3.10.0-862.14.4.el7 (RHEL7.5)
126 3.10.0-957.el7 (RHEL7.6)
127 4.4.120-92.70 (SLES12 SP2)
128 4.4.162-94.69 (SLES12 SP3)
129 4.4.0-131 (Ubuntu 16.04)
130 4.15.0-32 (Ubuntu 18.04)
131 vanilla linux 4.6.7 (ZFS only)
132 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
133 * Client known to build on unpatched kernels (others may also work):
134 2.6.32-696.18.7.el6 (RHEL6.9)
135 2.6.32-754.6.3.el6 (RHEL6.10)
136 3.10.0-862.14.4.el7 (RHEL7.5)
137 3.10.0-957.el7 (RHEL7.6)
138 4.14.0-49.13.1.el7a (RHEL7.5)
139 4.14.0-115.2.2.el7a (RHEL7.6)
140 4.4.120-92.70 (SLES12 SP2)
141 4.4.162-94.69 (SLES12 SP3)
142 4.4.0-131 (Ubuntu 16.04)
143 4.15.0-32 (Ubuntu 18.04)
144 vanilla linux 4.15.0 (ZFS only)
145 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
146 * Recommended ZFS / SPL version: 0.7.9
147 * Tested with ZFS / SPL version: 0.7.9
148 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
149 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
150 stack overflow. For more information, please refer to bugzilla 17630.
151 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
152 the re-exporting nodes to be mounted with "32bitapi" mount option
153 * Note for Ubuntu 14.04.5:
154 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
155 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
157 --------------------------------------------------------------------------------
160 2018-04-03 Intel Corporation
162 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
163 for currently supported client and server kernel versions.
164 * Server known to build on patched kernels:
165 2.6.32-431.29.2.el6 (RHEL6.5)
166 2.6.32-504.30.3.el6 (RHEL6.6)
167 2.6.32-573.26.1.el6 (RHEL6.7)
168 2.6.32-642.15.1.el6 (RHEL6.8)
169 2.6.32-696.18.7.el6 (RHEL6.9)
170 3.10.0-693.21.1.el7 (RHEL7.4)
171 3.0.101-0.47.71 (SLES11 SP3)
172 3.0.101-107 (SLES11 SP4)
173 3.12.74-60.64.40 (SLES12 SP1)
174 4.4.114-92.64 (SLES12 SP2)
175 4.4.114-94.11 (SLES12 SP3)
176 3.13.0-101 (Ubuntu 14.04, ZFS only)
177 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
178 4.4.0-116 (Ubuntu 16.04)
179 vanilla linux 4.6.7 (ZFS only)
180 * Client known to build on unpatched kernels:
181 2.6.32-431.29.2.el6 (RHEL6.5)
182 2.6.32-504.30.3.el6 (RHEL6.6)
183 2.6.32-573.26.1.el6 (RHEL6.7)
184 2.6.32-642.15.1.el6 (RHEL6.8)
185 2.6.32-696.18.7.el6 (RHEL6.9)
186 3.10.0-693.21.1.el7 (RHEL7.4)
187 3.0.101-0.47.71 (SLES11 SP3)
188 3.0.101-107 (SLES11 SP4)
189 3.12.74-60.64.40 (SLES12 SP1)
190 4.4.114-92.64 (SLES12 SP2)
191 4.4.114-94.11 (SLES12 SP3)
192 3.13.0-101 (Ubuntu 14.04)
193 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
194 4.4.0-116 (Ubuntu 16.04)
196 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
197 * Recommended ZFS / SPL version: 0.7.6
198 * Tested with ZFS / SPL version: 0.7.6
199 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
200 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
201 stack overflow. For more information, please refer to bugzilla 17630.
202 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
203 the re-exporting nodes to be mounted with "32bitapi" mount option
204 * Note for Ubuntu 14.04.5:
205 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
206 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
208 --------------------------------------------------------------------------------
210 2017-07-05 Intel Corporation
212 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
213 for currently supported client and server kernel versions.
214 * Server known to build on patched kernels:
215 2.6.32-431.29.2.el6 (RHEL6.5)
216 2.6.32-504.30.3.el6 (RHEL6.6)
217 2.6.32-573.26.1.el6 (RHEL6.7)
218 2.6.32-642.15.1.el6 (RHEL6.8)
219 2.6.32-696.3.1.el6 (RHEL6.9)
220 3.10.0-514.21.1.el7 (RHEL7.3)
221 3.0.101-0.47.71 (SLES11 SP3)
222 3.0.101-100 (SLES11 SP4)
223 3.12.74-60.64.40 (SLES12 SP1)
224 4.4.59-92.17 (SLES12 SP2)
225 vanilla linux 4.6.7 (ZFS only)
226 * Client known to build on unpatched kernels:
227 2.6.32-431.29.2.el6 (RHEL6.5)
228 2.6.32-504.30.3.el6 (RHEL6.6)
229 2.6.32-573.26.1.el6 (RHEL6.7)
230 2.6.32-642.15.1.el6 (RHEL6.8)
231 2.6.32-696.3.1.el6 (RHEL6.9)
232 3.10.0-514.21.1.el7 (RHEL7.3)
233 3.0.101-0.47.71 (SLES11 SP3)
234 3.0.101-100 (SLES11 SP4)
235 3.12.74-60.64.40 (SLES12 SP1)
236 4.4.59-92.17 (SLES12 SP2)
238 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
239 * Recommended ZFS / SPL version: 0.6.5.9
240 * Tested with ZFS / SPL version: 0.6.5.9
241 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
242 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
243 stack overflow. For more information, please refer to bugzilla 17630.
244 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
245 the re-exporting nodes to be mounted with "32bitapi" mount option
247 --------------------------------------------------------------------------------
249 12-07-2016 Intel Corporation
251 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
252 for currently supported client and server kernel versions.
253 * Server known to build on patched kernels:
254 2.6.32-431.29.2.el6 (RHEL6.5)
255 2.6.32-504.30.3.el6 (RHEL6.6)
256 2.6.32-573.26.1.el6 (RHEL6.7)
257 2.6.32-642.11.1.el6 (RHEL6.8)
258 3.10.0-514.el7 (RHEL7.3)
259 3.0.101-0.47.71 (SLES11 SP3)
260 3.0.101-80 (SLES11 SP4)
261 3.12.62-60.64.8 (SLES12 SP1)
262 vanilla linux 4.5.7 (ZFS only)
263 * Client known to build on unpatched kernels:
264 2.6.32-431.29.2.el6 (RHEL6.5)
265 2.6.32-504.30.3.el6 (RHEL6.6)
266 2.6.32-573.26.1.el6 (RHEL6.7)
267 2.6.32-642.11.1.el6 (RHEL6.8)
268 3.10.0-514.el7 (RHEL7.3)
269 3.0.101-0.47.71 (SLES11 SP3)
270 3.0.101-80 (SLES11 SP4)
271 3.12.62-60.64.8 (SLES12 SP1)
273 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
274 * Recommended ZFS / SPL version: 0.6.5.8
275 * Tested with ZFS / SPL version: 0.6.5.8
276 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
277 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
278 stack overflow. For more information, please refer to bugzilla 17630.
279 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
280 the re-exporting nodes to be mounted with "32bitapi" mount option
282 --------------------------------------------------------------------------------
284 02-29-2016 Intel Corporation
286 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
287 for currently supported client and server kernel versions.
288 * Server known to build on patched kernels:
289 2.6.32-431.29.2.el6 (RHEL6.5)
290 2.6.32-504.30.3.el6 (RHEL6.6)
291 2.6.32-573.12.1.el6 (RHEL6.7)
292 3.10.0-327.3.1.el7 (RHEL7.2)
293 3.0.101-0.47.71 (SLES11 SP3)
294 3.0.101-68 (SLES11 SP4)
295 vanilla linux 4.2.1 (ZFS only)
296 * Client known to build on unpatched kernels:
297 2.6.32-431.29.2.el6 (RHEL6.5)
298 2.6.32-504.30.3.el6 (RHEL6.6)
299 2.6.32-573.12.1.el6 (RHEL6.7)
300 3.10.0-327.3.1.el7 (RHEL7.2)
301 3.0.101-0.47.71 (SLES11 SP3)
302 3.0.101-68 (SLES11 SP4)
305 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
306 * Recommended ZFS / SPL version: 0.6.4.2
307 * Tested with ZFS / SPL version: 0.6.4.2
308 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
309 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
310 stack overflow. For more information, please refer to bugzilla 17630.
311 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
312 the re-exporting nodes to be mounted with "32bitapi" mount option
314 --------------------------------------------------------------------------------
316 03-10-2015 Intel Corporation
318 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
319 for currently supported client and server kernel versions.
320 * Server known to build on patched kernels:
321 2.6.32-431.29.2.el6 (RHEL6.5)
322 2.6.32-504.8.1.el6 (RHEL6.6)
323 3.0.101-0.46 (SLES11 SP3)
324 * Client known to build on unpatched kernels:
325 2.6.32-431.29.2.el6 (RHEL6.5)
326 2.6.32-504.8.1.el6 (RHEL6.6)
327 3.10.0-123.20.1.el7 (RHEL7)
328 3.0.101-0.46 (SLES11 SP3)
329 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
330 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
331 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
332 stack overflow. For more information, please refer to bugzilla 17630.
333 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
334 the re-exporting nodes to be mounted with "32bitapi" mount option
336 Severity : enhancement
338 Description: control OST-index in IDIF via ROCOMPAT flag.
339 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
340 last_rcvd file. For new formatted OST device, it will be auto set;
341 for the case of upgrading from old OST device, you can enable it
342 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
343 enabled, for new created OST-object, its IDIF-in-LMA will contain
344 the OST-index; for the existing OST-object, the OSD will convert
345 old format IDIF as new format IDIF with OST-index stored in the
346 LMA EA when accessing such OST-object or via OI scrub. Once such
347 flag is enabled, it cannot be reverted back, so the system cannot
348 be downgraded to the orignal incompatible version.
350 --------------------------------------------------------------------------------
352 07-30-2014 Intel Corporation
354 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
355 for currently supported client and server kernel versions.
356 * Server known to build on patched kernels:
357 2.6.32-431.20.3.el6 (RHEL6.5)
358 3.0.101-0.31 (SLES11 SP3)
359 * Client known to build on unpatched kernels:
360 2.6.32-431.20.3.el6 (RHEL6.5)
361 3.0.101-0.31 (SLES11 SP3)
364 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
365 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
366 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
367 stack overflow. For more information, please refer to bugzilla 17630.
368 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
369 the re-exporting nodes to be mounted with "32bitapi" mount option
371 --------------------------------------------------------------------------------
373 10-31-2013 Intel Corporation
375 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
376 * Server builds for kernels:
377 2.6.32-358.18.1.el6 (RHEL6.4)
378 * Client builds for unpatched kernels:
379 2.6.32-358.18.1.el6 (RHEL6.4)
380 3.0.92-0.8 (SLES11 SP3)
382 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
383 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
384 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
385 stack overflow. For more information, please refer to bugzilla 17630.
386 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
387 the re-exporting nodes to be mounted with "32bitapi" mount option
389 --------------------------------------------------------------------------------
391 05-30-2013 Intel Corporation
393 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
394 * Server builds for kernels:
395 2.6.32-358.6.2.el6 (RHEL6.4)
396 * Client builds for unpatched kernels:
397 2.6.32-358.6.2.el6 (RHEL6.4)
398 3.0.74-0.6.6 (SLES11 SP2)
400 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
401 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
402 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
403 stack overflow. For more information, please refer to bugzilla 17630.
404 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
405 the re-exporting nodes to be mounted with "32bitapi" mount option
407 --------------------------------------------------------------------------------
409 9-30-2012 Whamcloud, Inc.
411 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
412 * Server support for kernels:
413 2.6.32-279.5.1.el6 (RHEL6)
414 * Client support for unpatched kernels:
415 2.6.32-279.5.1.el6 (RHEL6)
417 2.6.18-238.19.1.el5 (RHEL5)
418 2.6.16.60-0.69.1 (SLES10)
419 * Recommended e2fsprogs version: 1.41.90.wc4
420 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
421 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
422 stack overflow. For more information, please refer to bugzilla 17630.
423 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
424 the re-exporting nodes to be mounted with "32bitapi" mount option
426 Severity : enhancement
428 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
430 Severity : enhancement
432 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
434 Severity : enhancement
436 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
438 Severity : enhancement
440 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
442 Severity : enhancement
444 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
446 --------------------------------------------------------------------------------
448 04-30-2012 Whamcloud, Inc.
450 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
451 * Server support for kernels:
452 2.6.32-220.4.2.el6 (RHEL6)
453 * Client support for unpatched kernels:
454 2.6.32-220.4.2.el6 (RHEL6)
456 2.6.18-238.19.1.el5 (RHEL5)
457 2.6.16.60-0.69.1 (SLES10)
458 * Recommended e2fsprogs version: 1.41.90.wc4
459 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
460 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
461 stack overflow. For more information, please refer to bugzilla 17630.
462 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
463 the re-exporting nodes to be mounted with "32bitapi" mount option
465 Severity : enhancement
467 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
469 Severity : enhancement
471 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
473 Severity : enhancement
475 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
477 Severity : enhancement
479 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
481 --------------------------------------------------------------------------------
483 09-30-2011 Whamcloud, Inc.
485 * Server support for kernels:
486 2.6.18-238.19.1.el5 (RHEL5)
487 2.6.32-131.6.1.el6 (RHEL6)
488 * Client support for unpatched kernels:
489 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
490 2.6.18-238.19.1.el5 (RHEL5)
491 2.6.32-131.6.1.el6 (RHEL6)
492 2.6.29.4-167.fc11 (FC11)
493 2.6.16.60-0.69.1 (SLES10)
495 * Recommended e2fsprogs version: 1.41.90.wc3
496 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
497 Lustre filesystem with 4K stack may cause a stack overflow. For more
498 information, please refer to bugzilla 17630.
499 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
500 mounted with "32bitapi" mount option
502 Severity : enhancement
503 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
505 Severity : enhancement
506 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
508 Severity : enhancement
509 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
512 Frequency : only with SLES10
514 Description: Use OFED "KMP" provided by Novell
515 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
516 Lustre is now built against this package. That means you need to
517 install the ofed-kmp package from Novell for the patchless client.
518 Note that the ofed-kmp that Novell ships may not exactly match the
519 kernel version but should still be compatible.
521 Severity : enhancement
523 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
524 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
526 --------------------------------------------------------------------------------
528 2010-07-15 Oracle, Inc.
530 * Server support for kernels:
531 2.6.18-164.11.1.el5 (RHEL 5)
532 2.6.18-164.11.1.0.1.el5 (OEL 5)
533 * Client support for unpatched kernels:
534 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
535 2.6.18-164.11.1.el5 (RHEL 5),
536 2.6.18-164.11.1.0.1.el5 (OEL 5)
537 2.6.16.60-0.42.8 (SLES 10),
539 2.6.29.4-167.fc11 (FC11)
540 * Recommended e2fsprogs version: 1.41.10-sun2
542 Severity : enhancement
544 Description: Distribute lfs_migrate script with lustre
545 Details : The lfs_migrate mentioned in the manual is now included in the
546 lustre tree in lustre/scripts and will distributed in lustre RPMs.
548 Severity : enhancement
550 Description: Failover nid list added to import proc file.
552 Severity : enhancement
554 Description: "som_preview" mount option required to use SOM
555 Details : The clients will need to use som_preview mount option
559 Description: Lustre does not do 1MB IOs to HW RAID
560 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
562 Severity : enhancement
564 Description: lfs quota output is not very convenient for awk/sed-parsing
565 Details : Some positions in lfs quota output table could be empty or
566 non-empty which made it hard to parse it with scripts, now
567 a dash is put instead of space where there is not supposed
572 Description: Update to OFED 1.5.1
573 Details : OFED 1.5.1 will be used in environments that don't use
574 vendor supplied OFED.
578 Description: Remove lnet. prefix from lctl params display
579 Details : Also renamed lnet.memused to lnet_memused to
584 Description: Use conf_param -d to remove permanant settings.
588 Description: Add list_param -R
589 Details : lctl list_param now supports recursive mode.
592 Bugzilla : 21829, 22312
593 Description: Fix llobdstat.
594 Details : 1. Now updating the write_bytes counter
595 2. Port script fixes from 1.8.
599 Description: decrease the usage of memory on clients.
600 Details : 1. On clients, recycle dentries and inodes unused.
601 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
606 Description: llobdstat fix and enhancement
607 Details : add a counter to set a limit to how many samples will be returned
608 fix a wildcard in the path to limit to obdfilter stats only
612 Description: lfs check only allowed for root
613 Details : Code cleanups and added a regresion test of non-root use of
618 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
619 Details : Fix hr->hr_index overload.
623 Description: Fix FIEMAP functionality.
627 Description: FC11 patchless client support added
631 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
632 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
636 Description: New acc-sm script and various test-framework enhancments.
640 Description: Previously landed patch was reverted.
644 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
648 Description: Add ldiskfs maxdirsize mount option.
652 Description: Fix mkfs.lustre support for 16TB LUNs.
656 Description: Print mmp_check_interval
657 Details : print mmp_check_interval and make it possible to abort
658 mount operation in case it takes too long.
662 Description: Add cascading_rw to lustre mpi tests.
666 Description: Store SOM data in EA.
667 Details : Also several SOM related cleanups.
671 Description: Add refcount to md_open_data
675 Description: Use refcount on lov_request_set to prevent memory corruption.
679 Description: Statahead: use parent dir's i_mutex for synchronization.
683 Description: loadgen improvements.
687 Description: Enable peer health detection in ksocklnd and ko2iblnd.
691 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
695 Description: Fix race on obd_devcie:md_stats
699 Description: Fix SLES 11 client builds.
700 Details : Resolved uninitalized pointer gcc warnings.
704 Description: Allow non-root access for "lfs check".
705 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
709 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
711 Severity : enhancement
713 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
714 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
718 Description: Sanity test_27p fails.
719 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
720 Sanity tests 27 were cleaned up and improved.
724 Description: Unify req format on client/servers
725 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
726 buffers and swabbers approach..
731 Description: drop join file stuff.
733 Severity : enhancement
735 Description: Add 2.6.27 and SLES11 (patchless client) support
739 Description: Fix in ptlrpc_expire_one_request() to print the signed time
742 Severity : enhancement
744 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
747 Severity : enhancement
749 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
751 Severity : enhancement
753 Description: correctly handle big reply message.
754 Details : send LNet event if reply is bigger than buffer and adjust this buffer
759 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
761 Severity : enhancement
763 Description: Add support for OEL5.
767 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
769 Severity : enhancement
771 Description: "weak-modules" support
772 Details : Implement "weak-modules" support which enables kernel modules
773 to be used with any kernel that implements the same kABI. In
774 order to achieve this modules are now installed in
775 /lib/modules/$(uname -r)/updates/kernel on all distributions.
779 Description: client could not reconnect to OST because of an active request.
780 Details : The client now retries to reconnect to the same server, if a
781 connect request failed with EBUSY or -EAGAIN. Test case 35b
782 conf-sanity.sh has been created.
786 Description: Don't allow make backward step on assignin osc next id.
787 Details : race between allocation next id and ll_sync thread can be cause
788 of set wrong osc next id and can be kill valid ost objects.
792 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
796 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
800 Description: Error handling in osc_statfs_interpret() has been improved.
801 Details : Check in osc_statfs_interpret() for EBADR.
805 Description: Do not send statfs() requests to OSTs disabled by administrator.
806 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
808 Severity : enhancement
810 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
812 Severity : enhancement
814 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
816 Severity : enhancement
818 Description: Enhancement for lfs(1) command to use numeric uid/gid.
822 Description: Improved cleanup procedure for conf-sanity test_46a.
826 Description: Do not start bulk transfer, if client tried to reconnect.
827 Details : Do not start bulk transfer, if client tried to reconnect,
828 but failed due to active rpcs. Check for exp_abort_active_req
829 has been implemented.
833 Description: Prevent inconsistences between linux and lustre mount structures.
834 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
835 Make the sleep interruptible.
837 Severity : enhancement
839 Description: provide server to client comms path
840 Details : server to client communications path via new LDLM_SET_INFO rpc
844 Description: truncate starts GFP_FS allocation under transaction causing deadlock
845 Details : ldiskfs_truncate calls grab_cache_page which may start page
846 allocation under an open transaction. This may lead to
847 calling prune_icache with consequent lustre reentrance.
851 Description: send correctly lsm on open replay
852 Details : MDS is trust to LSM size on replay open, but client can set wrong size
857 Description: enable client interface failover
858 Details : When a child reconnects from another NID, properly update export
859 nid hash position and ldlm reverse import.
861 Severity : enhancement
863 Description: Communicate OST degraded/readonly state via statfs to MDS
864 Details : Flags in the statfs returned from OSTs indicate whether the
865 OST is in a degraded RAID state, or if the filesystem has
866 turned read-only after a filesystem error is detected.
871 Description: don't panic if EPROTO was hit when reading symlink
872 Details : correctly handling request reference in error cases.
874 Severity : enhancement
876 Description: Add LustreNetLink, a kernel-userspace communcation path.
878 Severity : enhancement
880 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
885 Description: access to llog context before init.
886 Details : move handling CATALOGS file at osc layer and forbid access to llog
890 Frequency : rare, only when client is writing to OST on local node
892 Description: Avoid deadlock for local client writes
893 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
894 memory freeing context. This allows OST threads to set the
895 PF_MEMALLOC flag on task structures in order to allocate memory
896 from reserved pools and complete IO.
897 Use GFP_HIGHUSER for OST allocations for non-local client writes,
898 so that the OST threads generate memory pressure and allow
899 inactive pages to be reclaimed.
901 Severity : enhancement
903 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
908 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
909 Details : move ldlm namespace creation in setup phase to avoid grab
910 _lprocfs_lock with cli_sem held.
914 Description: Temporarily disable grant shrink.
915 Details : Disable the feature for debugging.
918 Frequency : only when formatting test filesystems
920 Description: Unable to run several mkfs.lustre on loop devices at the same time
921 Details : mkfs.lustre returns error 256 on the concurrent loop devices
922 formatting. The solution is to proper handle the error.
924 Severity : enhancement
926 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
928 Severity : enhancement
930 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
935 Description: resolve race between obd_disconnect and class_disconnect_exports
936 Details : if obd_disconnect will be called to already disconnected export he
937 forget release one reference and osc module can't unloaded.
939 Severity : enhancement
941 Description: Allow tuning service thread via /proc
942 Details : For each service a new
943 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
944 created that can be used to set min/max thread counts, and get the
945 current number of running threads.
950 Descriptoin: don't return error if have partially created objects for file.
951 Details : lov_update_create_set uses set->set_success as index for created
952 objects, so if some requests will be failed, they will have hole
953 at end of array and qos_shrink_lsm for allocate correct lsm.
955 Severity : enhancement
957 Description: Update OFED support to 1.4.1
962 Description: Reduce small size read RPC
963 Details : Set read-ahead limit for every file and only do read-ahead when
964 available read-ahead pages are bigger than 1M to avoid small size
967 Severity : enhancement
969 Description: move AT tunable parameters for more consistent usage
970 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
972 Severity : enhancement
974 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
975 Details : allow skip disconnected ost for send statfs request and hide error
981 Description: fix racy locking of mballoc block bitmaps causing BUG
982 Details : The locking of the mballoc buddy bitmap and the in-memory
983 block bitmap was using two different spin locks in some
984 cases. This made it possible to incorrectly access the
985 mballoc bitmap while another process was modifying it,
986 causing a sanity assertion to fail. While no on-disk corruption
987 was reported, there was some risk of this happening.
990 Frequency : rare, on llog test 6
992 Descriptoin: don't allow connect to already connected import
993 Details : allowing connect to already connected import is hide connecting problem.
998 Descriptoin: allow kill process which wait statahead result
999 Details : for some reasons 'ls' can stick in waiting result from statahead,
1000 in this case need way for kill this process.
1002 Severity : enhancement
1004 Description: Add state history info file, enhance import info file
1005 Details : Track import connection state changes in a new osc/mdc proc file;
1006 add overview-type data to the osc/mdc import proc file.
