3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-957.27.2.el7 (RHEL7.6)
7 3.10.0-1062.1.1.el7 (RHEL7.7)
8 * Server known to build on the following kernels (others may also work):
9 2.6.32-696.18.7.el6 (RHEL6.9)
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 4.4.120-92.70 (SLES12 SP2)
12 4.4.180-94.97 (SLES12 SP3)
13 4.4.0-142 (Ubuntu 16.04)
14 4.15.0-45 (Ubuntu 18.04)
15 vanilla linux 4.6.7 (ZFS only)
16 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
17 * Client primary unpatched kernels built and tested during release cycle:
18 3.10.0-957.27.2.el7 (RHEL7.6)
19 3.10.0-1062.1.1.el7 (RHEL7.7)
20 4.18.0-80.7.1.el8 (RHEL8.0)
21 4.4.180-94.97 (SLES12 SP3)
22 4.12.14-95.29 (SLES12 SP4)
23 4.15.0-45 (Ubuntu 18.04)
24 * Client known to build on unpatched kernels (others may also work):
25 2.6.32-696.18.7.el6 (RHEL6.9)
26 2.6.32-754.6.3.el6 (RHEL6.10)
27 3.10.0-862.14.4.el7 (RHEL7.5)
28 4.14.0-49.13.1.el7a (RHEL7.5)
29 4.14.0-115.2.2.el7a (RHEL7.6)
30 4.4.120-92.70 (SLES12 SP2)
31 4.4.0-142 (Ubuntu 16.04)
32 vanilla linux 4.15.0 (ZFS only)
33 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
34 * Recommended ZFS / SPL version: 0.7.13
35 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
36 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
37 stack overflow. For more information, please refer to bugzilla 17630.
38 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
39 the re-exporting nodes to be mounted with "32bitapi" mount option
40 * Note for Ubuntu 14.04.5:
41 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
42 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
46 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
47 for currently supported client and server kernel versions.
48 * Server known to build on the following kernels (others may also work):
49 2.6.32-696.18.7.el6 (RHEL6.9)
50 3.10.0-862.14.4.el7 (RHEL7.5)
51 3.10.0-957.10.1.el7 (RHEL7.6)
52 4.4.120-92.70 (SLES12 SP2)
53 4.4.162-94.69 (SLES12 SP3)
54 4.4.0-142 (Ubuntu 16.04)
55 4.15.0-45 (Ubuntu 18.04)
56 vanilla linux 4.6.7 (ZFS only)
57 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
58 * Client known to build on unpatched kernels (others may also work):
59 2.6.32-696.18.7.el6 (RHEL6.9)
60 2.6.32-754.6.3.el6 (RHEL6.10)
61 3.10.0-862.14.4.el7 (RHEL7.5)
62 3.10.0-957.10.1.el7 (RHEL7.6)
63 4.14.0-49.13.1.el7a (RHEL7.5)
64 4.14.0-115.2.2.el7a (RHEL7.6)
65 4.4.120-92.70 (SLES12 SP2)
66 4.4.162-94.69 (SLES12 SP3)
67 4.12.14-95.13 (SLES12 SP4)
68 4.4.0-142 (Ubuntu 16.04)
69 4.15.0-45 (Ubuntu 18.04)
70 vanilla linux 4.15.0 (ZFS only)
71 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
72 * Recommended ZFS / SPL version: 0.7.13
73 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
74 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
75 stack overflow. For more information, please refer to bugzilla 17630.
76 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
77 the re-exporting nodes to be mounted with "32bitapi" mount option
78 * Note for Ubuntu 14.04.5:
79 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
80 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
85 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
86 for currently supported client and server kernel versions.