1008 Severity : enhancement
1010 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1012 Severity : enhancement
1014 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1019 Descriptoin: OST grant too much space to client even there are not enough space.
1020 Details : Client will shrink its grant cache to OST if there are no write
1021 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1022 this grant cache if there are already not enough avaible space
1023 (left_space < total_clients * 32M).
1026 Frequency : start MDS on uncleanly shutdowned MDS device
1028 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1029 Details : stay in waiting mds<>ost recovery finished produce random bugs
1030 due race between two ll_sync thread for one lov target. send
1031 ACTIVATE event only if connect realy finished and import have
1035 Frequency : rare, connect and disconnect target at same time
1037 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1038 Details : don't call obd_disconnect under lov_lock. this long time
1039 operation and can block ptlrpcd which answer to connect request.
1044 Descriptoin: don't lose wakeup for imp_recovery_waitq
1045 Details : recover_import_no_retry or invalidate_import and import_close can
1046 both sleep on imp_recovery_waitq, but we was send only one wakeup
1050 Frequency : always with long access acl
1052 Descriptoin: mds can't pack reply with long acl.
1053 Details : mds don't control size of acl but they limited by reint/getattr
1056 Severity : enhancement
1058 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1060 Severity : enhancement
1062 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1065 Frequency : start MDS on uncleanly shutdowned MDS device
1067 Descriptoin: aborting recovery hang on MDS
1068 Details : don't throttle destroy RPCs for the MDT.
1071 Frequency : on remount
1073 Description: external journal device not working after the remount
1074 Details : clear dev_rdonly flag for external journal devices in
1080 Description: shutdown vs evict race
1081 Details : client_disconnect_export vs connect request race.
1082 if client will evicted at this time - we start invalidate
1083 thread without referece to import and import can be freed
1087 Frequency : rare, need acl's on inode.
1089 Description: client can't handle ost additional correctly
1090 Details : if ost was added after client connected to mds client can have
1091 hit lnet_try_match_md ... to big messages to wide striped files.
1092 in this case need teach client to handle config events about add
1093 lov target and update client max ea size at that event.
1095 Severity : enhancement
1097 Description: Changelogs
1098 Details : Changelogs are a lightweight mechanism to track filesystem
1099 metadata and namespace changes. The changelog is recorded
1100 permanently on the MDTs, and is periodically "consumed" / purged
1101 when records are no longer needed.
1103 Severity : enhancement
1105 Description: compact fld format with extents
1106 Details : Store range of seq rather than every seq in FLD. Seq
1107 controller update FLD rather than clients. In Case of CMD, mdt0
1108 has FLD, all other metadata server act as non persistent proxy
1109 for FLD queries and cache fld entries in fld cache.
1114 Description: don't skip ost target if they assigned to file
1115 Details : Drop slow OSCs if we can, but not for requested start idx.
1116 This means "if OSC is slow and it is not the requested
1117 start OST, then it can be skipped, otherwise skip it only
1118 if it is inactive/recovering/out-of-space.
1122 Description: more cleanup in mds_lov
1123 Details : not send LOV EA under replay, we can't know about they size at this
1124 time. Don't allow client connect to mds before any ost connected,
1125 for avoid problems with LOV EA size and returning EIO to client.
1127 Severity : enhancement
1129 Description: Interoperability at server side (Disk interoperability)
1131 Severity : enhancement
1133 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1135 Severity : enhancement
1137 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1139 Severity : enhancement
1141 Description: New client IO stack (CLIO).
1143 Severity : enhancement
1145 Description: Commit on sharing. Eliminate inter-client dependencies between
1146 uncommitted transactions by doing transaction commits.
1147 Thereby clients may recovery independently.
1150 Frequency : Create a symlink file with a very long name
1152 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1153 Details : If there is no extra space in the request for early cancels,
1154 ldlm_req_handles_avail() returns 0 instead of a negative value.
1156 Severity : enhancement
1158 Description: Add /proc entry for import status
1159 Details : The mdc, osc, and mgc import directories now have
1160 an import directory that contains useful import data for debugging
1161 connection problems.
1163 Severity : enhancement
1165 Description: Re-disable certain /proc logging
1166 Details : Enable and disable client's offset_stats, extents_stats and
1167 extents_stats_per_process stats logging on the fly.
1170 Frequency : Only on FC kernels 2.6.22+
1172 Description: oops in statahead
1173 Details : Do not drop reference count for the dentry from VFS when lookup,
1174 VFS will do that by itself.
1176 Severity : enhancement
1178 Description: Generic /proc file permissions
1179 Details : Set /Proc file permissions in a more generic way to enable non-
1180 root users operate on some /proc files.
1184 Description: Hitting mdc_commit_close() ASSERTION
1185 Details : Properly handle request reference release in
1186 ll_release_openhandle().
1190 Description: quota recovery deadlock during mds failover
1191 Details : This patch includes att18982, att18236, att18237 in bz14840.
1193 1. fix osts hang when mds does failover with quotaon
1194 2. prevent watchdog storm when osts threads wait for the
1199 Frequency : only patchless client
1200 Description: add workaround for race between add/remove dentry from hash
1202 Severity : enhancement
1204 Description: Allow OST glimpses to return PW locks
1208 Description: LBUG when llog conf file is full
1209 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1213 Description: Prevent import from entering FULL state when server in recovery
1217 Description: service mount cannot take device name with ":"
1218 Details : Only when device name contains ":/" will mount treat it as
1224 Description: replace ptlrpcd with the statahead thread to interpret the async
1225 statahead RPC callback
1229 Frequency : on recovery
1230 Description: I/O failures after umount during fail back
1231 Details : if client reconnected to restarted server we need join to recovery
1232 instead of find server handler is changed and process self
1233 eviction with cancel all locks.
1235 Severity : enhancement
1237 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1239 Severity : enhancement
1241 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1243 Severity : enhancement
1245 Description: Upcall on Lustre log has been dumped
1246 Details : Allow for a user mode script to be called once a Lustre log has
1247 been dumped. It passes the filename of the dumped log to the
1248 script, the location of the script can be specified via
1249 /proc/sys/lnet/debug_log_upcall.
1254 Description: avoid idr_remove called for id which is not allocated.
1255 Details : Move assigment s_dev for clustered nfs to end of initialization,
1256 to avoid problem with error handling.
1261 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1262 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1263 into CONN_USED_HASH and this prodice warning when put connection
1264 again in unused hash.
1267 Severity : enhancement
1269 Description: File striping can now be set to use an arbitrary pool of OSTs.
1271 Severity : enhancement
1273 Description: Export bytes_read/bytes_write count on OSC/OST.
1277 Description: Early reply size mismatch, MGC loses connection
1278 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1279 the connect flags are properly negotiated.
1284 Description: quotas are not honored with O_DIRECT
1285 Details : all writes with the flag O_DIRECT will use grants which leads to
1286 this problem. Now using OBD_BRW_SYNC to guard this.
1290 Description: add quota statistics
1291 Details : 1. sort out quota proc entries and proc code.
1292 2. add quota statistics
1294 Severity : enhancement
1296 Description: enable quota support for HEAD.
1300 Description: Properly propagate oinfo flags from lov to osc for statfs
1301 Details : restore missing copy oi_flags to lov requests.
1303 Severity : enhancement
1305 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1306 : l_getgroups(8), lst(8), routerstat(8)
1307 Details : included man pages for llobdstat(8), llstat(8),
1308 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1310 Severity : enhancement
1312 Description: configure's --enable-quota should check the
1313 : kernel .config for CONFIG_QUOTA
1314 Details : configure is terminated if --enable-quota is passed but
1315 : no quota support is in kernel
1319 Description: Remove portals compatibility
1320 Details : Remove portals compatibility, not interoperable with releases
1325 Description: Resolve device initialization race
1326 Details : Prevent proc handler from accessing devices added to the
1327 obd_devs array but yet be intialized.
1329 Severity : enhancement
1331 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1333 Severity : enhancement
1335 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1340 Description: Using wrong pointer in osc_brw_prep_request
1341 Details : Access to array[-1] can produce panic if kernel compiled with
1342 CONFIG_PAGE_ALLOC enabled
1346 Description: Client runs out of low memory
1347 Details : Consider only lowmem when counting initial number of llap pages
1351 Description: *optional* service tags registration
1352 Details : if the "service tags" package is installed on a Lustre node
1353 When the filesystem is mounted, a local-node service tag will
1354 be created. See http://inventory.sun.com/ for more information
1355 about the Service Tags asset management system.
1359 Description: Kernel BUG tries to release flock
1360 Details : Lustre does not destroy flock lock before last reference goes
1361 away. So always drop flock locks when client is evicted and
1362 perform unlock regardless of successfulness of speaking to MDS.
1366 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1370 Description: sanity test 65a fails if stripecount of -1 is set
1371 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1376 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1377 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1380 Severity : enhancement
1382 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1386 Description: do not process already freed flock
1387 Details : flock can possibly be freed by another thread before it reaches
1388 to ldlm_flock_completion_ast.
1392 Description: LBUG during stress test
1393 Details : Need properly lock accesses the flock deadlock detection list.
1397 Description: oops in page fault handler
1398 Details : kernel page fault handler can return two special 'pages' in error
1399 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1403 Description: timeout with invalidate import.
1404 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1405 should be handled by ptlrpcd. This produce long age waiting and
1406 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1408 Severity : enhancement
1410 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1414 Description: enable MGS and MDT services start separately
1415 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1416 a co-located MGS without starting the MGS, which is a complement
1417 to 'nosvc' mount option.
1421 Frequency : after recovery
1422 Description: precreate to many object's after del orphan.
1423 Details : del orphan st in oscc last_id == next_id and this triger growing
1424 count of precreated objects. Set flag LOW to skip increase count
1425 of precreated objects.
1429 Frequency : rare, on clear nid stats
1430 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1431 Details : when clean nid stats sometimes try destroy live entry,
1432 and this produce panic in free.
1436 Description: Stack overflow during MDS log replay
1437 ease stack pressure by using a thread dealing llog_process.
1441 Description: wait until IO finished before start new when do lock cancel.
1442 Details : VM protocol want old IO finished before start new, in this case
1443 need wait until PG_writeback is cleared until check dirty flag and
1444 call writepages in lock cancel callback.
1446 Severity : enhancement
1448 Description: using special macro for print time and cleanup in includes.
1452 Description: mds_mfd_close() ASSERTION(rc == 0)
1453 Details : In mds_mfd_close(), we need protect inode's writecount change
1454 within its orphan write semaphore to prevent possible races.
1458 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1462 Frequency : rare, on shutdown ost
1463 Description: don't hit live lock with umount ost.
1464 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1465 use shrink_dcache_sb instead.
1469 Description: don't panic with use echo client
1470 Details : echo client pass NULL as client nid pointer and this produce null
1471 pointer dereference.
1475 Description: fix build on ppc32
1476 Details : compile code with -m64 flag produce wrong object file for ppc32.
1480 Description: add message levels for liblustreapi
1484 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1485 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1486 will result in an -EAGAIN and in the exit call path we call
1487 alloc_rr() which will with increasing aggressiveness attempt to
1488 aquire precreated objects on the minimum number of required OSCs.
1492 Description: Use old size assignment to avoid deadlock
1493 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1494 the scheduling while holding a spinlock. We do not need locking
1495 for size in ll_update_inode() because size is only updated from
1496 the MDS for directories or files without objects, so there is no
1497 other place to do the update, and concurrent access to such inodes
1498 are protected by the inode lock.
1502 Description: resolve "_IOWR redefined" build error on SLES10
1506 Description: dump the memory debugging after all modules are unloaded to
1507 suppress false negative in conf_sanity test 39
1509 Severity : enhancement
1511 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1516 Description: fix SLES kernel versioning
1517 Details : the kernel version for our SLES 10 kernel did not include a "-"
1518 before the "smp" at the end. while this was not a problem in
1519 general, it did mean that software trying to use the kernel
1520 version to try to detect a vendor specific kernel would fail.
1521 this was most evident by the OFED build scripts.
1525 Description: Don't update lov_desc members until making sure they are valid
1526 Details : When updating lov_desc members via proc fs, need fix their
1527 validities before doing the real update.
1531 Description: don't put request into delay list while invalidate in flight.
1532 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1533 import in flight. this produce timeout for invalidate and sometimes
1534 can cause stale data.
1538 Frequency : on ppc only
1539 Description: not convert ost objects for directory because it's not exist.
1540 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1544 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1546 Severity : enhancement
1548 Description: Update to support 2.6.22.14 vanilla kernel.
1552 Frequency : rare, on recovery
1553 Description: read procfs can produce deadlock in some situation
1554 Details : Holding lprocfs lock which send rpc can produce block for destroy
1555 obd objects and this also block reconnect with -EALREADY. This
1556 isn't fix all lprocfs bugs - but make it rare.
1558 Severity : enhancement
1560 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1563 Frequency : frequent on X2 node
1565 Description: mdc_set_open_replay_data LBUG
1566 Details : Set replay data for requests that are eligible for replay.
1570 Description: lustre_mgs: operation 101 on unconnected MGS
1571 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1572 MGC, and late on MGC cannot successfully connect to MGS and a lot
1573 of the error messages complaining that MGS is not connected.
1577 Frequency : on network error
1578 Description: panic with double free request if network error
1579 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1580 but it's true only for synchronus enqueue, for async enqueue
1581 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1584 Severity : enhancement
1586 Description: client-side metadata stat-ahead during readdir(directory readahead)
1587 Details : perform client-side metadata stat-ahead when the client detects
1588 readdir and sequential stat of dir entries therein
1591 Frequency : on start mds
1593 Description: Implement get_info(last_id) in obdfilter.
1596 Frequency : occasional
1598 Description: Correctly check stale fid, not start epoch if ost not support SOM
1599 Details : open with flag O_CREATE need set old fid in op_fid3 because
1600 op_fid2 was overwritten with new generated fid, but mds can answer
1601 with one of these two fids and both is not stale. Setattr
1602 incorrectly started an epoch and assume will be called
1603 done_writing, but without SOM done_writing ever being called.
1606 Frequency : rare, depends on device drivers and load
1608 Description: MDS or OSS nodes crash due to stack overflow
1609 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1610 In some cases, in conjunction with device drivers that use a lot
1611 of stack the MDS (or possibly OSS) service threads could overflow
1612 the stack. One change which was identified to consume additional
1613 stack has been reworked to avoid the extra stack usage.
1616 Frequency : occasional
1618 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1619 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1620 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1621 oscc_has_objects and see OSC has no objests and call
1622 oscc_internal_create to resend create request.
1624 Severity : enhancement
1626 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1628 Severity : enhancement
1630 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1633 Frequency : very rare
1635 Description: Fix replay if there is an un-replied request and open
1636 Details : In some cases, older replay request will revert the
1637 mcd->mcd_last_xid on MDS which is used to record the client's
1638 latest sent request.
1640 Severity : enhancement
1642 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1644 Severity : enhancement
1646 Description: Add rhel5 support to HEAD.
1648 Serverity : enhancement
1650 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1655 Description: Don't allow skipping OSTs if index has been specified.
1656 Details : Don't allow skipping OSTs if index has been specified, make
1657 locking in internal create lots better.
1661 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1662 Details : access to bitfield in structure is always rounded to long
1663 and this produce problem with not atomic change any bit.
1667 Description: Lustre make rpms failed.
1668 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1669 builds Lustre rpms from tarball.
1672 Frequency : rare on shutdown ost
1674 Description: If llog cancel was not send before clean_exports phase, this can
1675 produce deadlock in llog code.
1676 Details : If llog thread has last reference to obd and call class_import_put
1677 this produce deadlock because llog_cleanup_commit_master wait when
1678 last llog_commit_thread exited, but this never success because was
1679 called from llog_commit_thread.
1683 Description: allow userland application know is lost one of stripes.
1684 Details : fill lvb_blocks with error code on ost and return it to
1685 application if error flag found.
1689 Description: NULL lov_tgts causing MDS oops
1690 Details : more safe checks for NULL lov_tgts for avoid oops.
1692 Severity : enhancement
1694 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1698 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1699 Details : If an OST is down the MDS will hang indefinitely in
1700 obd_statfs_rqset() waiting for the statfs data. While for
1701 MDS QOS usage of statfs, it should not stuck in waiting.
1703 Severity : enhancement
1705 Description: remote_acl support
1706 Details : Support ACL-based permission check for remote user.
1707 Support setfacl/getfacl for remote user with the utils
1708 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1709 the system "{s,g}etfacl" utils.
1711 Severity : enhancement
1713 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1715 Severity : enhancement
1717 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1721 Description: make mgs_setparam() handle fsname containing dash
1722 Details : fsname containing a dash does not work with lctl conf_param
1724 Severity : enhancement
1726 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1728 Severity : enhancement
1730 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1734 Frequency : rare, at shutdown
1735 Description: access already free / zero obd_namespace.
1736 Details : if client_disconnect_export was called without force flag set,
1737 and exist connect request in flight, this can produce access to
1738 NULL pointer (or already free pointer) when connect_interpret
1739 store ocd flags in obd_namespace.
1743 Frequency : only at startup
1744 Description: not alloc memory with spinlock held.
1745 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1746 if any spinlock held.
1748 Severity : enhancement
1750 Description: make lustre randomly fail allocating memory
1751 Details : Make lustre randomly failed allocating memory for testing purpose.
1753 Severity : enhancement
1755 Description: lost problems with lov objid file
1756 Details : Fixes some scability and access to not inited memory problems
1757 in work with lov objdid file.
1762 Description: lfs find does not continue on file error
1763 Details : Continue other files processing when a file/dir is absent.
1767 Description: Inconsistent usage of lustre_pack_reply()
1768 Details : Standardize the usage of lustre_pack_reply() such that it
1769 always generate a CERROR on failure.
1772 Frequency : occasional
1774 Description: MDS hang or stay in waiting lock
1775 Details : If client receive lock with CBPENDING flag ldlm need send lock
1776 cancel as separate rpc, to avoid situation when cancel request
1777 can't processed due all i/o threads stay in wait lock.
1781 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1785 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1789 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1793 Description: Update to RHEL4 latest kernel.
1795 Severity : enhancement
1797 Description: Build SLES10 patchless client fails
1798 Details : The configure was broken by run ./configure with
1799 --with-linux-obj=.... argument for patchless client. When the
1800 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1801 can't search header adequately. Use absolute path such as
1802 -I($LINUX)/include instead.
1806 Description: interrupt oig_wait produce painc on resend.
1807 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1808 private set, and this produce situation when rq_ptlrpcd_data not
1809 copyed to new allocated request and triggered LBUG on assert
1810 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1811 ptlrpcd set if request is changed and can be safety changed to use
1816 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1817 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1818 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1819 these macros and replace them with their difinitions instead.
1823 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1824 Details : Update which_patch & target file for SLES10 latest kernel.
1827 Description: Remove client patches from SLES 10 kernel.
1828 Details : This causes SLES 10 clients to behave as patchless clients
1829 even on a Lustre-patched (server) kernel.
1831 Severity : enhancement
1833 Description: self-adjustable client's lru lists
1834 Details : use adaptive algorithm for managing client cached locks lru
1835 lists according to current server load, other client's work
1836 pattern, memory activities, etc. Both, server and client
1837 side namespaces provide number of proc tunables for controlling
1840 Severity : enhancement
1842 Description: light-weight GSS support
1843 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1844 and reduce performance overhead.
1846 Severity : enhancement
1848 Description: Linux keyring support
1849 Details : Support using service of Linux keyring for Lustre GSS internal
1850 context refresh/cache mechanism.
1854 Description: Fix errors in lfs documentation
1855 Details : Fixes man pages
1859 Description: don't use GFP_* in generic Lustre code.
1860 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1864 Description: obdclass is limited by single OBD_ALLOC(idarray)
1865 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1869 Description: warnings with build patchless client with vanila 2.6.19 and up
1870 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1871 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1875 Frequency : Only for SLES
1876 Description: sanity_quota fail test_1
1877 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1878 file, which makes TSTID[2] inunique.
1882 Frequency : Only for SLES9
1883 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1884 Details : Update target/ChangeLog/which_patch .
1888 Frequency : Only for test_10 in sanity-quota.sh
1889 Description: a bug in quota test code
1890 Details : checking a test flag in a wrong place
1894 Frequency : Only for test_13 in sanity-quota.sh
1895 Description: a bug in quota test script
1896 Details : shouldn't check overfull allocated quota
1900 Description: jbd statistics
1901 Details : Port older jbd statistics patch for sles10
1903 Severity : enhancement
1905 Description: Add printing inode info into message about error in writepage.
1910 Description: reply_lock_interpret crash due to race with it and lock cancel.
1911 Details : Do not replay locks that are being cancelled. Do not reference
1912 locks by their address during replay, just by their handle.
1916 Frequency : When flocks are used.
1917 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1918 Details : Flock locks might destroy just granted lock if it could be merged
1919 with another existing flock, this is done in completion handler,
1920 so teach ldlm_cli_enquque_fini that this is a valid case for
1926 Description: Oops in read and write path when failing to allocate lock.
1927 Details : Check if lock allocation failed and return error back.
1931 Description: lstripe command fails for valid OST index
1932 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1933 instead of lov->desc.ld_active_tgt_count.
1937 Description: sanity.sh failed test 103
1938 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1943 Description: df doesn't work properly if diskfs blocksize != 4K
1944 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1948 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1949 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1951 Severity : enhancement
1953 Description: Adaptive timeouts
1954 Details : RPC timeouts adapt to changing server load and network
1955 conditions to reduce resend attempts and improve recovery time.
1959 Description: llapi_file_create() does not allow some changes
1960 Details : add llapi_file_open() that allows specifying the mode and
1961 open flags, and also returns an open file handle.
1965 Description: merge and cleanup kernel patches.
1966 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1970 Description: Add journal checksum support.(Kernel part)
1971 Details : The journal checksum feature adds two new flags i.e
1972 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1973 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1974 indicates that the commit block contains the checksum for
1975 the blocks described by the descriptor blocks. Now commit
1976 record can be sent to disk without waiting for descriptor
1977 blocks to be written to disk. This behavior is controlled
1978 using JBD2_FEATURE_ASYNC_COMMIT flag.
1982 Description: OSS needs mutliple precreate threads
1983 Details : Add ability to start more than one create thread per OSS.
1987 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1988 Details : execute lfs setstripe on client
1992 Description: mds_obd_create error creating tmp object
1993 Details : When the user sets quota on root, llog will be affected and can't
1994 create files and write files.
1997 Frequency : Always on ia64 patchless client, and possibly others.
1999 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2000 Details : This allows the patchless client to be loaded on architectures
2001 without this export.
2005 Description: RedHat Update kernel for RHEL5
2006 Details : Add the kernel config file for RHEL5.
2010 Description: RedHat Update kernel for RHEL5
2011 Details : Modify the kernel config file more closer RHEL5.
2015 Description: Build failure against Centos5 (RHEL5)
2016 Details : Define PAGE_SIZE when it isn't present.
2020 Description: test_7 on sanity-quota.sh can't be used on separate machines
2021 Details : add facet to handle it
2025 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2026 Details : with NFS, the anon dentry's parent was set to itself in
2027 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2028 corresponding dentry other than use rec->ur_name.
2030 Severity : enhancement
2032 Description: lfs setstripe enhancement
2033 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2037 Description: enable data checksumming by default
2038 Details : enable checksum by default, allow --disable-checksum
2039 configure option and "-o nochecksum" mount option. Checksums
2040 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2044 Description: lustre support for RHEL5
2045 Details : Add support for RHEL5.
2049 Description: Client eviction due to failover config
2050 Details : after a connection loss, the lustre client should attempt to
2051 reconnect to the last active server first before trying the
2052 other potential connections.
2056 Description: when mds and osts use different quota unit(32bit and 64bit),
2057 quota will be released repeatly.
2058 Details : void sending multiple quota reqs to mds, which will keep the
2059 status between the reqs.
2063 Description: osts not allocated evenly to files
2064 Details : change the condition to increase offset_idx
2067 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2069 Description: Data corruption for OSTs that are formatted larger than 2TB
2071 Details : When generating the bio request for lustre file writes the
2072 sector number would overflow a temporary variable before being
2073 used for the IO. The data reads correctly from Lustre (which
2074 will overflow in a similar manner) but other file data or
2075 filesystem metadata may be corrupted in some cases.