87 * Server known to build on the following kernels (others may also work):
88 2.6.32-696.18.7.el6 (RHEL6.9)
89 3.10.0-862.14.4.el7 (RHEL7.5)
90 3.10.0-957.10.1.el7 (RHEL7.6)
91 4.4.120-92.70 (SLES12 SP2)
92 4.4.162-94.69 (SLES12 SP3)
93 4.4.0-142 (Ubuntu 16.04)
94 4.15.0-45 (Ubuntu 18.04)
95 vanilla linux 4.6.7 (ZFS only)
96 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
97 * Client known to build on unpatched kernels (others may also work):
98 2.6.32-696.18.7.el6 (RHEL6.9)
99 2.6.32-754.6.3.el6 (RHEL6.10)
100 3.10.0-862.14.4.el7 (RHEL7.5)
101 3.10.0-957.10.1.el7 (RHEL7.6)
102 4.14.0-49.13.1.el7a (RHEL7.5)
103 4.14.0-115.2.2.el7a (RHEL7.6)
104 4.4.120-92.70 (SLES12 SP2)
105 4.4.162-94.69 (SLES12 SP3)
106 4.12.14-95.13 (SLES12 SP4)
107 4.4.0-142 (Ubuntu 16.04)
108 4.15.0-45 (Ubuntu 18.04)
109 vanilla linux 4.15.0 (ZFS only)
110 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
111 * Recommended ZFS / SPL version: 0.7.13
112 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
113 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
114 stack overflow. For more information, please refer to bugzilla 17630.
115 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
116 the re-exporting nodes to be mounted with "32bitapi" mount option
117 * Note for Ubuntu 14.04.5:
118 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
119 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
123 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
124 for currently supported client and server kernel versions.
125 * Server known to build on the following kernels (others may also work):
126 2.6.32-696.18.7.el6 (RHEL6.9)
127 3.10.0-862.14.4.el7 (RHEL7.5)
128 3.10.0-957.el7 (RHEL7.6)
129 4.4.120-92.70 (SLES12 SP2)
130 4.4.162-94.69 (SLES12 SP3)
131 4.4.0-131 (Ubuntu 16.04)
132 4.15.0-32 (Ubuntu 18.04)
133 vanilla linux 4.6.7 (ZFS only)
134 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
135 * Client known to build on unpatched kernels (others may also work):
136 2.6.32-696.18.7.el6 (RHEL6.9)
137 2.6.32-754.6.3.el6 (RHEL6.10)
138 3.10.0-862.14.4.el7 (RHEL7.5)
139 3.10.0-957.el7 (RHEL7.6)
140 4.14.0-49.13.1.el7a (RHEL7.5)
141 4.14.0-115.2.2.el7a (RHEL7.6)
142 4.4.120-92.70 (SLES12 SP2)
143 4.4.162-94.69 (SLES12 SP3)
144 4.4.0-131 (Ubuntu 16.04)
145 4.15.0-32 (Ubuntu 18.04)
146 vanilla linux 4.15.0 (ZFS only)
147 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
148 * Recommended ZFS / SPL version: 0.7.9
149 * Tested with ZFS / SPL version: 0.7.9
150 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
151 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
152 stack overflow. For more information, please refer to bugzilla 17630.
153 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
154 the re-exporting nodes to be mounted with "32bitapi" mount option
155 * Note for Ubuntu 14.04.5:
156 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
157 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
159 --------------------------------------------------------------------------------
162 2018-04-03 Intel Corporation
164 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
165 for currently supported client and server kernel versions.
166 * Server known to build on patched kernels:
167 2.6.32-431.29.2.el6 (RHEL6.5)
168 2.6.32-504.30.3.el6 (RHEL6.6)
169 2.6.32-573.26.1.el6 (RHEL6.7)
170 2.6.32-642.15.1.el6 (RHEL6.8)
171 2.6.32-696.18.7.el6 (RHEL6.9)
172 3.10.0-693.21.1.el7 (RHEL7.4)
173 3.0.101-0.47.71 (SLES11 SP3)
174 3.0.101-107 (SLES11 SP4)
175 3.12.74-60.64.40 (SLES12 SP1)
176 4.4.114-92.64 (SLES12 SP2)
177 4.4.114-94.11 (SLES12 SP3)
178 3.13.0-101 (Ubuntu 14.04, ZFS only)
179 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
180 4.4.0-116 (Ubuntu 16.04)
181 vanilla linux 4.6.7 (ZFS only)
182 * Client known to build on unpatched kernels:
183 2.6.32-431.29.2.el6 (RHEL6.5)
184 2.6.32-504.30.3.el6 (RHEL6.6)
185 2.6.32-573.26.1.el6 (RHEL6.7)
186 2.6.32-642.15.1.el6 (RHEL6.8)
187 2.6.32-696.18.7.el6 (RHEL6.9)
188 3.10.0-693.21.1.el7 (RHEL7.4)
189 3.0.101-0.47.71 (SLES11 SP3)
190 3.0.101-107 (SLES11 SP4)
191 3.12.74-60.64.40 (SLES12 SP1)
192 4.4.114-92.64 (SLES12 SP2)
193 4.4.114-94.11 (SLES12 SP3)
194 3.13.0-101 (Ubuntu 14.04)
195 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
196 4.4.0-116 (Ubuntu 16.04)
198 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
199 * Recommended ZFS / SPL version: 0.7.6
200 * Tested with ZFS / SPL version: 0.7.6
201 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
202 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
203 stack overflow. For more information, please refer to bugzilla 17630.
204 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
205 the re-exporting nodes to be mounted with "32bitapi" mount option
206 * Note for Ubuntu 14.04.5:
207 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
208 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
210 --------------------------------------------------------------------------------
212 2017-07-05 Intel Corporation
214 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
215 for currently supported client and server kernel versions.
216 * Server known to build on patched kernels:
217 2.6.32-431.29.2.el6 (RHEL6.5)
218 2.6.32-504.30.3.el6 (RHEL6.6)
219 2.6.32-573.26.1.el6 (RHEL6.7)
220 2.6.32-642.15.1.el6 (RHEL6.8)
221 2.6.32-696.3.1.el6 (RHEL6.9)
222 3.10.0-514.21.1.el7 (RHEL7.3)
223 3.0.101-0.47.71 (SLES11 SP3)
224 3.0.101-100 (SLES11 SP4)
225 3.12.74-60.64.40 (SLES12 SP1)
226 4.4.59-92.17 (SLES12 SP2)
227 vanilla linux 4.6.7 (ZFS only)
228 * Client known to build on unpatched kernels:
229 2.6.32-431.29.2.el6 (RHEL6.5)
230 2.6.32-504.30.3.el6 (RHEL6.6)
231 2.6.32-573.26.1.el6 (RHEL6.7)
232 2.6.32-642.15.1.el6 (RHEL6.8)
233 2.6.32-696.3.1.el6 (RHEL6.9)
234 3.10.0-514.21.1.el7 (RHEL7.3)
235 3.0.101-0.47.71 (SLES11 SP3)
236 3.0.101-100 (SLES11 SP4)
237 3.12.74-60.64.40 (SLES12 SP1)
238 4.4.59-92.17 (SLES12 SP2)
240 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
241 * Recommended ZFS / SPL version: 0.6.5.9
242 * Tested with ZFS / SPL version: 0.6.5.9
243 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
244 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
245 stack overflow. For more information, please refer to bugzilla 17630.
246 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
247 the re-exporting nodes to be mounted with "32bitapi" mount option
249 --------------------------------------------------------------------------------
251 12-07-2016 Intel Corporation
253 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
254 for currently supported client and server kernel versions.