2079 Description: Tune the kernel for good SCSI performance.
2080 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2081 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2086 Description: rewrite ext2-derived code in obdclass/uuid.c
2087 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2088 from scratch preserving functionality.
2092 Description: "lfs find -obd UUID" prints directories
2093 Details : "lfs find -obd UUID" will return all directory names instead
2094 of just file names. It is incorrect because the directories
2095 do not reside on the OSTs.
2099 Description: use i_size_read and i_size_write in 2.6 port
2100 Details : replace inode->i_size access with i_size_read/write()
2103 Frequency : when removing large files
2105 Description: scheduling issue during removal of large Lustre files
2106 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2107 It causes scheduling issues when removing large files (17TB in the
2111 Frequency : only with liblustre clients on XT3
2113 Description: evictions taking too long
2114 Details : allow llrd to evict clients directly on OSTs
2117 Frequency : only on ppc
2119 Description: /proc/fs/lustre/devices broken on ppc
2120 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2121 We should make sure the type of 'index' is loff_t and then cast
2122 explicitly as needed below. Do not assign an explicitly cast
2126 Frequency : only for rhel5
2128 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2129 Details : Modify the target file & which_kernel.
2131 Severity : enhancement
2133 Description: omit set fsid for export NFS
2134 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2135 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2139 Frequency : Always, for kernels after 2.6.16
2140 Description: Fix warning idr_remove called for id=.. which is not allocated.
2141 Details : Last kernels save old s_dev before kill super and not allow
2142 to restore from callback - restore it before call kill_anon_super.
2146 Description: buffer overruns could theoretically occur
2147 Details : llapi_semantic_traverse() modifies the "path" argument by
2148 appending values to the end of the origin string, and a
2149 overrun may occur. Adding buffer overrun check in liblustreapi.
2153 Description: Fix error on 'ls .' at the top of the Lustre mount.
2154 Details : Don't revalidate dentry if it is a root dentry.
2158 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2159 Details : Modify vm-tunables-rhel4.patch.
2163 Description: Kernel config for 2.6.18-vanilla.
2164 Details : Modify targets/2.6-vanilla.target.in.
2165 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2166 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2167 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2168 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2172 Description: improve handling recoverable errors
2173 Details : if request processig with error which can be recoverable on server
2174 request should be resend, otherwise page released from cache and
2179 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2180 Details : Modify target file & which_patch.
2181 A flaw was found in the IA32 system call emulation provided
2182 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2183 value could be stored in the %RAX register, which could trigger an
2184 out-of-bounds system call table access. An untrusted local user
2185 could exploit this flaw to run code in the kernel
2186 (ie a root privilege escalation). (CVE-2007-4573).
2190 Description: O_DIRECT bypasses client statistics.
2191 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2192 read_ahead_stats were updated. copied the stats section from
2193 osc_send_oap_rpc() into async_internal().
2197 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2201 Description: change order of libsysio includes
2202 Details : '#include sysio.h' should always come before '#include xtio.h'
2206 Description: handle "serious error: objid * already exists" more gracefully
2207 Details : If LAST_ID value on disk is smaller than the objects existing in
2208 the O/0/d* directories, it indicates disk corruption and causes an
2209 LBUG(). If the object is 0-length, then we should use the existing
2210 object. This will help to avoid a full fsck in most cases.
2212 Severity : enhancement
2214 Description: adapt the lustre_config script to support the upgrade case
2215 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2216 server targets to 1.6 in parallel.
2220 Description: To avoid grant space > avaible space when the disk is almost
2221 full. Without this patch you might see the error "grant XXXX >
2222 available" or some LBUG about grant, when the disk is almost
2224 Details : In filter_check_grant, for non_grant cache write, we should
2225 check the left space by if (*left > ungranted + bytes), instead
2226 of (*left > ungranted), because only we are sure the left space
2227 is enough for another "bytes", then the ungrant space should be
2228 increase. In client, we should update cl_avail_grant only there
2229 is OBD_MD_FLGRANT in the reply.
2233 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2234 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2235 described in CVE-2007-4573. This problem could allow untrusted
2236 local users to gain root access.
2239 Frequency : when using O_DIRECT and quotas
2241 Description: Incorrect file ownership on O_DIRECT output files
2242 Details : block usage reported by 'lfs quota' does not take into account
2243 files that have been written with O_DIRECT.
2248 Description: touch file failed when fs is not full
2249 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2250 otherwise we can get ENOSP while fs is not full.
2254 Description: parallel lock callbacks
2255 Details : Instead of sending blocking and completion callbacks as separated
2256 requests, adding them to a set and sending in parallel.
2259 Frequency : only for Cray XT3
2260 Bugzilla : 12829/13455
2261 Description: Changing primary group doesn't change the group lustre assigns to
2263 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2264 overridden with the primary group provided by the group upcall,
2265 whereas the supplied fsgid can be trusted if it is in the list of
2266 supplementary groups returned by the group upcall.
2268 Severity : enhancement
2270 Description: Allow masking D_WARNING, D_ERROR messages from console
2271 Details : Console messages can now be disabled via lnet.printk.
2275 Description: User code with malformed file open parameter crashes client node
2276 Details : Before packing join_file req, all the related reference should be
2277 checked carefully in case some malformed flags cause fake
2278 join_file req on client.
2282 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2283 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2284 mode is equal to the granted mode. However, it should also check
2285 that LDLM_FL_CP_REQD is not set.
2289 Description: Heavy nfs access might result in deadlocks
2290 Details : After ELC code landed, it is now improper to enqueue any mds
2291 locks under och_sem, because enqueue might want to decide to
2292 cancel open locks for same inode we are holding och_sem for.
2296 Description: Client eviction while running blogbench
2297 Details : A lot of unlink operations with concurrent I/O can lead to a
2298 deadlock causing evictions. To address the problem, the number of
2299 oustanding OST_DESTROY requests is now throttled to
2300 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2301 callbacks are processed in priority.
2305 Description: enable ACLs on MDS by default
2306 Details : ACLs must be enabled on MDS by default.
2309 Frequency : PPC/PPC64 only
2311 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2312 Details : fix duplicated definitions between asm-ppc64/types.h and
2313 lustre_types.h on PPC.
2316 Frequency : PPC/PPC64 only
2318 Description: asm-ppc/segment.h does not exist
2319 Details : fix compile issue on PPC.
2323 Description: better handle error messages in extents code
2326 Frequency : RHEL4 only
2328 Description: mkfs is very slow on IA64/RHEL4
2329 Details : A performance regression has been discovered in the MPT Fusion
2330 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2331 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2332 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2334 Severity : enhancement
2336 Description: SNMP support enhancement
2337 Details : Adding total number of sampled request for an MDS node in snmp
2340 Severity : enhancement
2342 Description: Optimize ldlm waiting list processing for PR extent locks
2343 Details : When processing waiting list for read extent lock and meeting read
2344 lock that is same or wider to it that is not contended, skip
2345 processing rest of the list and immediatelly return current
2346 status of conflictness, since we are guaranteed there are no
2347 conflicting locks in the rest of the list.
2351 Description: Time out and refuse to reconnect
2352 Details : When the failover node is the primary node, it is possible
2353 to have two identical connections in imp_conn_list. We must
2354 compare not conn's pointers but NIDs, otherwise we can defeat
2355 connection throttling.
2359 Description: port llog fixes from b1_6 into HEAD
2360 Details : Port llog reference couting and some llog cleanups from b1_6
2361 (bug 10800) into HEAD, for protect from panic and access to already
2362 free llog structures.
2366 Description: Detect stride IO mode in read-ahead
2367 Details : When a client does stride read, read-ahead should detect that and
2368 read-ahead pages according to the detected stride pattern.
2372 Description: data checksumming impacts single node performance
2373 Details : add support for several checksum algorithm. Currently, only CRC32
2374 and Adler-32 are supported. The checksum type can be changed on
2375 the fly via /proc/fs/lustre/osc/*/checksum_type.
2379 Description: use adler32 for page checksums
2380 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2385 Description: build for x2 fails
2386 Details : fix compile issue on Cray systems.
2390 Description: Properly match for duplicate locks
2391 Details : Due to different lock order from skiplists code, we need to
2392 traverse entire list for now
2395 Frequency : only on PPC/SLES10
2397 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2398 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2399 be defined. Add a hack in mkfs_lustre.c to work around this compile
2404 Description: LASSERT on MDS when client holding flock lock dies
2405 Details : ldlm pool logic depends on number of granted locks equal to
2406 number of released locks which is not true for flock locks, so
2407 just exclude such locks from consideration.
2411 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2412 Details : Use fsfilt_sync() for both the whole filesystem sync and
2413 individual file sync to eliminate dangerous inode locking
2414 with I_LOCK that can lead to a deadlock.
2418 Description: performance in 1.6.3
2419 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2420 sure that 1MB requests are not fragmented by the block layer.
2422 Severity : enhancement
2424 Description: organize the server-side client stats on per-nid basis
2425 Details : Change the structure of stats under obdfilter and mds to
2433 The "uuid"s file would list the uuids of _active_ exports.
2434 And the clear entry is to clear all stats and stale nids.
2436 Severity : enhancement
2438 Description: eliminate client locks in face of contention
2439 Details : file contention detection and lockless i/o implementation
2440 for contended files.
2444 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2447 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2449 Description: MDS may lose file striping (and hence file data) in some cases
2450 Details : If there are additional extended attributes stored on the MDS,
2451 in particular ACLs, SELinux, or user attributes (if user_xattr
2452 is specified for the client mount options) then there is a risk
2453 of attribute loss. Additionally, the Lustre file striping
2454 needs to be larger than default (e.g. striped over all OSTs),
2455 and an additional attribute must be stored initially in the
2456 inode and then increase in size enough to be moved to the
2457 external attribute block (e.g. ACL growing in size) for file
2462 Description: skiplist implementation simplification
2463 Details : skiplists are used to group compatible locks on granted list
2464 that was implemented as tracking first and last lock of each
2465 lock group the patch changes that to using doubly linked lists
2469 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2470 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2471 because LASSERT on both the data supplied by a client, and the
2472 data on disk is dangerous and incorrect.
2474 Severity : enhancement
2476 Description: Slow truncate/writes to huge files at high offsets.
2477 Details : Directly associate cached pages to lock that protect those pages,
2478 this allows us to quickly find what pages to write and remove
2479 once lock callback is received.
2483 Description: more ldlm soft lockups
2484 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2485 starve other threads from the resource lock for a long time in
2486 case of long waiting queue, so change the debug level from
2487 D_OTHER to the less frequently used D_INFO.
2489 Severity : enhancement
2491 Description: add -gid, -group, -uid, -user options to lfs find
2495 Description: Hung threads in invalidate_inode_pages2_range
2496 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2497 once one is completed. As a result, some RPCs are stuck in the
2498 queue and are never sent.
2502 Description: filter threads hungs on waiting journal commit
2503 Details : Cleanup filter group llog code, then only filter group llog will
2504 be only created in the MDS/OST syncing process.
2508 Description: Procfs and llog threads access destoryed import sometimes.
2509 Details : Sync the import destoryed process with procfs and llog threads by
2510 the import refcount and semaphore.
2512 Severity : enhancement
2514 Description: openlock cache of b1_6 port to HEAD
2519 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2520 Details : If insertion of an extent fails, then discard the inode
2521 preallocation and free data blocks else it can lead to duplicate
2526 Description: don't always update ctime in ext3_xattr_set_handle()
2527 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2528 In some cases the ctime should not be updated, for example for
2529 2.0->1.8 compatibility it is necessary to delete an xattr and it
2530 should not update the ctime.
2534 Bugzilla : 15713/16362
2535 Description: Assertion in iopen_connect_dentry in 1.6.3
2536 Details : looking up an inode via iopen with the wrong generation number can
2537 populate the dcache with a disconneced dentry while the inode
2538 number is in the process of being reallocated. This causes an
2539 assertion failure in iopen since the inode's dentry list contains
2540 both a connected and disconnected dentry.
2544 Description: assertion failure in ldlm_handle2lock()
2545 Details : fix a race between class_handle_unhash() and class_handle2object()
2546 introduced in lustre 1.6.5 by bug 13622.
2551 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2552 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2557 Description: oops in sd_iostats_seq_show()
2558 Details : unloading/reloading the scsi low level driver triggers a kernel
2559 bug when trying to access the sd iostat file.
2564 Description: Kernel panics during QLogic driver reload
2565 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2566 patch, causing memory corruption.
2571 Description: journal_dev option does not work in b1_6
2572 Details : pass mount option during pre-mount.
2574 Severity : enhancement
2576 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2577 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2578 extent information of a file. It can be used to map logical blocks
2579 in a file to physical blocks in the block device.
2583 Description: LDLM soft lockups - improvement
2584 Details : It is be possible to send the lock handle along with each read
2585 or write request because the client is already doing a lock match
2586 itself so there isn't any reason the OST should have to re-do that
2590 Frequency : only on Cray X2
2592 Description: X2 build failures
2593 Details : fix build failures on Cray X2.
2597 Description: xid & resent requests
2598 Details : Initialize RPC XID from clock at startup (randomly if clock is
2601 Severity : enhancement
2603 Description: Add lustre_start utility to start or stop multiple Lustre servers
2608 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2609 Details : In case of memory pressure, list_del() can be called twice on
2610 req->rq_history_list, causing a kernel oops.
2614 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2615 Details : fix stack overflow in the distributed lock manager by defering
2616 export eviction after a failed AST to the elt thread instead of
2617 handling it in the dlm interpret routine.
2621 Description: Convert some comments to new format.
2622 Details : Update documenting comments to match doxygen conventions.
2626 Description: Grammar fixes.
2627 Details : A couple of trivial spelling fixes.
2631 Description: OSD_COUNTERS-mandatory
2632 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2633 to simplify the code.
2637 Description: simplify cmm_device freeing logic.
2638 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2642 Description: Add lockdep support to dt_object_operations locking interface.
2643 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2644 parameter indicating lock ordering. Update mdd code to use new
2649 Description: Introduce failloc constants for lockless IO tests.
2650 Details : Add two new failloc constants to test lockless IO. Only one of
2651 them in implemented---another is checked in yet to be landed
2656 Description: Add lockdep support for inode mutex.
2657 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2658 in the situations where Lustre has to lock more than one inode
2663 Description: Add optional invariants checking support.
2664 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2665 configure switch. This macro is to be used for consistency and
2666 sanity checks that are too expensive to be left in `production'
2671 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2672 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2677 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2678 Details : Use spin_lock_nested() in (the only) situation where more than
2679 one ldlm_lock is locked simultaneously. Also, fix possible
2680 dead-lock in ldlm_lock_change_resource() by enforcing particular
2685 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2686 Details : Instead of specifying each ldlm_lock call-back through separate
2687 parameter, wrap them into struct ldlm_callback_suite.
2691 Description: Kill join_lru obd method and its callers.
2692 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2693 in memory---a requirement implemented through ->o_join_lru() obd
2694 method in HEAD. Remove this method and its users.
2698 Description: Add asynchronous ldlm ENQUEUE completion handler.
2699 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2700 case that stock ldlm_completion_ast() cannot handle as it waits
2701 until lock is granted. Introduce new ldlm_completion_ast_async()
2702 for this. Also comment ldlm_completion_ast().
2706 Description: ldlm_error <-> errno conversion.
2707 Details : Add functions to map (rather arbitrary) between LDLM error codes
2708 and standard errno values. CLIO needs this to prevent LDLM specific
2709 constants from escaping ldlm and osc.
2713 Description: Kill unused ldlm_handle2lock_ns() function.
2714 Details : Kill unused ldlm_handle2lock_ns() function.
2718 Description: Add lu_ref support to ldlm_lock
2719 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2720 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2721 and ldlm_resource. LDLM interface has to be changed, because code that
2722 releases a reference on a lock, has to "know" what reference this is.
2723 In the most frequent case
2725 lock = ldlm_handle2lock(handle);
2727 LDLM_LOCK_PUT(lock);
2729 no changes are required. When any other reference (received _not_ from
2730 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2731 instead of LDLM_LOCK_PUT().
2733 Arguably, changes are pervasive, and interface requires some discipline
2734 for proper use. On the other hand, it was very instrumental in finding
2735 a few leaked lock references.
2739 Description: Add ldlm_lock_addref_try().
2740 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2741 attempts to addref a lock that might be being canceled
2746 Description: Add ldlm_weigh_callback().
2747 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2748 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2749 hard-coded `number of pages' logic.
2753 Description: Add lockdep annotations to llog code.
2754 Details : Use appropriately tagged _nested() locking calls in the places
2755 where llog takes more than one ->lgh_lock lock.
2759 Description: Add loi_kms_set().
2760 Details : Wrap kms updates into a helper function.
2764 Description: Constify instances of struct lsm_operations.
2765 Details : Constify instances of struct lsm_operations.
2769 Description: lu_conf support.
2770 Details : On a server, a file system object is uniquely identified
2771 by a fid, which is sufficient to locate and load all object
2772 state (inode). On a client, on the other hand, more data are
2773 necessary instantiate an object. Change lu_object_find() and
2774 friends to take additional `lu_conf' argument describing object.
2775 Typically this includes layout information.
2779 Description: lu_context fixes.
2780 Details : Introduce new lu_context functions that are needed on the client
2781 side, where some system threads (ptlrpcd) are shared by multiple
2782 modules, and so cannot be stopped during module shutdown.
2786 Description: Add start and stop methods to lu_device_type_operations.
2787 Details : Introduce two new methods in lu_device_type_operations, that are
2788 invoked when first instance of a given type is created and last one
2789 is destroyed respectively. This is need by CLIO.
2793 Description: Add lu_ref support to struct lu_device.
2794 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2795 track leaked references.
2799 Description: Introduce lu_kmem_descr.
2800 Details : lu_kmem_descr and its companion interface allow to create
2801 and destroy a number of kmem caches at once.
2805 Description: Fix lu_object finalization race.
2806 Details : Fix a race between lu_object_find() finding an object and its
2807 concurrent finalization. This race is (most likely) not possible
2808 on the server, but might happen on the client.
2812 Description: Introduce lu_ref interface.
2813 Details : lu_ref is a debugging module allowing to track references to
2814 a given object. It is quite cpu expensive, and has to be
2815 explicitly enabled with --enable-lu_ref. See usage description
2820 Description: Factor lu_site procfs stats into a separate function.
2821 Details : Separate lu_site stats printing code into a separate function
2822 that can be reused on a client.
2826 Description: Constify instances of struct {lu,dt,md}_device_operations.
2827 Details : Constify instances of struct {lu,dt,md}_device_operations.
2831 Description: Introduce struct md_site and move meta-data specific parts of
2832 struct lu_site here.
2833 Details : Move md-specific fields out of struct lu_site into special struct
2834 md_site, so that lu_site can be used on a client.
2838 Description: Kill mdd_lov_destroy().
2839 Details : Remove unused mdd code.
2843 Description: Add st_block checking to multistat.c.
2844 Details : Add st_block checking to multistat.c.
2848 Description: Add lu_ref support to struct obd_device.
2849 Details : Add lu_ref tracking to obd_device.
2853 Description: Kill obd_set_fail_loc().
2854 Details : Remove unused code.
2858 Description: Add special type for ptlrpc_request interpret functions.
2859 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2860 there. Actual usage will be in CLIO.
2864 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2865 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2866 doesn't work with lockdep.
2870 Description: Add rwv.c test program.
2871 Details : New testing program exercising readv(2) and writev(2) (Qian).
2875 Description: Add sendfile.c test program.
2876 Details : New testing program exercising sendfile(2) (Jay).
2880 Description: Ratelimit a message that can be very frequent.
2881 Details : Ratelimit a memory allocation failure message that can
2886 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2887 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2891 Description: Liblustre build fixes.
2892 Details : Liblustre build fixes.
2896 Description: libcfs: add cfs_{need,cond}_resched() interface.
2897 Details : libcfs: add cfs_{need,cond}_resched() definition and
2898 implementations for Linux, NT, and liblustre.
2900 Severity : enhancement
2902 Description: More exported tunables for mballoc
2903 Details : Add support for tunable preallocation window and new tunables for
2904 large/small requests
2908 Description: Detect corruption of block bitmap and checking for preallocations
2909 Details : Checks validity of on-disk block bitmap. Also it does better
2910 checking of number of applied preallocations. When corruption is
2911 found, it turns filesystem readonly to prevent further corruptions.
2915 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2916 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2917 window to 0. If it does hit cache miss, and read-pattern is still
2918 stride-io mode, does not reset the stride window, but also does not
2919 increase the stride window length in this case.
2923 Frequency : only for big-endian servers
2924 Description: Check if system is big-endian while mounting fs with extents feature
2925 Details : Mounting a filesystem with extents feature will fail on big-endian
2926 systems since ext3-based ldiskfs is not supported on big-endian
2927 systems. This can be over-riden with "bigendian_extents" mount option.
2929 Severity : enhancement
2931 Description: The root squash functionality
2932 Details : A security feature, which is to prevent users from being able
2933 to mount lustre on their desktop, run as root, and delete
2934 all of the files in the filesystem. The goal is accomplished by
2935 remapping user id (UID) and group id (GID) of the root user to
2936 a UID and GID specified by the system administartor via Lustre
2937 configuration management server (MGS). The functionality also
2938 allows to specify sets of clients for which the remapping does
2943 Description: Excessive recovery window
2944 Details : With AT enabled, the recovery window can be excessively long (6000+
2945 seconds). To address this problem, we no longer use
2946 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2947 timeout no longer depends on the service time, it is set to
2948 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2949 time via pb_service_time.
2953 Description: Watchdog triggered on MDS failover
2954 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2955 the OSTs know that the MDS "UUID" can be reused for the same export
2956 from a different NID, so we do not need to wait for the export to be
2960 Frequency : rare, only if using MMP with Linux RAID
2962 Description: MMP doesn't work with Linux RAID
2963 Details : While using HA for Lustre servers with Linux RAID, it is possible
2964 that MMP will not detect multiple mounts. To make this work we
2965 need to unplug the device queue in RAID when the MMP block is being
2966 written. Also while reading the MMP block, we should read it from
2967 disk and not the cached one.
2969 Severity : enhancement
2971 Description: open file using fid
2972 Details : A file can be opened using just its fid, like
2973 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2976 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2979 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2980 Details : add proper locking for creation of jbd_4k slab cache
2984 Description: MMP check in ext3_remount() fails without displaying any error
2985 Details : When multiple mount protection fails during remount, proper error
2988 Severity : enhancement
2990 Description: Allow stripe size to be up to 4G-64k
2991 Details : Fix math logic to allow large stripe sizes.
2995 Description: add check for >8TB ldiskfs filesystems
2996 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2997 Don't allow >8TB ldiskfs filesystems to be mounted without
2998 force_over_8tb mount option
3003 Descriptoin: OSS complains "too many blocks" and crashes
3004 Details : In filter_commitrw_write():
3005 1. lquota_chkquota();
3006 2. obdo_from_inode();
3007 3. lquota_pending_commit();
3008 "2." has changed uid and gid in this case, then caused
3009 lquota_chkquota() and lquota_pending_commit() to use
3010 different uid and gid. That is the root of the bug.
3014 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3015 Details : Detect and print out the 'stripe count = -1' case properly.
3019 Description: Out or order replies might be lost on replay
3020 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3021 request with tid smaller than one currently being added, add it
3022 to the start, not end of the list.
3024 --------------------------------------------------------------------------------
3026 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3028 * Support for kernels:
3029 2.6.5-7.283 (SLES 9),
3030 2.6.9-55.EL (RHEL 4),
3031 2.6.16.46-0.14 (SLES 10),
3032 2.6.18.8 vanilla (kernel.org)
3033 * Client support for unpatched kernels:
3034 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3035 2.6.16 - 2.6.22 vanilla (kernel.org)
3036 * Due to recently discovered recovery problems, we do not recommend
3037 using patchless RHEL 4 clients with this or any earlier release.
3038 * Recommended e2fsprogs version: 1.39.cfs8
3039 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3040 * Starting with this release, the ldiskfs backing filesystem required
3041 by Lustre is now in its own package, lustre-ldiskfs. This package
3042 should be installed. It is versioned separately from Lustre and
3043 may be released separately in future.
3047 Description: block reactivating mgc import until all deactivates complete
3048 Details : Fix race when failing back MDT/MGS to itself (testing)
3051 Frequency : at statup only
3053 Description: mds_lov_synchronize race leads to various problems
3054 Details : simultaneous MDT->OST connections at startup can cause the
3055 sync to abort, leaving the OSC in a bad state.
3057 Severity : enhancement
3059 Description: add optional extra BUILD_VERSION info
3060 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3061 to override the lustre version.
3064 Frequency : 2.6.18 servers only
3066 Description: ll_kern_mount() doesn't release the module reference
3067 Details : The ldiskfs module reference count never drops down to 0
3068 because ll_kern_mount() doesn't release the module reference.
3073 Description: server LBUG when using old ost_num_threads parameter
3074 Details : Accept the old ost_num_threads parameter but warn that it
3075 is deprecated, and fix an off-by-one error that caused an LBUG.
3080 Description: Transient SCSI error results in persistent IO issue
3081 Details : iobuf->dr_error is not reinitialized to 0 between two
3085 Frequency : sometimes when underlying device returns I/O errors
3087 Description: OSTs not going read-only during write failures
3088 Details : OSTs are not remounted read-only when the journal commit threads
3089 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3090 instead of the ext3 wrappers.
3094 Description: poor connect scaling with increasing client count
3095 Details : Don't run filter_grant_sanity_check for more than 100 exports
3096 to improve scaling for large numbers of clients.
3099 Frequency : SLES10 only
3101 Description: sanity-quota.sh quotacheck failed: rc = -22
3102 Details : Quotas cannot be enabled on SLES10.
3105 Frequency : liblustre clients only
3107 Description: getdirentries does not give error when run on compute nodes
3108 Details : getdirentries does not fail when the size specified as an argument
3109 is too small to contain at least one entry
3111 Severity : enhancement
3113 Description: Add LNET router traceability for debug purposes
3114 Details : If a checksum failure occurs with a router as part of the
3115 IO path, the NID of the last router that forwarded the bulk data
3116 is printed so it can be identified.
3121 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3122 Details : Due to a race condition, liblustre clients were occasionally
3123 evicted incorrectly.
3125 Severity : enhancement
3127 Description: lfs setstripe use optional parameters instead of postional
3130 Severity : enhancement
3132 Description: Nanosecond timestamp support for ldiskfs
3133 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3134 resolution timestamps. There is not yet support for this at
3135 the Lustre filesystem level.
3138 Frequency : during server recovery
3140 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3141 Details : request with rq_no_resend flag not awake l_wait_event if they get
3145 Frequency : nfs export on patchless client
3147 Description: connectathon hang when test nfs export over patchless client
3148 Details : Disconnected dentry cannot be found with lookup, so we do not need
3149 to unhash it or make it invalid
3152 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3153 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3154 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3159 Description: open req refcounting wrong on reconnect
3160 Details : If reconnect happened between getting open reply from server and
3161 call to mdc_set_replay_data in ll_file_open, we will schedule
3162 replay for unreferenced request that we are about to free.
3163 Subsequent close will crash in variety of ways.
3164 Check that request is still eligible for replay in
3165 mdc_set_replay_data().
3170 Description: disable writes to filesystem when reading health_check file
3171 Details : the default for reading the health_check proc file has changed
3172 to NOT do a journal transaction and write to disk, because this
3173 can cause reads of the /proc file to hang and block HA state
3174 checking on a healthy but otherwise heavily loaded system. It
3175 is possible to return to the previous behaviour during configure
3176 with --enable-health-write.
3178 Severity : enhancement
3180 Description: 64-bit inode version
3181 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3182 to the inode. This will be required for version-based recovery.
3187 Description: MDS fails to start if a duplicate client export is detected
3188 Details : in some rare cases it was possible for a client to connect to
3189 an MDS multiple times. Upon recovery the MDS would detect this
3190 and fail during startup. Handle this more gracefully.
3192 Severity : enhancement
3194 Description: Add -o localflock option to simulate old noflock
3196 Details : This will achieve local-only flock/fcntl locks
3202 Description: log_commit_thread vs filter_destroy race leads to crash
3203 Details : Take import reference before releasing llog record semaphore
3208 Description: Wrong request locking in request set processing
3209 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3210 imp_delayed_list, in this place should be used imp_lock.
3213 Frequency : when reconnection
3215 Description: Grant Leak when osc reconnect to OST
3216 Details : When osc reconnect ost, OST(filter) should check whether it
3217 should grant more space to client by comparing fed_grant and
3218 cl_avail_grant, and return the granted space to client instead
3219 of "new granted" space, because client will call osc_init_grant
3220 to update the client grant space info.
3223 Frequency : when client reconnect to OST
3225 Description: Grant Leak when osc do resend and replay bulk write
3226 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3227 bulk write request, because the grant info while be sync between
3228 OSC and OST when reconnect, and we should ignore the grant info
3229 these of resend/replay write req.
3234 Description: Grant space more than available left space sometimes.
3235 Details : When then OST is about to be full, if two bulk writing from
3236 different clients came to OST. Accord the avaliable space of the
3237 OST, the first req should be permitted, and the second one
3238 should be denied by ENOSPC. But if the seconde arrived before
3239 the first one is committed. The OST might wrongly permit second
3240 writing, which will cause grant space > available space.
3243 Frequency : when client is evicted
3245 Description: Grant might be wrongly erased when osc is evicted by OST
3246 Details : when the import is evicted by server, it will fork another
3247 thread ptlrpc_invalidate_import_thread to invalidate the
3248 import, where the grant will be set to 0. While the original
3249 thread will update the grant it got when connecting. So if
3250 the former happened latter, the grant will be wrongly errased
3251 because of this race.
3256 Description: Checking Stale with correct fid
3257 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3258 to check whether it is stale, because sometimes, we want the
3259 enqueue happened anyway, and op_data.fid2 will not be initialized.
3261 Severity : enhancement
3263 Description: update patchless client
3264 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3267 Frequency : only with 2.4 kernel
3269 Description: random memory corruption
3270 Details : size of struct ll_inode_info is to big for union inode.u and this
3271 can be cause of random memory corruption.
3276 Description: Memory leak in recovery
3277 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3278 It should also check obd_fail before fsfilt_start, otherwise if
3279 fsfilt_start return -EROFS,(failover mds during mds recovery).
3280 then the req will return with repmsg->transno = 0 and rc = EROFS.
3281 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3282 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3283 be freed no matter whether fsfilt_commit success or not.
3286 Frequency : only with huge count clients
3288 Description: Prevents from taking the superblock lock in llap_from_page for
3290 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3291 for llap_from_page call in ll_removepage prevents from taking the
3292 superblock lock for a soon died page.
3297 Description: Not check open intent error before release open handle
3298 Details : in some rare cases, the open intent error is not checked before
3299 release open handle, which may cause
3300 ASSERTION(open_req->rq_transno != 0), because it tries to release
3301 the failed open handle.
3306 Description: Set cat log bitmap only after create log success.
3307 Details : in some rare cases, the cat log bitmap is set too early. and it
3308 should be set only after create log success.
3312 Description: Accessing a block bevice can re-enable I/O when Lustre is
3313 tearing down a device.
3314 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3319 Description: service threads may hog cpus when there are a lot of requests
3320 Details : Insert cond_resched to give other threads a chance to use some CPU
3325 Description: the cat log was not initialized in recovery
3326 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3327 unlink log on mds will not be initialized until mds post
3328 recovery. And also in mds post recovery, the unlink log will
3329 initialization will be done asynchronausly, so there will be race
3330 between add unlink log and unlink log initialization.
3334 Description: brw_stats were being printed incorrectly
3335 Details : brw_stats were being printed as log2 but all of them were not
3336 recorded as log2. Also remove some code duplication arising from
3337 filter_tally_{read,write}.
3341 Frequency : rare, only in recovery.
3342 Description: ASSERTION(req->rq_type != LI_POISON) failed
3343 Details : imp_lock should be held while iterating over imp_sending_list for
3344 prevent destroy request after get timeout in ptlrpc_queue_wait.
3348 Description: replay-single.sh test 52 fails
3349 Details : A lock's skiplist need to be cleanup when it being unlinked
3350 from its resource list.
3354 Description: Short directio read returns full requested size rather than
3356 Details : Direct I/O operations should return actual amount of bytes
3357 transferred rather than requested size.
3359 Severity : enhancement
3361 Description: metadata RPC reduction (e.g. for rm performance)
3362 Details : decrease the amount of synchronous RPC between clients and servers
3363 by canceling conflicing lock before the operation on the client
3364 and packing thier handles into the main operation RPC to server.
3366 Severity : enhancement
3368 Description: Async OSC create to avoid the blocking unnecessarily.
3369 Details : If an OST has no remaining object, system will block on the
3370 creation when it needs to create a new object on this OST. Now,
3371 ways use pre-created objects when available, instead of blocking on
3372 an empty osc while others are not empty. If we must block, we
3373 block for the shortest possible period of time.
3377 Description: improve handling recoverable errors
3378 Details : if request processig with error which can be recoverable on server
3379 request should be resend, otherwise page released from cache and
3382 Severity : enhancement
3384 Description: refine locking for avoid write wrong info into lov_objid file
3385 Details : fix possible races with add new target and write/update data in
3388 --------------------------------------------------------------------------------
3390 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3395 Frequency : on some architectures
3397 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3398 Details : Uninitialized flags sometimes cause configuration commands to
3402 Frequency : patchless clients only
3404 Description: missing __iget() symbol export
3405 Details : The __iget() symbol export is missing. To avoid the need for
3406 this on patchless clients the deathrow inode reaper is turned
3407 off, and we depend on the VM to clean up old inodes. This
3408 dependency was during via the fix for bug 12181.
3410 --------------------------------------------------------------------------------
3412 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3414 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3415 INTEROPERATE with older versions automatically. In many cases a
3416 special upgrade step is needed. Please read the
3417 user documentation before upgrading any part of a live system.
3418 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3419 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3420 * WARNING: Lustre configuration and startup changes are required with
3421 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3425 Severity : enhancement
3427 Description: MountConf
3428 Details : Lustre configuration is now managed via mkfs and mount
3429 commands instead of lmc and lconf. New obd types (MGS, MGC)
3430 are added for dynamic configuration management. See
3431 https://mail.clusterfs.com/wikis/lustre/MountConf for
3434 Severity : enhancement
3436 Description: dynamic OST addition
3437 Details : OSTs can now be added to a live filesystem
3439 Severity : enhancement
3441 Description: startup order invariance
3442 Details : MDTs and OSTs can be started in any order. Clients only
3443 require the MDT to complete startup.
3445 Severity : enhancement
3447 Description: parallel, asynchronous orphan cleanup
3448 Details : orphan cleanup is now performed in separate threads for each
3449 OST, allowing parallel non-blocking operation.
3451 Severity : enhancement
3453 Description: optimized stripe assignment
3454 Details : stripe assignments are now made based on ost space available,
3455 ost previous usage, and OSS previous usage, in order to try
3456 to optimize storage space and networking resources.
3458 Severity : enhancement
3460 Description: Permanently set tunables
3461 Details : All writable /proc/fs/lustre tunables can now be permanently
3462 set on a per-server basis, at mkfs time or on a live system.
3464 Severity : enhancement
3466 Description: Lustre message v2
3467 Details : Add lustre message format v2.
3469 Severity : enhancement
3471 Description: client OST exclusion list
3472 Details : Clients can be started with a list of OSTs that should be
3473 declared "inactive" for known non-responsive OSTs.
3477 Description: SPEC SFS validation failure on NFS v2 over lustre.
3478 Details : Changes the blocksize for regular files to be 2x RPC size,
3479 and not depend on stripe size.
3481 Severity : enhancement
3483 Description: Multiple MD RPCs in flight.
3484 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3485 To avoid overly-overloading MDS, introduce a limit on number of
3486 MDS RPCs in flight for a single client and add /proc controls
3487 to adjust this limit.
3489 Severity : enhancement
3491 Description: client read/write statistics
3492 Details : Add client read/write call usage stats for performance
3493 analysis of user processes.
3494 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3495 file access. extents_stats shows chunk size distribution.
3496 extents_stats_per_process show chunk size distribution per
3499 Severity : enhancement
3501 Description: mds statistics
3502 Details : Add detailed mds operations statistics in
3503 /proc/fs/lustre/mds/*/stats.
3507 Description: Failure of copying files with lustre special EAs.
3508 Details : Client side always return success for setxattr call for lustre
3509 special xattr (currently only "trusted.lov").
3514 Description: Refcount LNET uuids
3515 Details : The global LNET uuid list grew linearly with every startup;
3516 refcount repeated list entries instead of always adding to
3519 Severity : enhancement
3521 Description: Dynamic service threads
3522 Details : Within a small range, start extra service threads
3523 automatically when the request queue builds up.
3526 Frequency : mixed-endian client/server environments
3528 Description: mixed-endian crashes
3529 Details : The new msg_v2 system had some failures in mixed-endian
3532 Severity : enhancement
3534 Description: Easy OST removal
3535 Details : OSTs can be permanently deactivated with e.g. 'lctl
3536 conf_param lustre-OST0001.osc.active=0'
3538 Severity : enhancement
3540 Description: MGS proc entries
3541 Details : Added basic proc entries for the MGS showing what filesystems
3544 Severity : enhancement
3546 Description: provide MGS failover
3547 Details : Added config lock reacquisition after MGS server failover.
3549 Severity : enhancement
3551 Description: add Linux 2.4 support
3552 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3556 Description: a large application tries to do I/O to the same resource and dies
3557 in the middle of it.
3558 Details : Check the req->rq_arrival time after the call to
3559 ost_brw_lock_get(), but before we do anything about
3560 processing it & sending the BULK transfer request. This
3561 should help move old stale pending locks off the queue as
3562 quickly as obd_timeout.
3565 Frequency : when an incorrect nid is specified during startup
3567 Description: ptlrpc connect to non-existant node causes kernel crash
3568 Details : LNET can't be re-entered from an event callback, which
3569 happened when we expire a message after the export has been
3570 cleaned up. Instead, hand the zombie cleanup off to another
3573 Severity : enhancement
3575 Description: plain/inodebits lock performance improvement
3576 Details : Grouping plain/inodebits in granted list by their request modes
3577 and bits policy, thus improving the performance of search through
3581 Frequency : only if OST filesystem is corrupted
3583 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3584 Details : for a short time RPCs with bulk IO are in the replay list,
3585 but replay of bulk IOs is unimplemented. If the OST filesystem
3586 is corrupted due to disk cache incoherency and then replay is
3587 started it is possible to trip an assertion. Avoid putting
3588 committed RPCs into the replay list at all to avoid this issue.
3591 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3593 Description: System hang on startup
3594 Details : This bug allowed the liblustre (e.g. catamount) client to
3595 return to the app before handling all startup RPCs. This
3596 could leave the node unresponsive to lustre network traffic
3597 and manifested as a server ptllnd timeout.
3599 Severity : enhancement
3601 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3602 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3603 boolean to control whether to print peer debug info when a
3604 client's RPC times out.
3607 Frequency : only for kernels with patches from Lustre below 1.4.3
3609 Description: Remove old rdonly API
3610 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3613 Frequency : only for devices with external journals
3615 Description: Set external device read-only also
3616 Details : During a commanded failover stop, we set the disk device
3617 read-only while the server shuts down. We now also set any
3618 external journal device read-only at the same time.
3621 Frequency : when upgrading from 1.4 while trying to change parameters
3623 Description: The wrong (new) MDC name was used when setting parameters for
3624 upgraded MDT's. Also allows changing of OSC (and MDC)
3625 parameters if --writeconf is specified at tunefs upgrade time.
3628 Frequency : when setting specific ost indicies
3630 Description: QOS code breaks on skipped indicies
3631 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3632 created with --index need not be sequential.
3634 Severity : enhancement
3636 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3637 Details : The uninit_groups feature works in conjunction with the kernel
3638 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3639 up the pass1 processing of e2fsck. This is a read-only feature
3640 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3641 mount filesystems that have had this feature enabled.
3643 Severity : enhancement
3645 Description: Improve multi-block allocation algorithm to avoid fragmentation
3646 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3647 allocation locality and avoid filesystem fragmentation.
3649 ------------------------------------------------------------------------------
3651 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3653 * Support for kernels:
3654 2.4.21-47.0.1.EL (RHEL 3)
3655 2.6.5-7.283 (SLES 9)
3656 2.6.9-42.0.10.EL (RHEL 4)
3657 2.6.12.6 vanilla (kernel.org)
3658 2.6.16.27-0.9 (SLES 10)
3659 * Recommended e2fsprogs version: 1.39.cfs5
3661 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3665 Frequency : occasional, depends on client load and configuration
3666 Bugzilla : 12181, 12203
3667 Description: data loss for recently-modified files
3669 Details : In some cases it is possible that recently written or created
3670 files may not be written to disk in a timely manner (this should
3671 normally be within 30s unless client IO load is very high).
3672 The problem appears as zero-length files or files that are a
3673 multiple of 1MB in size after a client crash or client eviction
3674 that are missing data at the end of the file.
3676 This problem is more likely to be hit on clients where files are
3677 repeatedly created and unlinked in the same directory, clients
3678 have a large amount of RAM, have many CPUs, the filesystem has
3679 many OSTs, the clients are rebooted frequently, and/or the files
3680 are not accessed by other nodes after being written.
3682 The presence of the problem can be detected by looking at
3683 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3684 smaller than the second (nr_unused) then dirty files will not
3685 be flushed automatically to disk. "sync; sleep 10" should be
3686 run several times on the node before unmounting it to update
3687 Lustre (this is also safe to run on nodes without this problem).
3689 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3690 that can cause this same problem, so customers using that kernel
3691 also need to update the kernel in addition to Lustre. In order
3692 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3694 It is normal that files written just before a client crash (less
3695 than 30s) may not yet have been flushed to disk, even for local
3699 Frequency : frequent on thin XT3 nodes
3701 Description: UUID collision on thin XT3 Linux nodes
3702 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3703 randomly, since we relied on an insufficiently-seeded PRNG.
3708 Description: OSS hangs after "All ost request buffers busy"
3709 Details : A deadlock between quota and journal operations caused OSS
3710 hangs after printing "All ost request buffers busy."
3713 Frequency : always on liblustre builds
3715 Description: Cleanup compiler warnings on liblustre
3718 Frequency : always on liblustre builds on XT3
3720 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3724 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3725 > 32000 subdirectories
3726 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3727 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3728 subdirectory count crosses 32000. This will aid e2fsck to
3729 correctly handle more than 32000 subdirectories.
3732 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3734 Description: System hang on startup
3735 Details : This bug allowed the liblustre (e.g. catamount) client to
3736 return to the app before handling all startup RPCs. This
3737 could leave the node unresponsive to lustre network traffic
3738 and manifested as a server ptllnd timeout.
3740 Severity : enhancement
3742 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3743 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3744 boolean to control whether to print peer debug info when a
3745 client's RPC times out.
3750 Description: make O_SYNC working on 2.6 kernels
3751 Details : 2.6 kernels use different method for mark pages for write,
3752 so need add a code to lustre for O_SYNC work.
3757 Description: Failure to close file and release space on NFS
3758 Details : Put inode details into lock acquired in ll_intent_file_open.
3759 Use mdc_intent_lock in ll_intent_open to properly
3760 detect all kind of errors unhandled by mdc_enqueue.
3765 Description: proc file read during shutdown sometimes raced obd removal,
3767 Details : Add lock to prevent obd access after proc file removal.
3770 Frequency : Only for files larger than 4GB on 32-bit clients.
3772 Description: improperly doing page alignment of locks
3773 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3774 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3779 Description: RPCs being resent when they shouldn't be
3780 Details : Some RPCs that should not be resent are being resent. This
3781 can cause inconsistencies in the RPC state machine. Do not
3782 resend such requests.
3785 Frequency : rare, only with NFS export
3787 Description: Crash on NFS re-export node
3788 Details : under very unusual load conditions an assertion is hit in
3789 ll_intent_file_open()
3792 Frequency : only if OST filesystem is corrupted
3794 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3795 Details : for a short time RPCs with bulk IO are in the replay list,
3796 but replay of bulk IOs is unimplemented. If the OST filesystem
3797 is corrupted due to disk cache incoherency and then replay is
3798 started it is possible to trip an assertion. Avoid putting
3799 committed RPCs into the replay list at all to avoid this issue.
3804 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3805 Details : Large single O_DIRECT read and write calls can fail to allocate
3806 a sufficiently large buffer to process the request. In case of
3807 allocation failure the allocation is retried with a smaller
3808 buffer and broken into smaller requests.
3810 Severity : enhancement
3812 Description: Add -o localflock option to simulate old noflock behaviour.
3813 Details : This will achieve local-only flock/fcntl locks coherentness.
3818 Description: versioning check is incomplete
3819 Details : Checking the version difference of client vs. server, report
3820 error if the gap is too big.
3825 Description: add support PG_writeback bit
3826 Details : add support for PG_writeback bit for Lustre, for more carefull
3827 work with page cache in 2.6 kernel. This also fix some deadlocks
3828 and remove hack for work O_SYNC with 2.6 kernel.
3830 Severity : enhancement
3832 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3833 Details : The uninit_groups feature works in conjunction with the kernel
3834 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3835 up the pass1 processing of e2fsck. This is a read-only feature
3836 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3837 mount filesystems that have had this feature enabled.
3839 Severity : enhancement
3841 Description: Improve multi-block allocation algorithm to avoid fragmentation
3842 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3843 allocation locality and avoid filesystem fragmentation.
3845 ------------------------------------------------------------------------------
3847 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3849 * Support for kernels:
3850 2.6.9-42.0.3.EL (RHEL 4)
3851 2.6.5-7.276 (SLES 9)
3852 2.4.21-47.0.1.EL (RHEL 3)
3853 2.6.12.6 vanilla (kernel.org)
3854 2.6.16.21-0.8 (SLES10)
3855 * Recommended e2fsprogs version: 1.39.cfs2-0
3857 * The backwards-compatible /proc/sys/portals symlink has been removed
3858 in this release. Before upgrading, please ensure that you change
3859 any configuration scripts or /etc/sysctl.conf files that access
3860 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3861 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3862 in advance of the upgrade on any system running Lustre 1.4.6 or
3863 newer, since /proc/sys/lnet was added in that version.
3864 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3868 Frequency : only when quota is used
3870 Description: avoid scanning export list for quota master
3871 Details : Change the algorithms to avoid scanning export list in order
3872 to improve the efficiency.
3875 Frequency : MDS failover only, very rarely
3877 Description: "went back in time" messages on mds failover
3878 Details : The greatest transno may be lost when the current operation
3879 finishes with an error (transno==0) and the client's last_rcvd
3880 record is over-written. Save the greatest transno in the
3881 mds_last_transno for this case.
3884 Frequency : always for specific kernels and striping counts
3886 Description: client may get "Matching packet too big" without ACL support
3887 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3888 when trying to access files in certain configurations. The
3889 clients should in fact be denied when mounting because they do
3890 not understand ACLs.
3893 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3895 Description: many clients connecting with IO in progress causes connect timeouts
3896 Details : Avoid synchronous journal commits to avoid delays caused by many
3897 clients connecting/disconnecting when bulk IO is in progress.
3898 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3899 OST_IO_PORTAL to avoid delays behind potentially many pending
3903 Frequency : occasionally with multiple writers to a single file
3905 Description: shared writes to file may result in wrong size reported by stat()
3906 Details : Allow growing of kms when extent lock is cancelled
3909 Frequency : always with random mmap IO to multi-striped file
3911 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3912 Details : Only if the hole is at the end of OST object so that kms is too
3913 small. Fix is to increase kms accordingly in ll_nopage.
3916 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3918 Description: writes to a missing object would leak memory on the OST
3919 Details : If there is an inconsistency between the MDS and OST filesystems,
3920 such that the MDS references an object that doesn't exist, writes
3921 to that object will leak memory due to incorrect cleanup in the
3922 error handling path, eventually running out of memory on the OST.