255 * Server known to build on patched kernels:
256 2.6.32-431.29.2.el6 (RHEL6.5)
257 2.6.32-504.30.3.el6 (RHEL6.6)
258 2.6.32-573.26.1.el6 (RHEL6.7)
259 2.6.32-642.11.1.el6 (RHEL6.8)
260 3.10.0-514.el7 (RHEL7.3)
261 3.0.101-0.47.71 (SLES11 SP3)
262 3.0.101-80 (SLES11 SP4)
263 3.12.62-60.64.8 (SLES12 SP1)
264 vanilla linux 4.5.7 (ZFS only)
265 * Client known to build on unpatched kernels:
266 2.6.32-431.29.2.el6 (RHEL6.5)
267 2.6.32-504.30.3.el6 (RHEL6.6)
268 2.6.32-573.26.1.el6 (RHEL6.7)
269 2.6.32-642.11.1.el6 (RHEL6.8)
270 3.10.0-514.el7 (RHEL7.3)
271 3.0.101-0.47.71 (SLES11 SP3)
272 3.0.101-80 (SLES11 SP4)
273 3.12.62-60.64.8 (SLES12 SP1)
275 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
276 * Recommended ZFS / SPL version: 0.6.5.8
277 * Tested with ZFS / SPL version: 0.6.5.8
278 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
279 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
280 stack overflow. For more information, please refer to bugzilla 17630.
281 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
282 the re-exporting nodes to be mounted with "32bitapi" mount option
284 --------------------------------------------------------------------------------
286 02-29-2016 Intel Corporation
288 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
289 for currently supported client and server kernel versions.
290 * Server known to build on patched kernels:
291 2.6.32-431.29.2.el6 (RHEL6.5)
292 2.6.32-504.30.3.el6 (RHEL6.6)
293 2.6.32-573.12.1.el6 (RHEL6.7)
294 3.10.0-327.3.1.el7 (RHEL7.2)
295 3.0.101-0.47.71 (SLES11 SP3)
296 3.0.101-68 (SLES11 SP4)
297 vanilla linux 4.2.1 (ZFS only)
298 * Client known to build on unpatched kernels:
299 2.6.32-431.29.2.el6 (RHEL6.5)
300 2.6.32-504.30.3.el6 (RHEL6.6)
301 2.6.32-573.12.1.el6 (RHEL6.7)
302 3.10.0-327.3.1.el7 (RHEL7.2)
303 3.0.101-0.47.71 (SLES11 SP3)
304 3.0.101-68 (SLES11 SP4)
307 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
308 * Recommended ZFS / SPL version: 0.6.4.2
309 * Tested with ZFS / SPL version: 0.6.4.2
310 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
311 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
312 stack overflow. For more information, please refer to bugzilla 17630.
313 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
314 the re-exporting nodes to be mounted with "32bitapi" mount option
316 --------------------------------------------------------------------------------
318 03-10-2015 Intel Corporation
320 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
321 for currently supported client and server kernel versions.
322 * Server known to build on patched kernels:
323 2.6.32-431.29.2.el6 (RHEL6.5)
324 2.6.32-504.8.1.el6 (RHEL6.6)
325 3.0.101-0.46 (SLES11 SP3)
326 * Client known to build on unpatched kernels:
327 2.6.32-431.29.2.el6 (RHEL6.5)
328 2.6.32-504.8.1.el6 (RHEL6.6)
329 3.10.0-123.20.1.el7 (RHEL7)
330 3.0.101-0.46 (SLES11 SP3)
331 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
332 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
333 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
334 stack overflow. For more information, please refer to bugzilla 17630.
335 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
336 the re-exporting nodes to be mounted with "32bitapi" mount option
338 Severity : enhancement
340 Description: control OST-index in IDIF via ROCOMPAT flag.
341 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
342 last_rcvd file. For new formatted OST device, it will be auto set;
343 for the case of upgrading from old OST device, you can enable it
344 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
345 enabled, for new created OST-object, its IDIF-in-LMA will contain
346 the OST-index; for the existing OST-object, the OSD will convert
347 old format IDIF as new format IDIF with OST-index stored in the
348 LMA EA when accessing such OST-object or via OI scrub. Once such
349 flag is enabled, it cannot be reverted back, so the system cannot
350 be downgraded to the orignal incompatible version.
352 --------------------------------------------------------------------------------
354 07-30-2014 Intel Corporation
356 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
357 for currently supported client and server kernel versions.
358 * Server known to build on patched kernels:
359 2.6.32-431.20.3.el6 (RHEL6.5)
360 3.0.101-0.31 (SLES11 SP3)
361 * Client known to build on unpatched kernels:
362 2.6.32-431.20.3.el6 (RHEL6.5)
363 3.0.101-0.31 (SLES11 SP3)
366 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
367 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
368 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
369 stack overflow. For more information, please refer to bugzilla 17630.
370 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
371 the re-exporting nodes to be mounted with "32bitapi" mount option
373 --------------------------------------------------------------------------------
375 10-31-2013 Intel Corporation
377 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
378 * Server builds for kernels:
379 2.6.32-358.18.1.el6 (RHEL6.4)
380 * Client builds for unpatched kernels:
381 2.6.32-358.18.1.el6 (RHEL6.4)
382 3.0.92-0.8 (SLES11 SP3)
384 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
385 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
386 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
387 stack overflow. For more information, please refer to bugzilla 17630.
388 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
389 the re-exporting nodes to be mounted with "32bitapi" mount option
391 --------------------------------------------------------------------------------
393 05-30-2013 Intel Corporation
395 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
396 * Server builds for kernels:
397 2.6.32-358.6.2.el6 (RHEL6.4)
398 * Client builds for unpatched kernels:
399 2.6.32-358.6.2.el6 (RHEL6.4)
400 3.0.74-0.6.6 (SLES11 SP2)
402 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
403 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
404 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
405 stack overflow. For more information, please refer to bugzilla 17630.
406 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
407 the re-exporting nodes to be mounted with "32bitapi" mount option
409 --------------------------------------------------------------------------------
411 9-30-2012 Whamcloud, Inc.
413 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
414 * Server support for kernels:
415 2.6.32-279.5.1.el6 (RHEL6)
416 * Client support for unpatched kernels:
417 2.6.32-279.5.1.el6 (RHEL6)
419 2.6.18-238.19.1.el5 (RHEL5)
420 2.6.16.60-0.69.1 (SLES10)
421 * Recommended e2fsprogs version: 1.41.90.wc4
422 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
423 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
424 stack overflow. For more information, please refer to bugzilla 17630.
425 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
426 the re-exporting nodes to be mounted with "32bitapi" mount option
428 Severity : enhancement
430 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
432 Severity : enhancement
434 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
436 Severity : enhancement
438 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
440 Severity : enhancement
442 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
444 Severity : enhancement
446 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
448 --------------------------------------------------------------------------------
450 04-30-2012 Whamcloud, Inc.
452 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
453 * Server support for kernels:
454 2.6.32-220.4.2.el6 (RHEL6)
455 * Client support for unpatched kernels:
456 2.6.32-220.4.2.el6 (RHEL6)
458 2.6.18-238.19.1.el5 (RHEL5)
459 2.6.16.60-0.69.1 (SLES10)
460 * Recommended e2fsprogs version: 1.41.90.wc4
461 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
462 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
463 stack overflow. For more information, please refer to bugzilla 17630.
464 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
465 the re-exporting nodes to be mounted with "32bitapi" mount option
467 Severity : enhancement
469 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
471 Severity : enhancement
473 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
475 Severity : enhancement
477 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
479 Severity : enhancement
481 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
483 --------------------------------------------------------------------------------
485 09-30-2011 Whamcloud, Inc.
487 * Server support for kernels:
488 2.6.18-238.19.1.el5 (RHEL5)
489 2.6.32-131.6.1.el6 (RHEL6)
490 * Client support for unpatched kernels:
491 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
492 2.6.18-238.19.1.el5 (RHEL5)
493 2.6.32-131.6.1.el6 (RHEL6)
494 2.6.29.4-167.fc11 (FC11)
495 2.6.16.60-0.69.1 (SLES10)
497 * Recommended e2fsprogs version: 1.41.90.wc3
498 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
499 Lustre filesystem with 4K stack may cause a stack overflow. For more
500 information, please refer to bugzilla 17630.
501 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
502 mounted with "32bitapi" mount option
504 Severity : enhancement
505 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
507 Severity : enhancement
508 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
510 Severity : enhancement
511 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
514 Frequency : only with SLES10
516 Description: Use OFED "KMP" provided by Novell
517 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
518 Lustre is now built against this package. That means you need to
519 install the ofed-kmp package from Novell for the patchless client.