3927 Description: Creating too long symlink causes lustre errors
3928 Details : Check symlink and name lengths before sending requests to MDS.
3931 Frequency : only if flock is enabled (not on by default)
3933 Description: posix locks not released on fd closure on 2.6.9+
3934 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3935 kernels, this caused such locks not to be released on fd close and
3936 then assertions on fs unmount about still used locks.
3939 Frequency : MDS failover only, very rarely
3941 Description: clients may get ASSERTION(granted_lock != NULL)
3942 Details : When request was taking a long time, and a client was resending
3943 a getattr by name lock request. The were multiple lock requests
3944 with the same client lock handle and
3945 mds_getattr_name->fixup_handle_for_resent_request found one of the
3946 lock handles but later failed with ASSERTION(granted_lock != NULL).
3951 Description: handle->h_buffer_credits > 0, assertion failure
3952 Details : h_buffer_credits is zero after truncate, causing assertion
3953 failure. This patch extends the transaction or creates a new
3957 Frequency : NFS re-export or patchless client
3958 Bugzilla : 11179, 10796
3959 Description: Crash on NFS re-export node (__d_move)
3960 Details : We do not want to hash the dentry if we don't have a lock.
3961 But if this dentry is later used in d_move, we'd hit uninitialised
3962 list head d_hash, so we just do this to init d_hash field but
3963 leave dentry unhashed.
3966 Frequency : NFS re-export or patchless client
3968 Description: NFS exports has problem with symbolic link
3969 Details : lustre client didn't properly install dentry when re-exported
3970 to NFS or running patchless client.
3973 Frequency : NFS re-export or patchless client
3975 Description: Various nfs/patchless fixes.
3976 Details : fixes reuse disconected alias for lookup process - this fixes
3977 warning "find_exported_dentry: npd != pd",
3978 fix permission error with open files at nfs.
3979 fix apply umask when do revalidate.
3982 Frequency : occasional
3984 Description: Crash on NFS re-export node
3985 Details : calling clear_page() on the wrong pointer triggered oops in
3986 generic_mapping_read().
3989 Frequency : rarely, using O_DIRECT IO
3991 Description: unaligned directio crashes client with LASSERT
3992 Details : check for unaligned buffers before trying any requests.
3995 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3997 Description: stale data returned from RAID cache
3998 Details : If only a small amount of IO is done to the RAID device before
3999 reading it again it is possible to get stale data from the RAID
4000 cache instead of reading it from disk.
4003 Frequency : always for sles10 kernel
4005 Description: sles10 support
4006 Details : ll_follow_link: compile fixes and using of nd_set_link
4007 under newer kernels.
4010 Frequency : depends on arch, kernel and compiler version, always on sles10
4013 Description: recursive or deep enough symlinks cause stack overflow
4014 Details : getting rid of large stack-allocated variable in
4018 Frequency : depends on hardware
4020 Description: lustre write performance loss in the SLES10 kernel
4021 Details : the performance loss is caused by using of write barriers in the
4022 ext3 code. The SLES10 kernel turns barrier support on by
4023 default. The fix is to undo that change for ldiskfs.
4025 ------------------------------------------------------------------------------
4027 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4029 * Support for kernels:
4030 2.6.9-42.0.3EL (RHEL 4)
4031 2.6.5-7.276 (SLES 9)
4032 2.4.21-47.0.1.EL (RHEL 3)
4033 2.6.12.6 vanilla (kernel.org)
4037 Frequency : quota enabled and large files being deleted
4039 Description: releasing more than 4GB of quota at once hangs OST
4040 Details : If a user deletes more than 4GB of files on a single OST it
4041 will cause the OST to spin in an infinite loop. Release
4042 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4047 Description: statfs data retrieved from /proc may be stale or zero
4048 Details : When reading per-device statfs data from /proc, in the
4049 {kbytes,files}_{total,free,avail} files, it may appear
4050 as zero or be out of date.
4053 Frequency : systems with MD RAID1 external journal devices
4055 Description: lconf's call to blkid is confused by RAID1 journal devices
4056 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4057 of returning all block devices that match the journal UUID.
4060 Frequency : always, for aggregate stripe size over 4GB
4062 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4063 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4064 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4067 Frequency : always if "lfs find" used on a local file/directory
4069 Description: "lfs find" segfaults if used on a local file/directory
4070 Details : The case where a directory component was not specified wasn't
4071 handled correctly. Handle this properly.
4074 Frequency : always on ppc64
4076 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4077 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4078 perform endianness conversion.
4081 Frequency : rarely (truncate to non-zero file size after write under load)
4082 Bugzilla : 10730, 10687
4083 Description: Files padded with zeros to next 4K multiple
4084 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4085 it is possible that files that are truncated to a non-zero size
4086 immediately after being written are filled with zero bytes beyond
4087 the truncated size. No file data is lost.
4089 Severity : enhancement
4091 Description: Allow recovery/failover for liblustre clients.
4092 Details : liblustre clients were unaware of failover configurations until
4095 Severity : enhancement
4097 Description: user file locks should fail when not mounting with flock option
4098 Details : Set up an error-returning stub in ll_file_operations.lock field
4099 to prevent incorrect behaviour when client is mounted without
4100 flock option. Also, set up properly f_op->flock field for
4104 Frequency : always on ia64
4106 Description: "lfs df" loops on printing out MDS statfs information
4107 Details : The obd_ioctl_data was not initialized and in some systems
4108 this caused a failure during the ioctl that did not return
4109 an error. Initialize the struct and return an error on failure.
4112 Frequency : SLES 9 only
4114 Description: Error of copying files with lustre special EAs as root
4115 Details : Client side always return success for setxattr call for lustre
4116 special xattr (currently only "trusted.lov").
4119 Frequency : rarely on clusters with both ia64+i386 clients
4121 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4122 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4123 clients are doing concurrent writes to the same file it is
4124 possible that the ia64 clients may LASSERT because the OST
4125 extent locks are not PAGE_SIZE aligned. Ensure that grown
4126 locks are always aligned on the request boundary.
4129 Frequency : specific use, occasional
4131 Description: Overwriting in use executable truncates on-disk binary image
4132 Details : If one node attempts to overwrite an executable in use by
4133 another node, we now correctly return ETXTBSY instead of
4134 truncating the file.
4136 Severity : enhancement
4138 Description: Async OSC create to avoid the blocking unnecessarily.
4139 Details : If an OST has no remaining object, system will block on the creating
4140 when need to create a new object on this OST. Now, ways use
4141 pre-created objects when available, instead of blocking on an
4142 empty osc while others are not empty. If we must block, we block
4143 for the shortest possible period of time.
4148 Description: chmod on Lustre root is propagated to other clients
4149 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4150 invalid by the follow_mount time.
4153 Frequency : liblustre clients only
4155 Description: Race in 'instant cancel' lock handling could lead to such locks
4156 never to be granted in case of SMP MDS
4157 Details : Do not destroy not yet granted but cbpending locks in
4161 Frequency : replay/resend of open
4163 Description: non null lock assetion failure in mds_intent_policy
4164 Details : Trying to replay/resend lockless open requests resulted in
4165 mds_open() returning 0 with no lock. Now it sets a flag if
4166 a lock is going to be returned.
4168 Severity : enhancement
4170 Description: Checksum enhancements
4171 Details : New checksum enhancements allow for resending RPCs that failed
4174 Severity : enhancement
4176 Description: Tunables on number of dirty pages in cacche
4177 Details : Allow to set limit on number of dirty pages cached.
4182 Description: client crash on unmount - lock still has references
4183 Details : In some error handling cases it was possible to leak a lock
4184 reference on a client while accessing a file. This was not
4185 harmful to the client during operation, but would cause the
4186 client to crash when the filesystem is unmounted.
4189 Frequency : specific case, rare
4191 Description: ETXTBSY on mds though file not in use by client
4192 Details : ETXTBSY is no longer incorrectly returned when attempting to
4193 chmod or chown a directory that the user previously tried to
4194 execute or a currently-executing binary.
4197 Frequency : extremely rare except on liblustre-based clients
4199 Description: Lustre space not freed when files are deleted
4200 Details : Clean up open-unlinked files after client eviction. Previously
4201 the unlink was skipped and the files remained as orphans.
4206 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4207 Details : In some cases it was possible to send a blocking callback to
4208 a client doing a glimpse, even though that client didn't get
4209 a lock granted. When the glimpse lock is cancelled on the OST
4210 the freed lock is left on the waiting list and corrupted the list.
4213 Frequency : all core dumps
4215 Description: Broke core dumps to lustre
4216 Details : Negative dentry may be unhashed if parent does not have UPDATE
4217 lock, but some callers, e.g. do_coredump, expect dentry to be
4218 hashed after successful create, hash it in ll_create_it.
4220 ------------------------------------------------------------------------------
4222 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4224 * Support for kernels:
4225 2.6.9-42.0.2.EL (RHEL 4)
4226 2.6.5-7.276 (SLES 9)
4227 2.4.21-40.EL (RHEL 3)
4228 2.6.12.6 vanilla (kernel.org)
4232 Frequency : always on RHEL 3
4234 Description: Number of open files grows over time
4235 Details : The number of open files grows over time, whether or not
4236 Lustre is started. This was due to a filp leak introduced
4237 by one of our kernel patches.
4239 ------------------------------------------------------------------------------
4241 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4243 * Support for kernels:
4244 2.6.9-42.EL (RHEL 4)
4245 2.6.5-7.276 (SLES 9)
4246 2.4.21-40.EL (RHEL 3)
4247 2.6.12.6 vanilla (kernel.org)
4252 Bugzilla : 5719, 9635, 9792, 9684
4253 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4254 Details : If a server is under heavy load and cannot reply to new
4255 connection requests before the client resends the (re)connect,
4256 the connection handling code can behave badly if two service
4257 threads are concurrently handing separate (re)connections from
4258 the same client. Add better locking to the connection handling
4259 code, and ensure that only a single connection will be processed
4260 for a given client UUID, even if the lock is dropped.
4262 Severity : enhancement
4264 Description: add TCP zero-copy support to kernel
4265 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4266 sends if the hardware supports scatter-gather and checksumming.
4267 This allows socklnd to do client-write and server-read more
4268 efficiently and reduce CPU utilization from skbuf copying.
4271 Frequency : only if NFS exporting from client
4273 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4274 Details : If setting attributes on a file created under NFS that had
4275 never been opened it would be possible to oops the client
4276 if the file had no objects.
4279 Frequency : always for liblustre
4281 Description: liblustre client does MDS+OSTs setattr RPC for each write
4282 Details : When doing a write from a liblustre client, the client
4283 incorrectly issued an RPC to the MDS and each OST the file was
4284 striped over in order to update the timestamps. When writing
4285 with small chunks and many clients this could overwhelm the MDS
4286 with RPCs. In all cases it would slow down the write because
4287 these RPCs are unnecessary.
4289 Severity : enhancement
4291 Description: allow number of MDS service threads to be changed at module load
4292 Details : It is now possible to change the number of MDS service threads
4293 running. Adding "options mds mds_num_threads={N}" to the MDS's
4294 /etc/modprobe.conf will set the number of threads for the next
4295 time Lustre is restarted (assuming the "mds" module is also
4296 reloaded at that time). The default number of threads will
4297 stay the same, 32 for most systems.
4302 Description: OST crash if filesystem is unformatted or corrupt
4303 Details : If an OST is started on a device that has never been formatted
4304 or if the filesystem is corrupt and cannot even mount then the
4305 error handling cleanup routines would dereference a NULL pointer.
4310 Description: NULL pointer deref in llap_from_page.
4311 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4312 incorrect) mapping if the page was truncated/reclaimed while it was
4313 searched for. Check for this condition and skip such pages when
4314 doing readahead. Introduce extra check to llap_from_page() to
4315 verify page->mapping->host is non-NULL (so page is not anonymous).
4318 Frequency : Sometimes when using sys_sendfile
4320 Description: "page not covered by a lock" warnings from ll_readpage
4321 Details : sendfile called ll_readpage without right page locks present.
4322 Now we introduced ll_file_sendfile that does necessary locking
4323 around call to generic_file_sendfile() much like we do in
4327 Frequency : with certain MDS communication failures at client mount time
4329 Description: NULL pointer deref after failed client mount
4330 Details : a client connection request may delayed by the network layer
4331 and not be sent until after the PTLRPC layer has timed out the
4332 request. If the client fails the mount immediately it will try
4333 to clean up before the network times out the request. Add a
4334 reference from the request import to the obd device and delay
4335 the cleanup until the network drops the request.
4338 Frequency : occasionally during client (re)connect
4340 Description: assertion failure during client (re)connect
4341 Details : processing a client connection request may be delayed by the
4342 client or server longer than the client connect timeout. This
4343 causes the client to resend the connection request. If the
4344 original connection request is replied in this interval, the
4345 client may trip an assertion failure in ptlrpc_connect_interpret()
4346 which thought it would be the only running connect process.
4349 Frequency : only with obd_echo servers and clients that are rebooted
4351 Description: kernel BUG accessing uninitialized data structure
4352 Details : When running an obd_echo server it did not start the ping_evictor
4353 thread, and when a client was evicted an uninitialized data
4354 structure was accessed. Start the ping_evictor in the RPC
4355 service startup instead of the OBD startup.
4357 Severity : enhancement
4358 Bugzilla : 10193 (patchless)
4359 Description: Remove dependency on various unexported kernel interfaces.
4360 Details : No longer need reparent_to_init, exit_mm, exit_files,
4361 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4364 Frequency : rare (only users of deprecated and unsupported LDAP config)
4366 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4367 Details : LDAP apparently contains 'lustreName' attributes instead of
4368 'name'. A simple remapping of the name is sufficient.
4371 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4373 Description: waiting_locks_callback trips kernel BUG if client is evicted
4374 Details : Running with the dump_on_timeout debug flag turned on makes
4375 it possible that the waiting_locks_callback() can try to dump
4376 the Lustre kernel debug logs from an interrupt handler. Defer
4377 this log dumping to the expired_lock_main() thread.
4379 Severity : enhancement
4381 Description: Support NFS exporting on 2.6 kernels.
4382 Details : Implement non-rawops metadata methods for NFS server to use without
4383 changing NFS server code.
4386 Frequency : very rare (synthetic metadata workload only)
4388 Description: two racing renames might cause an MDS thread to deadlock
4389 Details : Running the "racer" program may cause one MDS thread to rename
4390 a file from being the source of a rename to being the target of
4391 a rename at exactly the same time that another thread is doing
4392 so, and the second thread has already enqueued these locks after
4393 doing a lookup of the target and is trying to relock them in
4394 order. Ensure that we don't try to re-lock the same resource.
4397 Frequency : only very large systems with liblustre clients
4399 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4400 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4401 all OSTs in parallel. This allows the eviction of stale liblustre
4402 clients to proceed much faster than if they were done in series,
4403 and also offers similar improvements for other set_info RPCs.
4408 Description: excessive CPU usage during initial read phase on client
4409 Details : During the initial read phase on a client, it would agressively
4410 retry readahead on the file, consuming too much CPU and impacting
4411 performance (since 1.4.5.8). Improve the readahead algorithm
4412 to avoid this, and also improve some other common cases (read
4413 of small files in particular, where "small" is files smaller than
4414 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4419 Description: MDS crash when receiving packet with unknown intent.
4420 Details : Do not LBUG in unknown intent case, just return -EFAULT
4422 Severity : enhancement
4423 Bugzilla : 9293, 9385
4424 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4425 Details : Do not serialize getattr (non-intent version) and statfs.
4428 Frequency : occasional, when OST network is overloaded/intermittent
4430 Description: client evicted by OST after bulk IO timeout
4431 Details : If a client sends a bulk IO request (read or write) the OST
4432 may evict the client if it is unresposive to its data GET/PUT
4433 request. This is incorrect if the network is overloaded (takes
4434 too long to transfer the RPC data) or dropped the OST GET/PUT
4435 request. There is no need to evict the client at all, since
4436 the pinger and/or lock callbacks will handle this, and the
4437 client can restart the bulk request.
4440 Frequency : Always when mmapping file with no objects
4442 Description: client crashes when mmapping file with no objects
4443 Details : Check that we actually have objects in a file before doing any
4444 operations on objects in ll_vm_open, ll_vm_close and
4450 Description: Request leak when working with deleted CWD
4451 Details : Introduce advanced request refcount tracking for requests
4452 referenced from lustre intent.
4454 Severity : Enhancement
4456 Description: Cache open file handles on client.
4457 Details : MDS now will return special lock along with openhandle, if
4458 requested and client is allowed to hold openhandle, even if unused,
4459 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4460 closing files for every read/write openration.
4462 Severity : Enhancement
4464 Description: Cache open negative dentries on client when possible.
4465 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4466 negative dentries on lock revocation.
4471 Description: Remounting a client read-only wasn't possible with a zconf mount
4472 Details : It wasn't possible to remount a client read-only with llmount.
4474 Severity : enhancement
4475 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4476 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4477 MPICH is included to add Lustre-specific ADIO interfaces.
4478 This is based closely on the UFS ADIO layer and only differs
4479 in file creation, in order to allow the OST striping to be set.
4480 This is user-contributed code and not supported by CFS.
4485 Description: extended inode attributes (immutable, append-only) work improperly
4486 when 2.4 and 2.6 kernels are used on client/server or vice versa
4487 Details : Introduce kernel-independent values for these flags.
4489 Severity : enhancement
4492 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4493 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4494 granularity. Now, Fractional MB granularity is be supported,
4495 this is very useful for low memory system.
4497 Severity : enhancement
4499 Description: Getattr by fid
4500 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4501 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4502 have locks and avoids getattr rpc requests for it on every stat.
4505 Frequency : Always, for filesystems larger than 2TB
4507 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4508 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4509 filesystems larger than 2TB to be mounted with mballoc enabled.
4512 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4514 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4515 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4516 filesystems larger than 2TB is mounted then the kernel will
4517 silently corrupt the start of the filesystem. CONFIG_LBD is
4518 enabled for all CFS-supported kernels, but the possibility of
4519 this happening with a modified kernel config exists.
4521 Severity : enhancement
4523 Description: add client O_DIRECT support for 2.6 kernels
4524 Details : It is now possible to do O_DIRECT reads and writes to files
4525 in the Lustre client mountpoint on 2.6 kernel clients.
4527 Severity : enhancement
4529 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4530 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4531 parallel, not waiting for response from every OST before sending
4532 a rpc to the next OST.
4537 Description: setattr vs write race when updating file timestamps
4538 Details : Client processes that update a file timestamp into the past
4539 right after writing to the file (e.g. tar) it is possible that
4540 the updated file modification time can be reset to the current
4541 time due to a race between processing the setattr and write RPC.
4543 Severity : enhancement
4545 Description: Bring 'lfs find' closer in line with regular Linux find.
4546 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4547 -print0 options and obtains all the needed info through the lustre
4550 Severity : enhancement
4552 Description: support up to 1024 configured devices on one node
4553 Details : change obd_dev array from statically allocated to dynamically
4554 allocated structs as they are first used to reduce memory usage
4559 Description: Flush dirty partially truncated pages during truncate
4560 Details : Immediatelly flush partially truncated pages in filter_setattr,
4561 this way we completely avoid having any pages in page cache on OST
4562 and can retire ugly workarounds during writes to flush such pages.
4567 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4568 Details : protect inode from truncation within vfs_unlink() context
4569 just take a reference before calling vfs_unlink() and release it
4570 when parent's i_sem is free.
4575 Description: last_id value checked outside lock on OST caused LASSERT failure
4576 Details : If there were multiple MDS->OST object precreate requests in
4577 flight, it was possible that the OST's last object id was checked
4578 outside a lock and incorrectly tripped an assertion. Move checks
4579 inside locks, and discard old precreate requests.
4582 Frequency : always, if extents are used on OSTs
4584 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4585 Details : OSTs using the extents format would not zero the high 16 bits of
4586 the index physical block number. This is not a problem for any
4587 OST filesystems smaller than 16TB, and no kernels support ext3
4588 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4589 new/modified files) and can be fixed for existing filesystems
4590 with e2fsprogs-1.39-cfs1.
4595 Description: import connection selection may be incorrect if timer wraps
4596 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4597 import connections to be ignored if the 32-bit jiffies counter
4598 wraps. Use a 64-bit jiffies counter.
4601 Frequency : very large clusters immediately after boot
4603 Description: LNET request buffers exhausted under heavy short-term load
4604 Details : If a large number of client requests are generated on a service
4605 that has previously never seen so many requests it is possible
4606 that the request buffer growth cannot keep up with the spike in
4607 demand. Instead of dropping incoming requests, they are held in
4608 the LND until the RPC service can accept more requests.
4611 Frequency : Sometimes during replay
4613 Description: Assertion failure in ll_local_open after replay.
4614 Details : If replay happened on an open request reply before we were able
4615 to set replay handler, reply will become not swabbed tripping the
4616 assertion in ll_local_open. Now we set the handler right after
4617 recognising of open request
4620 Frequency : very rare
4622 Description: kernel reports "badness in vsnprintf"
4623 Details : Reading from the "recovery_status" /proc file in small chunks
4624 may cause a negative length in lprocfs_obd_rd_recovery_status()
4625 call to vsnprintf() (which is otherwise harmless). Exit early
4626 if there is no more space in the output buffer.
4628 Severity : enhancement
4630 Description: clear OBD RPC statistics by writing to them
4631 Details : It is now possible to clear the OBD RPC statistics by writing
4632 to the "stats" file.
4637 Description: Client mtime is not the same on different clients after utimes
4638 Details : In some cases, the client was using the utimes() syscall on
4639 a file cached on another node. The clients now validate the
4640 ctime from the MDS + OSTs to determine which one is right.
4645 Description: Inability to activate failout mode
4646 Details : lconf script incorrectly assumed that in pythong string's numeric
4647 value is used in comparisons.
4650 Frequency : always with multiple stripes per file
4652 Description: Inefficient object allocation for mutli-stripe files
4653 Details : When selecting which OSTs to stripe files over, for files with
4654 a stripe count that divides evenly into the number of OSTs,
4655 the MDS is always picking the same starting OST for each file.
4656 Return the OST selection heuristic to the original design.
4661 Description: mount failures may take full timeout to return an error
4662 Details : Under some heavy load conditions it is possible that a
4663 failed mount can wait for the full obd_timeout interval,
4664 possibly several minutes, before reporting an error.
4665 Instead return an error as soon as the status is known.
4667 Frequency : quota enabled and large files being deleted
4669 Description: releasing more than 4GB of quota at once hangs OST
4670 Details : If a user deletes more than 4GB of files on a single OST it
4671 will cause the OST to spin in an infinite loop. Release
4672 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4677 Description: statfs data retrieved from /proc may be stale or zero
4678 Details : When reading per-device statfs data from /proc, in the
4679 {kbytes,files}_{total,free,avail} files, it may appear
4680 as zero or be out of date.
4683 Frequency : systems with MD RAID1 external journal devices
4685 Description: lconf's call to blkid is confused by RAID1 journal devices
4686 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4687 of returning all block devices that match the journal UUID.
4690 Frequency : always, for aggregate stripe size over 4GB
4692 Description: assertion fails when trying to use 4GB stripe size
4693 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4694 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4697 Frequency : always on ppc64
4699 Description: the first write on an ext3 filesystem with mballoc got stuck
4700 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4701 perform endianness conversion.
4703 ------------------------------------------------------------------------------
4705 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4707 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4708 INTEROPERATE with older versions automatically. Please read the
4709 user documentation before upgrading any part of a live system.
4710 * WARNING: Lustre networking configuration changes are required with
4711 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4714 * Support for kernels:
4715 2.6.9-22.0.2.EL (RHEL 4)
4716 2.6.5-7.244 (SLES 9)
4717 2.6.12.6 vanilla (kernel.org)
4720 Severity : enhancement
4721 Bugzilla : 7981/8208
4722 Description: Introduced Lustre Networking (LNET)
4723 Details : LNET is new networking infrastructure for Lustre, it includes
4724 a reorganized network configuration mode (see the user
4725 documentation for full details) as well as support for routing
4726 between different network fabrics. Lustre Networking Devices
4727 (LNDs) for the supported network fabrics have also been
4728 created for this new infrastructure.