520 Note that the ofed-kmp that Novell ships may not exactly match the
521 kernel version but should still be compatible.
523 Severity : enhancement
525 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
526 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
528 --------------------------------------------------------------------------------
530 2010-07-15 Oracle, Inc.
532 * Server support for kernels:
533 2.6.18-164.11.1.el5 (RHEL 5)
534 2.6.18-164.11.1.0.1.el5 (OEL 5)
535 * Client support for unpatched kernels:
536 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
537 2.6.18-164.11.1.el5 (RHEL 5),
538 2.6.18-164.11.1.0.1.el5 (OEL 5)
539 2.6.16.60-0.42.8 (SLES 10),
541 2.6.29.4-167.fc11 (FC11)
542 * Recommended e2fsprogs version: 1.41.10-sun2
544 Severity : enhancement
546 Description: Distribute lfs_migrate script with lustre
547 Details : The lfs_migrate mentioned in the manual is now included in the
548 lustre tree in lustre/scripts and will distributed in lustre RPMs.
550 Severity : enhancement
552 Description: Failover nid list added to import proc file.
554 Severity : enhancement
556 Description: "som_preview" mount option required to use SOM
557 Details : The clients will need to use som_preview mount option
561 Description: Lustre does not do 1MB IOs to HW RAID
562 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
564 Severity : enhancement
566 Description: lfs quota output is not very convenient for awk/sed-parsing
567 Details : Some positions in lfs quota output table could be empty or
568 non-empty which made it hard to parse it with scripts, now
569 a dash is put instead of space where there is not supposed
574 Description: Update to OFED 1.5.1
575 Details : OFED 1.5.1 will be used in environments that don't use
576 vendor supplied OFED.
580 Description: Remove lnet. prefix from lctl params display
581 Details : Also renamed lnet.memused to lnet_memused to
586 Description: Use conf_param -d to remove permanant settings.
590 Description: Add list_param -R
591 Details : lctl list_param now supports recursive mode.
594 Bugzilla : 21829, 22312
595 Description: Fix llobdstat.
596 Details : 1. Now updating the write_bytes counter
597 2. Port script fixes from 1.8.
601 Description: decrease the usage of memory on clients.
602 Details : 1. On clients, recycle dentries and inodes unused.
603 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
608 Description: llobdstat fix and enhancement
609 Details : add a counter to set a limit to how many samples will be returned
610 fix a wildcard in the path to limit to obdfilter stats only
614 Description: lfs check only allowed for root
615 Details : Code cleanups and added a regresion test of non-root use of
620 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
621 Details : Fix hr->hr_index overload.
625 Description: Fix FIEMAP functionality.
629 Description: FC11 patchless client support added
633 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
634 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
638 Description: New acc-sm script and various test-framework enhancments.
642 Description: Previously landed patch was reverted.
646 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
650 Description: Add ldiskfs maxdirsize mount option.
654 Description: Fix mkfs.lustre support for 16TB LUNs.
658 Description: Print mmp_check_interval
659 Details : print mmp_check_interval and make it possible to abort
660 mount operation in case it takes too long.
664 Description: Add cascading_rw to lustre mpi tests.
668 Description: Store SOM data in EA.
669 Details : Also several SOM related cleanups.
673 Description: Add refcount to md_open_data
677 Description: Use refcount on lov_request_set to prevent memory corruption.
681 Description: Statahead: use parent dir's i_mutex for synchronization.
685 Description: loadgen improvements.
689 Description: Enable peer health detection in ksocklnd and ko2iblnd.
693 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
697 Description: Fix race on obd_devcie:md_stats
701 Description: Fix SLES 11 client builds.
702 Details : Resolved uninitalized pointer gcc warnings.
706 Description: Allow non-root access for "lfs check".
707 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
711 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
713 Severity : enhancement
715 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
716 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
720 Description: Sanity test_27p fails.
721 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
722 Sanity tests 27 were cleaned up and improved.
726 Description: Unify req format on client/servers
727 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
728 buffers and swabbers approach..
733 Description: drop join file stuff.
735 Severity : enhancement
737 Description: Add 2.6.27 and SLES11 (patchless client) support
741 Description: Fix in ptlrpc_expire_one_request() to print the signed time
744 Severity : enhancement
746 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
749 Severity : enhancement
751 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
753 Severity : enhancement
755 Description: correctly handle big reply message.
756 Details : send LNet event if reply is bigger than buffer and adjust this buffer
761 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
763 Severity : enhancement
765 Description: Add support for OEL5.
769 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
771 Severity : enhancement
773 Description: "weak-modules" support
774 Details : Implement "weak-modules" support which enables kernel modules
775 to be used with any kernel that implements the same kABI. In
776 order to achieve this modules are now installed in
777 /lib/modules/$(uname -r)/updates/kernel on all distributions.
781 Description: client could not reconnect to OST because of an active request.
782 Details : The client now retries to reconnect to the same server, if a
783 connect request failed with EBUSY or -EAGAIN. Test case 35b
784 conf-sanity.sh has been created.
788 Description: Don't allow make backward step on assignin osc next id.
789 Details : race between allocation next id and ll_sync thread can be cause
790 of set wrong osc next id and can be kill valid ost objects.
794 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
798 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
802 Description: Error handling in osc_statfs_interpret() has been improved.
803 Details : Check in osc_statfs_interpret() for EBADR.
807 Description: Do not send statfs() requests to OSTs disabled by administrator.
808 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
810 Severity : enhancement
812 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
814 Severity : enhancement
816 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
818 Severity : enhancement
820 Description: Enhancement for lfs(1) command to use numeric uid/gid.
824 Description: Improved cleanup procedure for conf-sanity test_46a.
828 Description: Do not start bulk transfer, if client tried to reconnect.
829 Details : Do not start bulk transfer, if client tried to reconnect,
830 but failed due to active rpcs. Check for exp_abort_active_req
831 has been implemented.
835 Description: Prevent inconsistences between linux and lustre mount structures.
836 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
837 Make the sleep interruptible.
839 Severity : enhancement
841 Description: provide server to client comms path
842 Details : server to client communications path via new LDLM_SET_INFO rpc
846 Description: truncate starts GFP_FS allocation under transaction causing deadlock
847 Details : ldiskfs_truncate calls grab_cache_page which may start page
848 allocation under an open transaction. This may lead to
849 calling prune_icache with consequent lustre reentrance.
853 Description: send correctly lsm on open replay
854 Details : MDS is trust to LSM size on replay open, but client can set wrong size
859 Description: enable client interface failover
860 Details : When a child reconnects from another NID, properly update export
861 nid hash position and ldlm reverse import.
863 Severity : enhancement
865 Description: Communicate OST degraded/readonly state via statfs to MDS
866 Details : Flags in the statfs returned from OSTs indicate whether the
867 OST is in a degraded RAID state, or if the filesystem has
868 turned read-only after a filesystem error is detected.
873 Description: don't panic if EPROTO was hit when reading symlink
874 Details : correctly handling request reference in error cases.
876 Severity : enhancement
878 Description: Add LustreNetLink, a kernel-userspace communcation path.
880 Severity : enhancement
882 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
887 Description: access to llog context before init.
888 Details : move handling CATALOGS file at osc layer and forbid access to llog
892 Frequency : rare, only when client is writing to OST on local node
894 Description: Avoid deadlock for local client writes
895 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
896 memory freeing context. This allows OST threads to set the
897 PF_MEMALLOC flag on task structures in order to allocate memory
898 from reserved pools and complete IO.
899 Use GFP_HIGHUSER for OST allocations for non-local client writes,
900 so that the OST threads generate memory pressure and allow
901 inactive pages to be reclaimed.
903 Severity : enhancement
905 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
910 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
911 Details : move ldlm namespace creation in setup phase to avoid grab
912 _lprocfs_lock with cli_sem held.