4730 Severity : enhancement
4731 Description: Introduced Access control lists
4732 Details : clients can set ACLs on files and directories in order to have
4733 more fine-grained permissions than the standard Unix UGO+RWX.
4734 The MDS must be started with the "-o acl" mount option.
4736 Severity : enhancement
4737 Description: Introduced filesystem quotas
4738 Details : Administrators may now establish per-user quotas on the
4741 Severity : enhancement
4743 Description: Configuration change for the XT3
4744 The PTLLND is now used to run Lustre over Portals on the XT3
4745 The configure option(s) --with-cray-portals are no longer used.
4746 Rather --with-portals=<path-to-portals-includes> is used to
4747 enable building on the XT3. In addition to enable XT3 specific
4748 features the option --enable-cray-xt3 must be used.
4753 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4754 Details : A race between allocating a new llcd and re-getting the llcd_lock
4755 allowed another thread to grab newly-allocated llcd.
4757 Severity : enhancement
4759 Description: 2.6 OST async journal commit and locking fix to improve performance
4760 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4761 2.6 kernels are now async as they already were in 2.4 kernels so
4762 that they can commit concurrently with the network bulk transfer.
4763 For block-allocated files the filter allocation semaphore is held
4764 to avoid filesystem fragmentation during allocation. BKL lock
4765 removed for 2.6 xattr operations where it is no longer needed.
4770 Description: lconf incorrectly determined whether two IP networks could talk
4771 Details : In some more complicated routing and multiple-network
4772 configurations, lconf will avoid trying to make a network
4773 connection to a disjoint part of the IP space. It was doing the
4774 math incorrectly for one set of cases.
4779 Description: Fix for potential infinite loop processing records in an llog.
4780 Details : If an llog record is corrupted/zeroed, it is possible to loop
4781 forever in llog_process(). Validate the llog record length
4782 and skip the remainder of the block on error.
4785 Frequency : occasional (liblustre only)
4787 Description: liblustre could not open files whose last component is a symlink
4788 Details : sysio_path_walk() would incorrectly pass the open intent to
4789 intermediate path components.
4792 Frequency : rare (liblustre only with non-standard tuning)
4793 Bugzilla : 7201 (7350)
4794 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4795 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4796 after it has been released, only do this for referenced locks
4798 Severity : enhancement
4800 Description: specifying an (invalid) directory default stripe_size of -1
4801 would reset the directory default striping
4802 Details : stripe_size -1 was used internally to signal directory stripe
4803 removal, now use "all default" to signal dir stripe removal
4804 as a directory striping of "all default" is not useful
4807 Frequency : common for large clusters running liblustre clients
4809 Description: doing an ls when liblustre clients are running is slow
4810 Details : sending a glimpse AST to a liblustre client waits for every AST
4811 to time out, as liblustre clients will not respond. Since they
4812 cannot cache data we refresh the OST lock LVB from disk instead.
4814 Severity : enhancement
4816 Description: doing an ls at the same time as file IO can be slow
4817 Details : enqueue and other "small" requests can be blocked behind many
4818 large IO requests. Create a new OST IO portal for non-IO
4819 requests so they can be processed faster.
4822 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4824 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4825 created file with mode 000
4826 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4827 allow an ftruncate() as a non-root user to a file with mode 000.
4828 Lustre now respects this flag to disable mode checks when
4829 truncating a file owned by the user
4832 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4835 Description: Revoking locks from clients that went dead or catatonic might take
4837 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4838 cancellation of such locks instant on servers without waiting for
4839 any reply from clients. Clients drops these locks when cancel
4840 notification from server is received without replying.
4843 Frequency : liblustre-only, when liblustre client dies or becomes busy
4845 Description: Doing ls on Linux clients can take a long time with active
4847 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4848 granting such locks to it in the first place if possible. Locks
4849 are taken by proxy on the OST during the read or write and
4850 dropped immediately afterward. Add connect flags handling, do
4851 not grant locks to liblustre clients for glimpse ASTs.
4853 Severity : enhancement
4855 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4856 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4857 sequential reads and behaves badly with applications doing random
4858 reads. Improve it by reading ahead at least read region, and
4859 avoiding excessive large RPC for small reads.
4861 Severity : enhancement
4863 Description: Creating more than 1000 files for a single job may cause a load
4864 imbalance on the OSTs if there are also a large number of OSTs.
4865 Details : qos_prep_create() uses an OST index reseed value that is an
4866 even multiple of the number of available OSTs so that if the
4867 reseed happens in the middle of the object allocation it will
4868 still utilize the OSTs as uniformly as possible.
4873 Description: OST or MDS may oops in ping_evictor_main()
4874 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4875 but doesn't restart at beginning of obd_exports_timed list
4878 Severity : enhancement
4880 Description: improve by-nid export eviction on the MDS and OST
4881 Details : allow multiple exports with the same NID to be evicted at one
4882 time without re-searching the exports list.
4885 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4887 Description: MDS may oops in groups_free()
4888 Details : in rare race conditions a newly allocated group_info struct is
4889 freed again, and this can be NULL. The 2.4 compatibility code
4890 for groups_free() checked for a NULL pointer, but 2.6 did not.
4893 Frequency : common for liblustre clients doing little filesystem IO
4894 Bugzilla : 9352, 7313
4895 Description: server may evict liblustre clients accessing contended locks
4896 Details : if a client is granted a lock or receives a completion AST
4897 with a blocking AST already set it would not reply to the AST
4898 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4901 Frequency : lfs setstripe, only systems with more than 160 OSTs
4903 Description: unable to set striping with a starting offset beyond OST 160
4904 Details : llapi_create_file() incorrectly limited the starting stripe
4905 index to the maximum single-file stripe count.
4908 Frequency : LDAP users only
4910 Description: lconf did not handle in-kernel recovery with LDAP properly
4911 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4913 Severity : enhancement
4915 Description: bind OST threads to NUMA nodes to improve performance
4916 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4917 node and do their allocations there to localize memory access
4919 Severity : enhancement
4921 Description: llmount can determine client NID directly from Myrinet (GM)
4922 Details : the client NID code from gmnalnid was moved directly into
4923 llmount, removing the need to use this or specifying the
4924 client NID explicitly when mounting GM clients with zeroconf
4927 Frequency : if client is started with down MDS
4929 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4930 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4931 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4932 ptlrpc_import_delayed_req() because we didn't check if the
4933 request was interrupted, and we also didn't break out of the
4934 event loop if there was no timeout
4939 Description: data loss during non-page-aligned writes to a single file from
4940 both multiple nodes and multiple threads on one node at same time
4941 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4942 inconsistency led to false short-reads, that were cached and later
4943 used by ->prepare_write() to fill in partially written page,
4944 leading to data loss.
4947 Frequency : always, if lconf --abort_recovery used
4949 Description: lconf --abort_recovery fails with 'Operation not supported'
4950 Details : lconf was attempting to abort recovery on the MDT device and not
4953 Severity : enhancement
4955 Description: remove cleanup logs
4956 Details : replace lconf-generated cleanup logs with lustre internal
4957 cleanup routines. Eliminates the need for client-cleanup and
4960 Severity : enhancement
4962 Description: add support for EAs (user and system) on lustre filesystems
4963 Details : it is now possible to store extended attributes in the Lustre
4964 client filesystem, and with the user_xattr mount option it
4965 is possible to allow users to store EAs on their files also
4967 Severity : enhancement
4969 Description: Add possibility (config option) to show minimal available OST free
4971 Details : When compiled with --enable-mindf configure option, statfs(2)
4972 (and so, df) will return least minimal free space available from
4973 all OSTs as amount of free space on FS, instead of summary of
4974 free spaces of all OSTs.
4976 Severity : enhancement
4978 Description: do not expand extent locks acquired on OST-side
4979 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4980 by server: they are not cached anyway.
4983 Frequency : when mmap is used/binaries executed from Lustre
4985 Description: Unmmap pages before throwing them away from read cache.
4986 Details : llap_shrink cache now attempts to unmap pages before discarding
4987 them (if unmapping failed - do not discard). SLES9 kernel has
4988 extra checks that trigger if this unmapping is not done first.
4993 Description: lconf didn't resolve symlinks before checking to see whether a
4994 given mountpoint was already in use
4997 Frequency : when migrating failover services
4998 Bugzilla : 6395, 9514
4999 Description: When migrating a subset of services from a node (e.g. failback
5000 from a failover service node) the remaining services would
5001 time out and evict clients.
5002 Details : lconf --force (implied by --failover) sets the global obd_timeout
5003 to 5 seconds in order to quickly disconnect, but this caused
5004 other RPCs to time out too quickly. Do not change the global
5005 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5007 Severity : enhancement
5008 Frequency : if MDS is started with down OST
5009 Bugzilla : 9439,5706
5010 Description: Allow startup/shutdown of an MDS without depending on the
5011 availability of the OSTs.
5012 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5013 Add appropriate locking and lov-osc refcounts for safe
5014 cleaning. Add osc abort_inflight calls in case the
5015 synchronize never started.
5018 Frequency : occasional (Cray XT3 only)
5020 Description: root not authorized to access files in CRAY_PORTALS environment
5021 Details : The client process capabilities were not honoured on the MDS in
5022 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5023 been accessed by an authorized user then root was able to access
5024 the file on the local client also. The root user capabilities
5025 are now allowed on the MDS, as this environment has secure UID.
5028 Frequency : occasional
5030 Description: ldiskfs "too long searching" message happens too often
5031 Details : A debugging message (otherwise harmless) prints too often on
5032 the OST console. This has been reduced to only happen when
5033 there are fragmentation problems on the filesystem.
5038 Description: Division by zero in statfs when all OSCs are inactive
5039 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5040 lov_statfs divides by value returned by lov_get_stripecnt().
5044 Bugzilla : 9489, 3273
5045 Description: First write from each client to each OST was only 4kB in size,
5046 to initialize client writeback cache, which caused sub-optimal
5047 RPCs and poor layout on disk for the first written file.
5048 Details : Clients now request an initial cache grant at (re)connect time
5049 and so that they can start streaming writes to the cache right
5050 away and always do full-sized RPCs if there is enough data.
5051 If the OST is rebooted the client also re-establishes its grant
5052 so that client cached writes will be honoured under the grant.
5057 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5058 Details : Now I/O RPCs go to different portal number and (presumably) fast
5059 lock requests (and glimses) and other RPCs get their own service
5060 threads pool that should be able to service those RPCs
5063 Severity : enhancement
5065 Description: Ability to exchange lustre version between client and servers and
5066 issue warnings at client side if client is too old. Also for
5067 liblustre clients there is ability to refuse connection of too old
5069 Details : New 'version' field is added to connect data structure that is
5070 filled with version info. That info is later checked by server and
5074 Frequency : rare, liblustre only.
5075 Bugzilla : 9296, 9581
5076 Description: Two simultaneous writes from liblustre at offset within same page
5077 might proceed at the same time overwriting eachother with stale
5079 Details : I/O lock withing llu_file_prwv was released too early, before data
5080 actually was hitting the wire. Extended lock-holding time until
5081 server acknowledges receiving data.
5084 Frequency : extremely rare. Never observed in practice.
5086 Description: avoid generating lustre_handle cookie of 0.
5087 Details : class_handle_hash() generates handle cookies by incrementing
5088 global counter, and can hit 0 occasionaly (this is unlikely, but
5089 not impossible, because initial value of cookie counter is
5090 selected randonly). Value of 0 is used as a sentinel meaning
5091 "unassigned handle" --- avoid it. Also coalesce two critical
5092 sections in this function into one.
5094 Severity : enhancement
5096 Description: allow liblustre clients to delegate truncate locking to OST
5097 Details : To avoid overhead of locking, liblustre client instructs OST to
5098 take extent lock in ost_punch() on client's behalf. New connection
5099 flag is added to handle backward compatibility.
5101 Severity : enhancement
5102 Bugzilla : 4928, 7341, 9758
5103 Description: allow number of OST service threads to be specified
5104 Details : a module parameter allows the number of OST service threads
5105 to be specified via "options ost ost_num_threads={N}" in the
5106 OSS's /etc/modules.conf or /etc/modprobe.conf.
5110 Bugzilla : 6146, 9635, 9895
5111 Description: servers crash with bad pointer in target_handle_connect()
5112 Details : In rare cases when a client is reconnecting it was possible that
5113 the connection request was the last reference for that export.
5114 We would temporarily drop the export reference and get a new
5115 one, but this may have been the last reference and the export
5116 was just destroyed. Get new reference before dropping old one.
5118 Severity : enhancement
5119 Frequency : if client is started with failover MDS
5121 Description: Allow multiple MDS hostnames in the mount command
5122 Details : Try to read the configuration from all specified MDS
5123 hostnames during a client mount in case the "primary"
5126 Severity : enhancement
5128 Description: Stop sending data to evicted clients as soon as possible.
5129 Details : Check if the client we are about to send or are sending data to
5130 was evicted already. (Check is done every second of waiting,
5131 for which l_wait_event interface was extended to allow checking
5132 of exit condition at specified intervals).
5135 Frequency : rare, normally only when NFS exporting is done from client
5137 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5139 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5142 Frequency : always, if srand() is called before liblustre initialization
5144 Description: Liblustre uses system PRNG disturbing its usage by user application
5145 Details : Introduce internal to lustre fast and high-quality PRNG for
5146 lustre usage and make liblustre and some other places in generic
5147 lustre code to use it.
5149 Severity : enhancement
5150 Bugzilla : 9477, 9557, 9870
5151 Description: Verify that the MDS configuration logs are updated when xml is
5152 Details : Check if the .xml configuration logs are newer than the config
5153 logs stored on the MDS and report an error if this is the case.
5154 Request --write-conf, or allow starting with --old_conf.
5156 Severity : enhancement
5158 Description: Handle symlinks in the path when checking if Lustre is mounted.
5159 Details : Resolve intermediate symlinks when checking if a client has
5160 mounted a filesystem to avoid duplicate client mounts.
5165 Description: lconf can hit an error exception but still return success.
5166 Details : The lconf command catches the Command error exception at the top
5167 level script context and will exit with the associated exit
5168 status, but doesn't ensure that this exit status is non-zero.
5173 Description: failure of ptlrpc thread startup can cause oops
5174 Details : Starting a ptlrpc service thread can fail if there are a large
5175 number of threads or the server memory is very fragmented.
5176 Handle this without oopsing.
5179 Frequency : always, only if liblustre and non-default acceptor port was used
5181 Description: liblustre cannot connect to servers with non-default acceptor port
5182 Details : tcpnal_set_default_params() was not called and was therefore
5183 ignoring the environment varaible TCPNAL_PORT, as well as other
5184 TCPNAL_ environment variables
5189 Description: two objects could be created on the same OST for a single file
5190 Details : If an OST is down, in some cases it was possible to create two
5191 objects on a single OST for a single file. No problems other
5192 than potential performance impact and spurious error messages.
5196 Bugzilla : 5681, 9562
5197 Description: Client may oops in ll_unhash_aliases
5198 Details : Client dcache may become inconsistent in race condition.
5199 In some cases "getcwd" can fail if the current directory is
5205 Description: Inode refcounting problems in NFS export code
5206 Details : link_raw functions used to call d_instantiate without obtaining
5207 extra inode reference first.
5211 Bugzilla : 9942, 9903
5212 Description: Referencing freed requests leading to crash, memleaks with NFS.
5213 Details : We used to require that call to ll_revalidate_it was always
5214 followed by ll_lookup_it. Also with revalidate_special() it is
5215 possible to call ll_revalidate_it() twice for the same dentry
5216 even if first occurence returned success. This fix changes semantic
5217 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5218 reference on a request referred from the intent.
5219 ll_intent_release() then releases such a request.
5222 Frequency : rare, normally benchmark loads only
5224 Description: unlinked inodes were kept in memory on the client
5225 Details : If a client is repeatedly creating and unlinking files it
5226 can accumulate a lot of stale inodes in the inode slab cache.
5227 If there is no other client load running this can cause the
5228 client node to run out of memory. Instead flush old inodes
5229 from client cache that have the same inode number as a new inode.
5232 Frequency : SLES9 2.6.5 kernel and long filenames only
5233 Bugzilla : 9969, 10379
5234 Description: utime reports stale NFS file handle
5235 Details : SLES9 uses out-of-dentry names in some cases, which confused
5236 the lustre dentry revalidation. Change it to always use the
5240 Frequency : rare, unless heavy write-truncate concurrency is continuous
5241 Bugzilla : 4180, 6984, 7171, 9963, 9331
5242 Description: OST becomes very slow and/or deadlocked during object unlink
5243 Details : filter_destroy() was holding onto the parent directory lock
5244 while truncating+unlinking objects. For very large objects this
5245 may block other threads for a long time and slow overall OST
5246 responsiveness. It may also be possible to get a lock ordering
5247 deadlock in this case, or run out of journal credits because of
5248 the combined truncate+unlink. Solution is to do object truncate
5249 first in one transaction without parent lock, and then do the
5250 final unlink in a new transaction with the parent lock. This
5251 reduces the lock hold time dramatically.
5254 Frequency : rare, 2.4 kernels only
5256 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5257 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5258 before unmounting their filesystem in order to avoid deadlock.
5259 The kernel_locked() function in 2.4 kernels only checks whether
5260 the kernel lock is held, not whether it is this process that is
5261 holding it as 2.6 kernels do.
5266 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5267 Details : The client ptlrpc code may be trying to reconnect to a down
5268 server before a previous connection attempt has timed out.
5269 Increase the reconnect interval to be longer than the connection
5270 timeout interval to avoid sending duplicate connections to
5274 Frequency : echo_client brw_test command
5276 Description: fix echo_client to work with OST preallocated code
5277 Details : OST preallocation code (5137) didn't take echo_client IO path
5278 into account: echo_client calls filter methods outside of any
5279 OST thread and, hence, there is no per-thread preallocated
5280 pages and buffers to use. Solution: hijack pga pages for IO. As
5281 a byproduct, this avoids unnecessary data copying.
5285 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5286 Description: Client can oops in mdc_commit_close() after open replay
5287 Details : It was possible for the MDS to return an open request with no
5288 transaction number in mds_finish_transno() if the client was
5289 evicted, but without actually returning an error. Clients
5290 would later try to replay that open and may trip an assertion
5291 Simplify the client close codepath, and always return an error
5292 from the MDS in case the open is not successful.
5295 Frequency : rare, 2.6 OSTs only
5297 Description: OST may deadlock under high load on fragmented files
5298 Details : If there was a heavy load and highly-fragmented OST filesystems
5299 it was possible to have all the OST threads deadlock waiting on
5300 allocation of biovecs, because the biovecs were not released
5301 until the entire RPC IO was completed. Instead, release biovecs
5302 as soon as they are complete to ensure forward IO progress.
5304 Severity : enhancement
5306 Description: Support for specifying external journal device at mount
5307 Details : If an OST or MDS device is formatted with an external journal
5308 device, this device major/minor is stored in the ext3 superblock
5309 and may not be valid for failover. Allow detecting and
5310 specifying the external journal at mount time.
5315 Description: Mounting an MDS with pending unlinked files may cause oops
5316 Details : target_finish_recovery() calls mds_postrecov() which returned
5317 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5318 considers this an error and immediately begins cleaning up the
5319 lov, just after starting the mds_lov process
5321 Severity : enhancement
5323 Description: Implement 'lfs df' to report actual free space on per-OST basis
5324 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5325 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5326 usage of objects; '-h' to report in human readable format.
5328 ------------------------------------------------------------------------------
5330 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5337 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5338 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5339 at startup during orphan unlinking. Instead, skip buddy update
5340 before setup, as it will be regenerated after recovery is complete.
5345 Description: If an OST is inactive, its locks might reference stale inodes.
5346 Details : lov_change_cbdata() must iterate over all namespaces, even if
5347 they are inactive to clear inode references from the lock.
5349 Severity : enhancement
5350 Frequency : occasional, if non-standard max_dirty_mb used
5352 Description: Client will block write RPCs if not enough grant
5353 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5354 then the client will block writes while waiting for another RPC
5355 to complete instead of consuming its dirty limit. With change
5356 we get improved performance when max_dirty_mb is small.
5358 Severity : enhancement
5359 Bugzilla : 3389, 6253
5360 Description: Add support for supplementary groups on the MDS.
5361 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5362 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5363 lookups for user supplementary groups into a cache.
5367 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5368 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5369 for liblustre O_EXCL is enforced by MDS.
5373 Description: readdir never returns NULL in liblustre.
5374 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5375 dentry in struct dirent.
5379 Frequency : liblustre only, depends on application IO pattern
5380 Description: liblustre clients evicted if not contacting servers
5381 Details : Don't put liblustre clients into the ping_evictor list, so
5382 they will not be evicted by the pinger ever.
5384 Severity : enhancement
5386 Description: Add ability to evict clients by NID from MDS.
5387 Details : By echoing "nid:$NID" string into
5388 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5389 $NID would be instantly evicted from this MDS and from all active
5390 OSTs connected to it.
5394 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5395 Details : lookup_it_finish() used to query file size from OSTs that was not
5400 Description: service threads change working directory to that of init
5401 Details : Starting lustre service threads may pin the working directory
5402 of the parent thread, making that filesystem busy. Threads
5403 now change to the working directory of init to avoid this.
5407 Frequency : during shutdown only
5408 Description: shutdown with a failed MDS or OST can cause unmount to hang
5409 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5414 Frequency : 2.6 only
5415 Description: chmod/chown may include an extra supplementary group
5416 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5417 supplementary group and if none is specified this is used.
5420 Bugzilla : 5479 (6816)
5422 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5423 Details : If lookup is in progress on a file that is unlinked we might try
5424 to revalidate the inode and fail in revalidate after lookup is
5425 complete and ll_file_open() enqueues the open again but
5426 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5429 Frequency : always, if lconf --abort_recovery used
5431 Description: lconf --abort_recovery fails with 'Operation not supported'
5432 Details : lconf was attempting to abort recovery on the MDT device and not
5435 ------------------------------------------------------------------------------
5437 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5442 Frequency : rare (only unsupported configurations with a node running as an
5444 Bugzilla : 6514, 5137
5445 Description: Mounting a Lustre file system on a node running as an OST could
5447 Details : OSTs now preallocates memory needed to write out data at
5448 startup, instead of when needed, to avoid having to
5449 allocate memory in possibly low memory situations.
5450 Specifically, if the file system is mounted on on OST,
5451 memory pressure could force it to try to write out data,
5452 which it needed to allocate memory to do. Due to the low
5453 memory, it would be unable to do so and the node would
5454 become unresponsive.
5456 Severity : enhancement
5458 Description: Addition of lconf --service command line option
5459 Details : lconf now accepts a '--service <arg>' option, which is
5460 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5462 Severity : enhancement
5464 Description: Failover mode is now the default for OSTs.
5465 Details : By default, OSTs will now run in failover mode. To return to
5466 the old behaviour, add '--failout' to the lmc line for OSTs.
5468 Severity : enhancement
5470 Description: Health checks are now provided for MDS and OSTs
5471 Details : Additional detailed health check information on MSD and OSTs
5472 is now provided through the procfs health_check value.
5475 Frequency : occasional, depends on IO load
5477 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5478 after numerous create/delete cycles
5479 Details : The ext3 inode allocation policy would not allocate new inodes
5480 very well on the OSTs because there are no new directories
5481 being created. Instead we look for groups with free space if
5482 the parent directories are nearly full.
5487 Description: Network or server problems during mount may cause partially
5488 mounted clients instead of returning an error.
5489 Details : The config llog parsing code may overwrite the error return
5490 code during mount error handling, returning success instead
5496 Description: MDS can fail to allocate large reply buffers
5497 Details : After long uptimes the MDS can fail to allocate large reply
5498 buffers (e.g. zconf client mount config records) due to memory
5499 fragmentation or consumption by the buffer cache. Preallocate
5500 some large reply buffers so that these replies can be sent even
5501 under memory pressure.
5505 Frequency : rare (liblustre)
5506 Description: fsx running with liblustre complained that using truncate() to
5507 extend the file doesn't work. This patch corrects that issue.
5508 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5509 ATTR_SIZE and lsm use in llu_setattr_raw.
5513 Frequency : rare, only 2.6 kernels
5514 Description: Unusual file access patterns on the MDS may result in inode
5515 data being lost in very rare circumstances.
5516 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5517 code in the 2.6 kernel caused the inode and/or EA data not to
5518 be read from disk, causing single-file corruption.
5522 Frequency : rare, only 2.6 filesystems using extents
5523 Description: Heavy concurrent write and delete load may cause data corruption.
5524 Details : It was possible under high-load situations to have an extent
5525 metadata block in the block device cache from a just-unlinked
5526 file overwrite a newly-allocated data block. We now unmap any
5527 metadata buffers that alias just-allocated data blocks.
5531 Frequency : filesystems with default stripe_count larger than 77
5532 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5533 Details : lconf specifies an inode size of 4096 bytes when the default
5534 stripe_count is larger than 77. This conflicts with the default
5535 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5536 this case to avoid pinning too much memory for large EAs.
5538 ------------------------------------------------------------------------------
5540 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5545 Frequency : rare (extremely heavy IO load with hundreds of clients)
5547 Description: Client is evicted, gets IO error writing to file
5548 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5549 caused clients to be evicted for AST timeouts. The fixes in
5550 bug 5192 mean we no longer need to have such short AST timeouts
5551 so ldlm_timeout has been increased.
5554 Frequency : occasional during --force or --failover shutdown under load
5555 Bugzilla : 5949, 4834
5556 Description: Server oops/LBUG if stopped with --force or --failover under load
5557 Details : a collection of import/export refcount and cleanup ordering
5558 issues fixed for safer force cleanup
5561 Frequency : only filesystems larger than 120 OSTs
5562 Bugzilla : 5990, 6223
5563 Description: lfs getstripe would oops on a very large filesystem
5564 Details : lov_getconfig used kfree on vmalloc'd memory
5567 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5568 Bugzilla : 6242, 6243
5569 Description: reading from files that had been truncated to a non-zero size
5570 but never opened returned no data
5571 Details : ll_file_read() reads zeros from no-object files to EOF
5576 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5577 Details : There's little harm in aborting MDS/OSS recovery and letting it
5578 try again, so I removed the LASSERT and return an error instead.
5580 Severity : enhancement
5582 Description: New debugging infrastructure for tracking down data corruption
5583 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5584 (b) cover more of the client-side code path, and (c) try to narrow
5585 down where problems occurred
5589 Bugzilla : 3819, 4364, 4397, 6313
5590 Description: Racing close and eviction MDS could cause assertion in mds_close
5591 Details : It was possible to get multiple mfd references during close and
5592 client eviction, leading to one thread referencing a freed mfd.
5594 Severity: : enhancement
5595 Bugzilla : 3262, 6359
5596 Description: Attempts to reconnect to servers are now more aggressive.
5597 Details : This builds on the enhanced upcall-less recovery that was added
5598 in 1.4.2. When trying to reconnect to servers, clients will
5599 now try each server in the failover group every 10 seconds. By
5600 default, clients would previously try one server every 25 seconds.
5605 Description: After recovery, certain operations trigger a failed
5606 assertion on a client.
5607 Details : Failing over an mds, using lconf -d --failover, while a
5608 client was doing a readdir() call would cause the client to
5609 LBUG after recovery completed and the readdir() was resent.
5611 Severity : enhancement
5613 Description: Default groups are now added by lconf
5614 Details : You can now run lconf --group <servicename> without having to
5615 manually add groups with lmc.
5618 Frequency : occasional
5620 Description: Nodes with an elan id of 0 trigger a failed assertion
5623 Frequency : always when accessing e.g. tty/console device nodes
5625 Description: tty and some other devices nodes cannot be used on lustre
5626 Details : file's private_data field is used by device data and lustre
5627 values in there got lost. New field was added to struct file to
5628 store fs-specific private data.
5631 Frequency : when exporting Lustre via NFS
5633 Description: NFSD failed occasionally when looking up a path component
5634 Details : NFSD is looking up ".." which was broken in ext3 directories
5635 that had grown large enough to become hashed.
5638 Frequency : Clusters with multiple interfaces not on the same subnet
5640 Description: Nodes will repeatedly try to reconnect to an interface which it
5641 cannot reach and report an error to the log.
5642 Details : Extra peer list entries will be created by lconf with some peers
5643 unreachable. lconf now validates the peer before adding it.
5646 Frequency : Only if a default stripe is set on the filesystem root.
5648 Description: Setting a default stripe on the filesystem root prevented the
5649 filesystem from being remounted.
5650 Details : The client was sending extra request flags in the root getattr
5651 request and did not allocate a reply buffer for the dir EA.
5654 Frequency : occasional, higher if lots of files are accessed by one client
5655 Bugzilla : 6159, 6097
5656 Description: Client trips assertion regarding lsm mismatch/magic
5657 Details : While revalidating inodes the VFS looks up inodes with ifind()
5658 and in rare cases can find an inode that is being freed.
5659 The ll_test_inode() code will free the lsm during ifind()
5660 when it finds an existing inode and then the VFS later attaches
5661 this free lsm to a new inode.
5665 Bugzilla : 6422, 7030
5666 Description: MDS deadlock between mkdir and client eviction
5667 Details : Creating a new file via mkdir or mknod (starting a transaction
5668 and getting the ns lock) can deadlock with client eviction
5669 (gets ns lock and trying to finish a synchronous transaction).
5672 Frequency : occasional
5673 Description: While starting a server, the fsfilt_ext3 module could not be
5675 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5676 kernels. Previously, lconf would still use the ext3 name
5677 when trying to load modules. Now, it will correctly use
5678 ext3 on 2.4 and ldiskfs on 2.6.
5680 Severity : enhancement
5681 Description: The default stripe count has been changed to 1
5682 Details : The interpretation of the default stripe count (0, to lfs
5683 or lmc) has been changed to mean striping across a single
5684 OST, rather than all available. For general usage we have
5685 found a stripe count of 1 or 2 works best.
5687 Severity : enhancement
5688 Description: Add support for compiling against Cray portals.
5689 Details : Conditional compiling for some areas that are different
5693 Frequency : occasional
5694 Bugzilla : 6409, 6834
5695 Description: Creating files with an explicit stripe count may lead to
5696 a failed assertion on the MDS
5697 Details : If some OSTs are full or unavailable, creating files may
5698 trigger a failed assertion on the MDS. Now, Lustre will
5699 try to use other servers or return an error to the
5703 Frequency : occasional
5705 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5706 caused lock timeout during readahead (since 1.4.2).
5707 Details : Processes doing readahead might match a lock that hasn't been
5708 granted yet if there are overlapping and conflicting lock
5709 requests. The readahead process waits on ungranted lock
5710 (original lock is CBPENDING), while OST waits for that process
5711 to cancel CBPENDING read lock and eventually evicts client.
5713 Severity : enhancement
5715 Description: Initial enabling of flock support for clients
5716 Details : Implements fcntl advisory locking and file status functions.
5717 This feature is provided as an optional mount flag (default
5718 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5719 locking are implemented yet, and those that are are not guaranteed
5720 to be completely correct in production environments.
5721 mount -t lustre -o [flock|noflock] ...
5724 Frequency : occasional
5726 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5727 assertion in the ext3 JBD (journaling) layer.
5728 Details : The b_committed_data struct is protected by the big kernel lock
5729 in 2.4 kernels, serializing journal_commit_transaction() and
5730 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5731 access to this struct. In 2.6 kernels there is finer grained
5732 locking to improve SMP performance of the JBD layer.
5736 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5741 Description: merge and cleanup kernel patches.
5744 -----------------------------------------------------------------------------
5746 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5748 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5749 versions of Lustre. Please update all servers and clients to
5750 version 1.4.2 or later at the same time. You must also run
5751 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5752 to update the configuration logs.
5754 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5755 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5756 - swab llog records of type '0' so we get proper header size/idx (5861)
5757 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5758 - fix rename of one directory over another leaking an inode (5953)
5759 - avoid SetPageDirty on 2.6 (5981)
5760 - don't re-add just-being-destroyed locks to the waiting list (5653)
5761 - when creating new directories, inherit the parent's custom
5762 striping settings if present parent (3048)
5763 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5764 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5765 - don't hold client locks on temporary worklist from l_lru (5666)
5766 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5767 - automatically evict dead clients (5921)
5768 - Update file size properly in create+truncate+fstat case (6196)
5769 - Do not unhash mountpoint dentries, do not allow removal of
5771 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5772 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5773 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5774 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5775 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5776 - Correctly handle reads of files with no objects (6243)
5777 - lctl recover will also mark a device active if deactivate used (5933)
5779 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5780 - allow --write-conf on an MDS with different nettype than client (5619)
5781 - don't write config llogs to MDS for mounts not from that MDS (5617)
5782 - lconf should create multiple TCP connections from a client (5201)
5783 - init scripts are now turned off by default; run chkconfig --on
5784 lustre and chkconfig --on lustrefs to use them
5785 - upcalls are no longer needed for clients to recover to failover
5787 - add --abort-recovery option to lconf to abort recovery on device
5789 - add support for an arbitrary number of OSTs (3026)
5790 - Quota support protocol changes.
5791 - forward compatibility changes to wire structs (6007)
5792 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5793 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5794 - improved "lustre status" script
5795 - initialize blocksize for non-regular files (6062)
5796 - added --disable-server and --disable-client configure options (5782)
5797 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5798 - Vanilla 2.4.29 support
5799 - increase maximum number of obd devices to 520 (6242)
5800 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5801 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5802 SLES 9 2.6.5 kernels
5803 - limit stripes per file to 160 (the maximum EA size) (6093)
5805 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5808 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5809 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5810 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5811 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5812 - small ext3 extents cleanups and fixes (5733)
5813 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5814 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5815 - avoid needless client->OST connect, fix handle mismatch (5317)
5816 - fix DLM error path that led to out-of-sync client, long delays (5779)
5817 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5818 - fix several locking issues related to i_size (5492,5624,5654,5672)
5819 - don't move pending lock onto export if it is already evicted (5683)
5820 - fix kernel oops when creating .foo in unlinked directory (5548)
5821 - fix deadlock in obdfilter statistics vs. object create (5811)
5822 - use time_{before,after} to avoid timer jiffies wrap (5882)
5823 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5824 - Do not leak request if server was not able to process it (5154)
5825 - If mds_open unable to find parent dir, make that negative lookup(5154)
5826 - don't create new directories with extent-mapping (5909, 5936)
5828 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5829 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5830 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5831 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5832 - support for RHEL4 kernel on i686 (5773)
5833 - provide error messages when incompatible logs are encountered (5898)
5835 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5836 * version 1.4.0.10 (1.4.1 release candidate 1)
5838 - don't keep a lock reference when lock is not granted (4238)
5839 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5840 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5841 - drop import inflight refcount on signal_completed_replay error (5255)
5842 - unlock page after async write error during send (3677)
5843 - handle missing objects in filter_preprw_read properly (5265)
5844 - no transno return for symlink open, don't save no-trasno open (3440)
5845 - don't try to complete elan receive that already failed (4012)
5846 - free RPC server reply state on error (5406)
5847 - clean up thread from ptlrpc_start_thread() on error (5160)
5848 - readahead could read extra page into cache that wasn't ejected (5388)
5849 - prevent races in class_attach/setup/cleanup/detach (5260)
5850 - don't dereference de->d_inode after l_dput of de (5458)
5851 - use "int" for stripe value returned from lock_to_stripe (5544)
5852 - mballoc allocation and error-checking fixes in 2.6 (5504)
5853 - block device patches to fix I/O request sizes in 2.6 (5482)
5854 - look up hostnames for IB nals (5602)
5855 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5856 - don't start multiple acceptors for the same port (5277)
5857 - fix incorrect LASSERT in mds_getattr_name (5635)
5858 - export a proc file for general "ping" checking (5628)
5859 - fix "lfs check" to not block when the MDS is down (5628)
5861 - service request history (4965)
5862 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5863 - create an "evict_client" /proc entry on OSTs, like the MDS has
5864 - fix mount usage message, return errors per mount(8) (5168)
5865 - change grep [] to grep "[]" in tests so they work in more UMLs
5866 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5867 - remove ancient LOV_MAGIC_V0 EA support (5047)
5868 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5869 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5870 - allow readahead allocations to fail when low on memory (5383)
5871 - mmap locking landed again, after considerable improvement (2828)
5872 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5874 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5877 - send OST transaction number in read/write reply to free req (4966)
5878 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5879 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5880 - the watchdog thread now runs as interruptible (5246)
5881 - flock/lockf fixes (but it's still disabled, pending 5135)
5882 - don't use EXT3 constants in llite code (5094)
5883 - memory shortage at startup could cause assertion (5176)
5885 - reorganization of lov code
5886 - single portals codebase
5888 - add extents/mballoc support (5025)
5889 - direct I/O reads in the obdfilter (4048)
5890 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5892 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5895 - send OST transaction number in read/write reply to free req (4966)
5896 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5897 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5898 - the watchdog thread now runs as interruptible (5246)
5899 - handle missing objects in filter_preprw_read properly (5265)
5900 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5901 - drop import inflight refcount on signal_completed_replay error (5255)
5902 - unlock page after async write error during send (3677)
5903 - return original error code on reconstructed replies (3761)
5904 - no transno return for symlink open, don't save no-trasno open (3440)
5906 - add pid to ldlm debugging output (4922)
5907 - bump the watchdog timeouts -- we can't handle 30sec yet
5908 - extra debugging for orphan dentry/inode bug (5259)
5910 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5913 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5914 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5915 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5916 - fix NULL dereference in /proc/sys/portals/routes (4827)
5917 - allow failed mdc_close() operations to be interrupted (4561)
5918 - stop precreate on OST before MDS would time out on it (4778)
5919 - don't send partial-page writes before EOF from client (4410)
5920 - discard client grant for sub-page writes on large-page clients (4520)
5921 - don't free dentries not owned by NFS code, check generation (4806)
5922 - fix lsm leak if mds_create_objects() fails (4801)
5923 - limit debug_daemon file size, always print CERROR messages (4789)
5924 - use transno after validating reply (3892)
5925 - process timed out requests if import state changes (3754)
5926 - update mtime on OST during writes, return in glimpse (4829)
5927 - add mkfsoptions to LDAP (4679)
5928 - use ->max_readahead method instead of zapping global ra (5039)
5929 - don't interrupt __l_wait_event() during strace
5931 - add software watchdogs to catch hung threads quickly (4941)
5932 - make lustrefs init script start after nfs is mounted
5933 - fix CWARN/ERROR duplication (4930)
5934 - return async write errors to application if possible (2248)
5935 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5936 - print NAL number in %x format (4645)
5937 - update barely-supported suse-2.4.21-171 series (4842)
5938 - support for sles 9 %post scripts
5939 - support for building 2.6 kernel-source packages
5940 - support for sles km_* packages
5942 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5945 - ignore -ENOENT errors in osc_destroy (3639)
5946 - notify osc create thread that OSC is being cleaned up (4600)
5947 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5948 - reconstruct ost_handle() like mds_handle() (4657)
5949 - create a new thread to do import eviction to avoid deadlock (3969)
5950 - let lconf resolve symlinked-to devices (4629)
5951 - don't unlink "objects" from directory with default EA (4554)
5952 - hold socknal file ref over connect in case target is down (4394)
5953 - allow more than 32000 subdirectories in a single directory (3244)
5954 - fix blocks count for O_DIRECT writes (3751)
5955 - OST returns ENOSPC from object create when no space left (4539)
5956 - don't send truncate RPC if file size isn't changing (4410)
5957 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5958 - bind to privileged port in socknal and tcpnal (3689)
5960 - rate limit CERROR/CWARN console message to avoid overload (4519)
5961 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5962 - basic mmap support (3918)
5963 - kernel patch series update from b1_4 (4711)
5965 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5968 - avoid crash during MDS cleanup with OST shut down (2775)
5969 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5970 - don't use bad inodes on the MDS (3744)
5971 - dynamic object preallocation to improve recovery speed (4236)
5972 - don't hold spinlock over lock dumping or change debug flags (4401)
5973 - don't zero obd_dev when it is force cleaned (3651)
5974 - print grants to console if they go negative (4431)
5975 - "lctl deactivate" will stop automatic recovery attempts (3406)
5976 - look for existing locks in ldlm_handle_enqueue() (3764)
5977 - don't resolve lock handle twice in recovery avoiding race (4401)
5978 - revalidate should check working dir is a directory (4134)
5980 - don't always mark "slow" obdfilter messages as errors (4418)
5982 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5985 - don't close LustreDB during write_conf until it is done (3860)
5986 - fix typo in lconf for_each_profile (3821)
5987 - allow dumping logs from multiple threads at one time (3820)
5988 - don't allow multiple threads in OSC recovery (3812)
5989 - fix debug_size parameters (3864)
5990 - fix mds_postrecov to initialize import for llog ctxt (3121)
5991 - replace config semaphore with spinlock (3306)
5992 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5993 - don't allow enqueue to complete on a destroyed export (3822)
5994 - down write_lock before checking llog header bitmap (3825)
5995 - recover from lock replay timeout (3764)
5996 - up llog sem before sending rpc (3652)
5997 - reduce ns lock hold times when setting kms (3267)
5998 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5999 - fix NULL deref and obd_dev leak on setup error (3312)
6000 - replace some LBUG about llog ops with error handling (3841)
6001 - don't match INVALID dentries from d_lookup and spin (3784)
6002 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6003 - fix invalid assertion in ptlrpc_set_wait (3880)
6005 - add libwrap support for the TCP acceptor (3996)
6006 - add /proc/sys/portals/routes for non-root route listing (3994)
6007 - allow setting MDS UUID in .xml (2580)
6008 - print the stack of a process that LBUGs (4228)
6010 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6013 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6014 - make sure to unset replay flag from failed open requests (3440)
6015 - if default stripe count is 0, use OST count for inode size (3636)
6016 - update parent mtime/ctime on client for create/unlink (2611)
6017 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6018 - free recovery state on server during a forced cleanup (3571)
6019 - unregister_reply for resent reqs (3063)
6020 - loop back devices mounting and status check on 2.6 (3563)
6021 - fix resource-creation race that can provoke i_size == 0 (3513)
6022 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6023 - more debugging for page-accounting assertion (3746)
6024 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6025 - avoid deadlock after precreation failure (3758)
6026 - fix race and lock order deadlock in orphan handling (3450, 3750)
6027 - add validity checks when grabbing inodes from l_ast_data (3599)
6029 - add /proc/.../recovery_status to obdfilter (3428)
6030 - lightweight CDEBUG infrastructure, debug daemon (3668)
6031 - change default OSC RPC parameters to be better on small clusters
6032 - turn off OST read cache for files smaller than 32MB
6033 - install man pages and include them in rpms (3100)
6034 - add new init script for (un)mounting lustre filesystems (2593)
6035 - run chkconfig in %post for init scripts (3701)
6036 - drop scimac NAL (unmaintained)
6038 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6041 - clean kiobufs before and after use (3485)
6042 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6043 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6044 - change init script to fail more clearly if not run as root (1528)
6045 - allow clients to reconnect during replay (1742)
6046 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6047 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6048 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6049 - fix cond_resched (3554)
6050 - only evict unfinished clients after recovery (3515)
6051 - allow bulk resend, prevent data loss (3570)
6052 - dynamic ptlrpc request buffer allocation (2102)
6053 - don't allow unlinking open directory if it isn't empty (2904)
6054 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6055 - remove extraneous obd dereference causing LASSERT failure (3334)
6056 - don't use get_cycles() when creating temp. files on the mds (3156)
6057 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6058 - handle EEXIST for set-stripe, set proper directory name (3336)
6060 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6061 - fix ksocknal_fmb_callback() error messages (2918)
6063 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6066 - don't copy lvb into (possibly NULL) reply on error (2983)
6067 - don't deref dentry after dput, don't free lvb on error (2922)
6068 - use the kms to determine writeback rpc length (2947)
6069 - increment oti_logcookies when osc is inactive (2948)
6070 - update client's i_blocks count via lvb messages (2543)
6071 - handle intent open/close of special files properly (1557)
6072 - mount MDS with errors=remount-ro, like obdfilter (2009)
6073 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6074 - don't use cancelling-locks' kms values (2947)
6075 - use highest lock extent for kms, not last one (2925)
6076 - don't dereference ERR_PTR() dentry in error handling path (3107)
6077 - fix thread race in portals_debug_dumplog() (3122)
6078 - create lprocfs device entries at setup instead of at attach (1519)
6079 - common AST error handler, don't evict client on completion race (3145)
6080 - zero nameidata in detach_mnt in 2.6 (3118)
6081 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6082 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6083 - initialize RPC timeout timer earlier for 2.6 (3219)
6084 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6085 - print nal/nid for unknown nid (3258)
6086 - additional checks for oscc recovery before doing precreate (3284)
6087 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6088 - don't crash in mdc_close for bad permissions on open (3285)
6089 - zero i_rdev for non-device files (3147)
6090 - clear page->private before handing to FS, better assertion (3119)
6091 - tune the read pipeline (3236)
6092 - fix incorrect decref of invalidated dentry (2350)
6093 - provide read-ahead stats and refine rpc in flight stats (3328)
6094 - don't hold journal transaction open across create RPC (3313)
6095 - update atime on MDS at close time (3265)
6096 - close LDAP connection when recovering to avoid server load (3315)
6097 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6098 - don't leak open file on MDS after open resend (3325)
6099 - serialize filter_precreate and filter_destroy_precreated (3329)
6100 - loop device shouldn't call sync_dev() for nul device (3092)
6101 - clear page cache after eviction (2766)
6102 - resynchronize MDS->OST in background (2824)
6103 - refuse to mount the same filesystem twice on same mountpoint (3394)
6104 - allow llmount to create routes for mounting behind routers (3320)
6105 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6106 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6107 - fix rare problem with rename on htree directories (3417)
6109 - allow default OST striping configuration per directory (1414)
6110 - fix compilation for qswnal for 2.6 kernels (3125)
6111 - increase maximum number of MDS request buffers for large systems
6112 - change liblustreapi to be useful for external progs like lfsck (3098)
6113 - increase local configuration timeout for slow disks (3353)
6114 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6116 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6119 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6120 - don't overwrite extent policy data in reply if lock was blocked (2901)
6121 - drop filter export grants atomically with removal from device (2663)
6122 - del obd_self_export from work_list in class_disconnect_exports (2908)
6123 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6124 - swab reply message in mdc_close, other PPC fixes (2464)
6125 - fix destroying of named logs (2325)
6126 - overwrite old logs when running lconf --write_conf (2264)
6127 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6128 - fix race in target_handle_connect (2898)
6129 - mds_reint_create() should take same inode create lock (2926)
6130 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6131 - don't close files for self_export to avoid uninitialized obd (2936)
6132 - allow MDS with the same name as client node (2939)
6133 - hold dentry reference for closed log files for unlink (2325)
6134 - reserve space for all logs during transactions (2059)
6135 - don't evict page beyond end of stripe extent (2925)
6136 - don't oops on a deleted current working directory (2399)
6137 - handle hard links to targets without a parent properly (2517)
6138 - don't dereference NULL lock when racing during eviction (2867)
6139 - don't grow lock extents when lots of conflicting locks (2919)
6141 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6144 - account for cache space usage on clients to avoid data loss (974)
6145 - lfsck support in lustre kernel code (2349)
6146 - reduce journal credits needed for BRW writes (2370)
6147 - orphan handling to avoid losing space on client/server crashes
6148 - ptlrpcd can be blocked, stopping ALL progress (2477)
6149 - use lock value blocks to assist in proper KMS, faster stat (1021)
6150 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6151 - recovery for initial connections (2355)
6152 - fixes for mds_cleanup_orphans (1934)
6153 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6154 - block all file creations until orphan recovery completes (1901)
6155 - client remove rq_connection from request struct (2423)
6156 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6157 - recovery timer race (2670)
6158 - mdc_close recovey bug (2532)
6159 - ptlrpc cleanup bug (2710)
6160 - mds timeout on local locks (2588)
6161 - namespace lock held during RPCs (2431)
6162 - handle interrupted sync write properly (2503)
6163 - don't try to handle a message that hasn't been replied to (2699)
6164 - client assert failure during cleanup after abort recovery (2701)
6165 - leak mdc device after failed mount (2712)
6166 - ptlrpc_check_set allows timedout requests to complete (2714)
6167 - wait for inflight reqs when ptlrpcd finishes (2710)
6168 - make sure unregistered services are removed from the srv_list
6169 - reset bulk XID's when resending them (caught by 1138 test)
6170 - unregister_bulk after timeout
6171 - fix lconf error (2694)
6172 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6173 - readahead locks pages, leaves pending causing memory pressure (2673)
6174 - increase OST request buffers to 4096 on large machines (2729)
6175 - fix up permission of existing directories in simple_mkdir (2661)
6176 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6177 - don't assign transno to errored transactions (2742)
6178 - don't delete objects on OST if given a bogus objid from MDS (2751)
6179 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6180 - if rq_no_resend, then timeout request after recovery (2432)
6181 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6182 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6183 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6184 - MDS recovery shouldn't delete objects that it has given out (2730)
6185 - if enqueue arrives after completion, don't clobber LVB (2819)
6186 - don't unlock pages twice when trigger_group_io returns error (2814)
6187 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6188 - don't write pages to disk if there was an error (1450)
6189 - don't ping imports that have recovery disabled (2676)
6190 - take buffered bytes into account when balancing socknal conn (2817)
6191 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6192 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6193 - remove little-endian swabbing of llog records (1987)
6194 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6195 - retry reposting request buffers if they fail (1191)
6196 - grow extent at grant time to avoid granting a revoked lock (2809)
6197 - lock revoke doesn't evict page if covered by a second lock (2765)
6198 - disable VM readahead to avoid reading outside lock extents (2805)
6200 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6201 - updated kernel patches for hp-2.4.20 kernel (2681)
6203 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6206 - fix truncated write corruption (2366)
6207 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6209 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6210 - ensure len(uuid) < 37 in lmc (1171)
6211 - fix ia64 OOPS in llog_test (2255)
6212 - zero end of page at obdfilter for partial page writes (2648)
6213 - don't leave stale dentries around after renames (bug 2428)
6214 - fix timeouts when evicting a client with a single lock held (2642)
6215 - set deadline for the initial HELLO message to drain (2634)
6216 - print out dotted-quad IP addresses in the socknal (2302)
6218 - additional debugging for MDS client eviction problem (2443)
6219 - fix mkfsoptions support for osts (2603, 2604)
6221 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6224 - add series for the vanilla 2.6.0 kernel
6225 - add series for the vanilla 2.4.24 kernel
6226 - add series for a cray x86/64 UL kernel drop
6227 - fix xattr patches for the vanilla 2.4.19 series
6229 - generate true UUIDs in lmc (1171)
6230 - have portals stack dumping break in UML (2466)
6231 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6232 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6233 - wake callback waiting threads on client eviction (2460)
6234 - Add --ptldebug and --subsystem to lmc (1719)
6235 - update assertion to allow safe interrupt allocation
6236 - set rq_no_resend for cancel requests (2432)
6237 - recalculate ptlrpcd timeout after resend (2494)
6238 - call vfs_rmdir when removing pending directories (2368)
6239 - fix renaming a file to itself (2429)
6240 - lmc creates a default one-stripe lov (2454)
6241 - expand procfs space to handle large clusters (2326)
6242 - increase UML stack to avoid overflow
6243 - update lconf's list of debug and subsystem masks
6244 - fix lfs find --obd (2510)
6245 - /proc tunable for disabling filter read caching (2591)
6246 - stop rpm packages from altering slapd.conf (2301)
6247 - disable nagle in the socknal under 0conf (2578)
6248 - choose mds inode size based on stripe count (2572)
6249 - fix kernel-source rpm problems (2516)
6251 - add --disable-doc to avoid pdf generation (2421)
6252 - update documentation, tests, type-os, comments
6253 - avoid format warnings on ia64
6254 - remove the TOE NAL
6255 - tiny code cleanups by removing unused fields
6257 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6260 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6261 - remove the most blatant lies from BUILDING file (2371)
6262 - change default debug level to reasonable production setting
6263 - reduce client side cache size to reduce cache flush time
6264 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6265 - make TCP zerocopy and pinger support enabled by default (2476)
6266 - sync writes completed after process exits caused crashes (2319)
6267 - maintain correct mount count on the MDS (2356)
6268 - backout 1557, because 2316 wasn't really fixed
6269 - better file I/O statistics gathering in /proc
6270 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6271 - another kernel patch to fix zero-copy TCP function export
6272 - don't take duplicate lock when processing re-sent getattr (2420)
6273 - lctl uses obd_self_export instead of creating new conn (2353)
6274 - MDS/OST recovery case which requires object creation asserted (2425)
6275 - move lfs from /usr/sbin to /usr/bin in packages
6276 - fix race between mds_client_add and mds_client_free (2417)
6277 - use kmalloc instead of slabs in portals (2430)
6278 - don't create duplicate records when a failover MDS is present (2442)
6279 - remove unnecessary mount age check (2332)
6280 - don't remove directory inodes from locks prematurely (2451)
6281 - don't break if MDS service name is the same as hostname (2103)
6282 - fix races in client write RPC generation when cache full (2482)
6284 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6287 - remove now-unused request->rq_obd (278)
6288 - if an allocation fails, print out how much memory we've used (1933)
6289 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6290 - add the "configurable stack size" patch to most series files (1256)
6291 - ability to write large log records, for 100+ OST configs (2306)
6292 - fix NULL deref when filter_prep fails (2314)
6293 - fix operator precedence error in filter_sync
6294 - dynamic allocation of socknal TX descriptors (2315)
6295 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6296 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6297 - ensure that CWARN messages in Portals always get to the syslog
6298 - __init/__exit are not for prototype decls (ldlm_init/exit)
6299 - x86-64 compile warning fixes
6300 - fix gateway LMC keyword conflict (2318)
6301 - fix MDS lock inversions in getattr/reint paths (1844)
6302 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6303 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6304 - DLM race condition prevented some lock evictions (2328)
6305 - ENOMEM detection and retry on socknal sends (2230)
6306 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6307 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6308 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6309 - better error messages when a client is rejected during recovery (1505)
6310 - avoid cancelling locks which were never granted, after failure (2330)
6311 - fix i_sem/journal inversion in mds_client_add (2333)
6312 - fix truncate/getattr lock cycle deadlock (2334)
6313 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6314 - fix two rare exit paths which could leak an l_lock() ref (2321)
6315 - fencepost error in MDS/OST orphan recovery (2226)
6316 - make log record alignment 8 bytes (1988)
6317 - lstripe now fails when requested offset > ost_count (2237)
6318 - ensure that all kernel series have a complete list.h (1607)
6319 - fix crashes in special-file operations (2316)
6320 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6322 - allow configurable automake binary, for testing new versions
6323 - small update to the lfs documentation
6325 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6327 * fix negative export reference count in fsfilt_sync (2312)
6329 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6330 * release candidate 0.9.1
6332 - orphans are moved into the PENDING directory for possible recovery
6333 - replayed opens now open by fid for orphan/rename safety (1042)
6334 - last close of an orphan inode generates a transno (683)
6335 - chdir() and mount() now pin the directory entry (1020)
6336 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6337 - discard very old requests without processing them (1502)
6338 - remove some common, well-understood CERRORs (1505)
6339 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6340 - clear "grant" flags in OST replies until OST grant code lands (1644)
6341 - fix read performance by not clobbering i_blksize on client (1598)
6342 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6343 - make LRU size a /proc tunable, clears locks when reduced (707)
6344 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6345 - server locks take references on exports now (1558)
6346 - build fixes for 2.4.20-rh trees (1663)
6347 - return an error from lov_create if all OSCs are inactive (1751)
6348 - fix import levels when a reconnect happens without a timeout (1597)
6349 - exit early from mds_open if we get a lookup error (1749)
6350 - partial page read at EOF wouldn't wait for disk before sending (1642)
6351 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6352 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6353 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6354 - write/truncate lock inversion (1639)
6355 - Don't auto-load obdclass, portals modules during cleanup (1495)
6356 - fix timestamps from jumping to "now" (1763)
6357 - extra journal assertions (1648)
6358 - add an extra multiunlink test (1771)
6359 - fix read_record/write_record API (1776)
6360 - fix leak of offset_extent, possible incorrect i_size later (1772)
6361 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6362 - Debugging for the kqswnal_get_idle_tx problems (1820)
6363 - Allow recovery to be attempted multiple times (1536)
6364 - Write out MDS last_rcvd file after it is first created (1600)
6365 - Fix tx_descriptor leak in failed transmit situations (1827)
6366 - ext3 journaling fixes for assertion failure after IO error (1871)
6367 - class_export_put() on freed export after completion AST error (1896)
6368 - Fix revalidate looping in VFS (1322)
6369 - Don't access a freed export during MDS_REINT timeout (1521)
6370 - Add open-unlink recovery support on the MDS (1673,1764)
6371 - Return an error if no MDS data was read from last_rcvd (1946)
6372 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6373 - Don't setup a disk device that doesn't match exported UUID (317)
6374 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6375 - avoid committing NULL handle in force close
6376 - local.sh is now a one-stripe LOV configuration
6377 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6378 - fix ext3 htree duplicate directory entry corruption (1516)
6379 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6380 - update to new LOV EA format (2097)
6381 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6383 2003-06-15 Phil Schwan <phil@clusterfs.com>
6386 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6387 - per-import/export recovery handling (958, 931, 959)
6388 - multiple last-rcvd slots, for serving multiple FSes (949)
6389 - connections are again shared between multiple imp/exports (963, 964)
6390 - "umount -f" would hang if any requests needed to be sent (393, 978)
6391 - avoid pinning large req buffer by copying for queued messages (989)
6392 - add "uuid" to "lctl device" command to help upcalls (991)
6393 - "open" RPCs with transnos would confuse recovery counters (1037)
6394 - do proper endian conversion of all wire messages (288, 340, 891)
6395 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6396 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6397 - fix ptlrpc_connection leak in target_handle_connect (1174)
6398 - fix import refcounting bug in OST and MDS cleanup (1134)
6399 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6400 - fix very unlikely obd_types race condition (501)
6401 - remove osc_open hack for echo_client (1187)
6402 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6403 - a failure in read_inode2 leads to deadlock (1139)
6404 - cancel ack-locks as soon as transaction is committed (1072)
6405 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6406 - make sure to commitrw after any preprw to avoid deadlock (1162)
6407 - failing to execute a file in a lustre FS would lock inode (1203)
6408 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6409 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6410 - obd_uuid2tgt would walk off the end of the list (1255)
6411 - on IA64 the osc would give portals incorrect bulk size (1258)
6412 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6413 - fix lock inversion caused by new llite matching code (1282)
6414 - limit the number of dirty pages on a client to 10MB (1286)
6415 - timed out locks were not being corrected cancelled (1289)
6416 - fix O_DIRECT above 4GB on IA-32 (1292)
6417 * major user-visible changes
6418 - fail out/fail over policy now controlled by the upcall (993)
6420 - add OBD_PING to check server availability and failure (954)
6421 - lustre messages are now sent in sending host order (288, 340, 891)
6422 - add eadatalen to MDS getattr reply (340)
6423 - OST read replies may contain second buffer, with per-page status (593)
6425 2003-03-11 Phil Schwan <phil@clusterfs.com>
6428 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6429 - failed open()s could cause deadlock; fixed (867, 869)
6430 - stop cancelling OST locks when files are closed (481)
6431 - overlapping XID spaces caused network corruption (851, 853)
6432 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6433 - setattr_raw added, to do single-RPC, server-side setattrs
6434 - lmc/lconf syntax change for OST UUIDs
6435 - fix crashy race condition between ptlrpc_free_req and osc_close
6436 - don't use request in mdc_enqueue if we hit a timeout (889)
6437 - don't set the inode i_size for regular files from the MDS (896)
6438 - handle out of order completion AST (842)
6439 - don't LBUG if a lock request times out after receiving AST (913)
6440 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6441 - if a bad lock AST arrives, send an error instead of dropping entirely
6442 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6443 - fix leak in bulk IO when only partially completed (899, 900, 926)
6444 - fix O_DIRECT for ia64 (55)
6445 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6446 - C-z after timeout could hang a process forever; fixed (977)
6448 - client-side I/O cache (678, 924, 929, 941, 970)
6450 - READPAGE and SETATTRs which don't take server-side locks get
6453 2003-02-11 Phil Schwan <phil@clusterfs.com>
6456 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6457 - Fix fsfilt_extN_readpage() to read a full page of directory
6458 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6459 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6460 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6461 - Most intent operations were removed, in favour of a new RPC mode
6462 that does a single RPC to the server and bypasses most of the VFS
6463 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6464 - Aggressively cancel local locks on DLM servers
6465 - mds_reint_unlink sends EA to the client if it's the last nlink.
6466 client uses that EA to unlink OST objects.
6467 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6468 - recursive symlinks were fixed (439)
6469 - fixed NULL deref in DEBUG_REQ
6470 - filter_update_lastobjid no longer calls sync, which annoyed extN
6471 - fixed multi-client small-writes to a single file problem (445)
6472 - fixed mtime updates during file writes (607)
6473 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6474 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6475 - fixed Linux OST size reporting problem (444, 656)
6476 - OST now updates object mtime with writes or setattr (607, 619)
6477 - client verifies file size before zeroing page past EOF (445)
6478 - OST now writes last allocated objid to disk with allocation (108)
6479 - LOV on echo now works (409)
6481 - mds_reint_unlink sends a new buffer, with the EA included. this
6482 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6483 set if a regular file was being unlinked, and it was the last link
6484 - use PtlGet from the target for bulk writes (315)
6485 - OST now updates object mtime with writes or setattr (607, 619)
6486 - LDLM now has a grant-time callback to revalidate locked items, if
6488 - Many MDS operations were reorganized to combat race conditions
6490 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6491 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6493 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6496 - Fully reactivate OST imports after reconnection (512, others)
6497 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6498 - More graceful error handling for truncating on dead OST (515)
6499 - Don't error out unless we're actually accessing dead stripes (474)
6500 - Fix garbage sizes when stripes are missing (410)
6501 - LRU counters were broken, causing constant lock purge (433, 432)
6502 - garbage on read from stripes with failed OSTs (441)
6503 - mark OSCs as active before reconnecting during recovery (438)
6504 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6505 - lfind did not preserve OST order in output (443)
6506 - symlinks cause hung clients, incorrect data (439)
6507 - stop dereferencing request after dropping refcount (457)
6508 - don't LASSERT(spin_is_locked) on non-SMP (455)
6509 - fixes for many rename() bugs
6510 - fstat didn't correctly synchronize attributes (399)
6511 - server must handle lock cancellation during blocking AST prep (487)
6512 - bulk descriptors were free()d too soon (511)
6513 - fix paths in lconf, which would load incorrect modules (451, 507)
6514 - fix confusing lconf 'host not found' error message (386)
6515 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6516 - fix race condition in mdc_blocking_ast() for inode access (526)
6517 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6518 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6519 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6520 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6521 - fix lstripe in several ways (kernel side) (527)
6522 - fix request leak in ldlm_cli_enqueue (262)
6523 - incorrect OSC was marked inactive after OST failure
6524 - call mds_fs_cleanup before unmounting filesystem (524)
6525 - fix races between taking ns_lock and ldlm_lock_change_resource
6526 - fix races updating LOV export open file list
6527 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6528 - fix recovery NULL deref in ldlm_cli_cancel_unused
6529 - fix some DLM races by using new hash table for lock handles (419)
6530 - permit the client to specify desired inodes, at replay
6531 - duplicate requests when we queue them for replay reintegration
6532 - fix last_rcvd offset calculation
6533 - sync after each recovered transaction, so we always make progress
6534 - never, not always, ERESTART requests without transnos
6535 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6537 - skip replay if the MDS says that the client is already connected
6538 - don't check for a recovery-enabled export to match lctl's UUID
6539 - don't INC_USE_COUNT for phantom exports
6540 - don't crash when cleaning up phantom exports (567)
6541 - don't double-finish or set replay data for errored mdc_open requests
6542 - abort requests when they time out, so we don't get old replies
6543 - send/receive replies for AST messages again
6544 - if the client says that it doesn't have the lock, cancel it on the
6546 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6547 mark it as destroyed, it will all work out when decref is called
6548 - fix module use counts (22, 581)
6550 - ASTs now expect a reply (server cancels lock on error reply)
6552 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6555 - fix many simultaneous client startup (392)
6556 - fix dentry->d_it clobbering
6557 - credentials weren't being shipped for readdir/getattr operations
6558 - remove invalid assertions triggered during some concurrent MD
6560 - proper Lustre versions added (336, 389)
6561 - fix memory leak for create error case (398)
6562 - fix LOV locking bug that would get cli/srv out of sync
6563 - fix echo client over LOV (409)
6564 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6565 - fix double-O_EXCL intent crash (424)
6566 - avoid sending multiple lock CANCELs (352)
6568 - MDS can do multi-client recovery (modulo bugs in new code)
6570 - many updates, edits, cleanups
6572 2002-11-18 Phil Schwan <phil@clusterfs.com>
6575 - fix null d_it dereference (346)
6576 - fix full OST/dbench hang (333)
6577 - fix permission problem with file removal (286)
6578 - fix removal of OSCs from LOV when they fail
6579 - fix NULL deref during bulk timeout (214)
6580 - fix problems related to multiple filesystems on one MDS (241)
6581 - fixed serious subtle metadata locking bugs
6582 - free locks on clients when inodes are removed due to memory
6584 - fix inode pointer in lock data (285)
6585 - partial support for multiple MDS on a single host (241)
6586 - data locks weren't cancelled at clear_inode time (290, 311)
6587 - intent locks could lead to unbounded lock growth (205)
6588 - added a maximum lock count, an LRU list, and a flusher
6589 - fix multiple rename (365)
6590 - properly abstracted the echo client
6591 - OSC locked 1 byte too many; fixed
6592 - rewrote brw callback code:
6593 - fixed recovery bugs related to LOVs (306)
6594 - fixed too-many-pages-in-one-write crash (191)
6595 - fixed (again) crash in sync_io_timeout (214)
6596 - probably fixed callback-related race (385)
6598 - Add capability to MDS protocol
6599 - LDLM cancellations and callbacks on different portals
6601 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6604 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6605 - defer open object destruction to close time (601981, 138)
6606 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6607 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6608 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6609 - don't call obd_setattr in ll_file_release for destroyed objects
6610 * protocol change to lustre_msg: move |version| and add |flags|
6611 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6612 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6613 * added replay of create, unlink, link and rename operations during
6614 MDS failover; recovery should be much more robust now
6615 * remove failed OSCs from LOVs (only lov_create uses this so far)
6616 * the lustre-HOWTO was brought (more) up to date (582544)
6618 2002-10-23 Phil Schwan <phil@clusterfs.com>
6621 - in-use dentries weren't being reused properly (617851)
6622 - prevent multiple LDLM setup (599178)
6623 - fix LOV size calculations for truncate (617853)
6624 - fix client handling of MDS intent errors (POSIX)
6625 - fix permission bug in lovstripe.c test (624321)
6626 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6627 - truncate past end of file could corrupt data
6628 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6629 - a race in recovery could return ETIMEDOUT to apps (623947)
6630 - building outside the source directory was fixed
6631 * the lustre-HOWTO was brought (more) up to date (582544)
6632 * major progress was made on recovery functionality
6634 2002-10-10 Phil Schwan <phil@clusterfs.com>
6637 - recovery deadlock fix
6638 - rm -rf causes LBUG fix (617817)
6639 - file open by multiple tasks fix (618962)
6640 - directory permissions bugs (602707 and 620007)
6641 - journal_stop fixed with locking (611313)
6642 - O_APPEND failures resolved (618273, perhaps 614459)
6643 - lconf PATH fix (619770)
6644 - IA64 build fix (621450)
6645 - RPC buffer sizes scale with amount of memory
6647 2002-10-01 Phil Schwan <phil@clusterfs.com>
6650 - locks would be cancelled without throwing away data pages,
6651 resulting in inconsistent data (605627)
6652 - inode attributes were not always being refreshed (605627, 612449)
6653 - lconf now continues to cleanup after lctl reports an error
6654 - MDS now enforces user permissions (602707)
6655 - lprocfs cleanup fixed, but not yet enabled (614157)
6656 - fixed infinite server hang, should a client not respond to an AST
6657 - avoid going into recovery if user calls readlink() with a buffer
6658 that's too small (613941)
6659 - AST RPCs no longer require replies (614867) -- this may be changed
6660 - don't crash server if client sends an IOV that's too big (611336)
6661 - fixed lock conversion deadlock (611892)
6662 - fixed the following of symlinks (614622)
6663 * recovery: the server can remove locks from a client that dies, other
6664 clients can make progress
6665 * more extN patch fixes
6666 * compile-time configurable ptlrpc buffer allocations
6668 - collaborative read cache document
6669 - Lustre Lite Performance CDR document-in-progress
6671 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6674 - fix typo in patch-2.4.18
6676 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6679 - clear ptlrpc request each time in handle_incoming_request()
6680 - unlink of files now destroys the object on the OST
6682 2002-09-19 Peter Braam <braam@clusterfs.com>
6684 * add hard link support
6685 * change obdfile creation method
6686 * kernel patch changed
6688 2002-09-19 Peter Braam <braam@clusterfs.com>
6691 - stack overflow bug in extN fixed
6693 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6695 * documentation updates
6696 - add man pages for config tools
6697 - update tests/README to describe testing with new config tools
6698 - finish metadata API descriptions
6699 * bug fixes and cleanups
6700 - statfs workaround for 16TB limit
6701 - LOV stripe allocation improved, can stripe on subset of OSTs
6702 - LOV file size/IO offset was wrong for files > 4GB in size
6703 - object EA data was being dropped, caused files to be unreadable
6704 - memory overflow with non-LOV OST caused memory corruption
6705 - fixed regression tests to work with new config tools, obdfilter
6706 - fixed bug when directory size became larger than 1 block
6707 - fixed bug (for single client case) when PWD was deleted
6708 - invalidate local directory pages when doing intent-based ops
6709 - avoid LDLM oops when lock callback contained bad data
6711 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6713 * documentation updates
6714 * bug fixes and cleanups
6715 - configuration tools
6718 - 64-bit compile warnings
6719 - 64-bit internal statfs data
6721 * test_brw on persistent OST devices
6723 * lprocfs (disabled)
6725 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6727 * documentation updates
6728 * bug fixes and cleanups
6729 * configuration tools
6731 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6734 * many small fixes to 0_5_4
6735 * io/network handling
6736 * thinkos in MDS operations
6738 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6741 * crucial basic fixes to 0.5.3
6742 * IOR, Iozone work over Elan
6745 2002-08-07 Phil Schwan <phil@clusterfs.com>
6746 * version 0_5_3, our first alpha
6747 * we use the new Portals iovs
6748 * documentation updates
6749 * bug fixes and cleanups
6750 * small changes in the DLM wire protocol
6752 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6753 * version 0_5_1 with some initial stability,
6754 * locking on MD and file I/O.
6755 * documentation updates
6756 * several bug fixes since 0.5.0
6757 * small changes in wire protocol
6759 2002-07-18 Phil Schwan <phil@clusterfs.com>
6761 * delivered as Lustre Light Alpha
6762 * fixed a crash after handling invalid MDS requests
6763 * fixed directory pages for architectures with non-4k pages sizes
6765 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6766 * release version v0_4_4
6767 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6769 * Fixes a number of interruption problems with OST operations.
6770 * Update documentation for portals header changes
6771 * Move all wire protocol structs/defines to lustre_idl.h
6772 * Fixes symlink length bug.
6773 * Add tcpdump to repository.
6775 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6776 * release version v0_4_3
6777 * Fixes statfs for inodes on extN.
6778 * Fixes bug in runtests which would delete /etc/hosts.
6779 * Use 64-bit object IDs wherever possible (not into VFS though)
6780 Remove ost_get_info, which is unused by lustre, and out of date.
6782 2002-07-03 Peter Braam <braam@clusterfs.com>
6783 * release version v0_4_2 Fixes a lookup error (type not passed)
6784 * move forward to head of Portals
6785 * move forward to latest Lustre kernel
6787 2002-06-25 Peter Braam <braam@clusterfs.com>
6788 * release version v0_4_1. Hopefully stable on single node use.