916 Description: Temporarily disable grant shrink.
917 Details : Disable the feature for debugging.
920 Frequency : only when formatting test filesystems
922 Description: Unable to run several mkfs.lustre on loop devices at the same time
923 Details : mkfs.lustre returns error 256 on the concurrent loop devices
924 formatting. The solution is to proper handle the error.
926 Severity : enhancement
928 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
930 Severity : enhancement
932 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
937 Description: resolve race between obd_disconnect and class_disconnect_exports
938 Details : if obd_disconnect will be called to already disconnected export he
939 forget release one reference and osc module can't unloaded.
941 Severity : enhancement
943 Description: Allow tuning service thread via /proc
944 Details : For each service a new
945 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
946 created that can be used to set min/max thread counts, and get the
947 current number of running threads.
952 Descriptoin: don't return error if have partially created objects for file.
953 Details : lov_update_create_set uses set->set_success as index for created
954 objects, so if some requests will be failed, they will have hole
955 at end of array and qos_shrink_lsm for allocate correct lsm.
957 Severity : enhancement
959 Description: Update OFED support to 1.4.1
964 Description: Reduce small size read RPC
965 Details : Set read-ahead limit for every file and only do read-ahead when
966 available read-ahead pages are bigger than 1M to avoid small size
969 Severity : enhancement
971 Description: move AT tunable parameters for more consistent usage
972 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
974 Severity : enhancement
976 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
977 Details : allow skip disconnected ost for send statfs request and hide error
983 Description: fix racy locking of mballoc block bitmaps causing BUG
984 Details : The locking of the mballoc buddy bitmap and the in-memory
985 block bitmap was using two different spin locks in some
986 cases. This made it possible to incorrectly access the
987 mballoc bitmap while another process was modifying it,
988 causing a sanity assertion to fail. While no on-disk corruption
989 was reported, there was some risk of this happening.
992 Frequency : rare, on llog test 6
994 Descriptoin: don't allow connect to already connected import
995 Details : allowing connect to already connected import is hide connecting problem.
1000 Descriptoin: allow kill process which wait statahead result
1001 Details : for some reasons 'ls' can stick in waiting result from statahead,
1002 in this case need way for kill this process.
1004 Severity : enhancement
1006 Description: Add state history info file, enhance import info file
1007 Details : Track import connection state changes in a new osc/mdc proc file;
1008 add overview-type data to the osc/mdc import proc file.
1010 Severity : enhancement
1012 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1014 Severity : enhancement
1016 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1021 Descriptoin: OST grant too much space to client even there are not enough space.
1022 Details : Client will shrink its grant cache to OST if there are no write
1023 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1024 this grant cache if there are already not enough avaible space
1025 (left_space < total_clients * 32M).
1028 Frequency : start MDS on uncleanly shutdowned MDS device
1030 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1031 Details : stay in waiting mds<>ost recovery finished produce random bugs
1032 due race between two ll_sync thread for one lov target. send
1033 ACTIVATE event only if connect realy finished and import have
1037 Frequency : rare, connect and disconnect target at same time
1039 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1040 Details : don't call obd_disconnect under lov_lock. this long time
1041 operation and can block ptlrpcd which answer to connect request.
1046 Descriptoin: don't lose wakeup for imp_recovery_waitq
1047 Details : recover_import_no_retry or invalidate_import and import_close can
1048 both sleep on imp_recovery_waitq, but we was send only one wakeup
1052 Frequency : always with long access acl
1054 Descriptoin: mds can't pack reply with long acl.
1055 Details : mds don't control size of acl but they limited by reint/getattr
1058 Severity : enhancement
1060 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1062 Severity : enhancement
1064 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1067 Frequency : start MDS on uncleanly shutdowned MDS device
1069 Descriptoin: aborting recovery hang on MDS
1070 Details : don't throttle destroy RPCs for the MDT.
1073 Frequency : on remount
1075 Description: external journal device not working after the remount
1076 Details : clear dev_rdonly flag for external journal devices in
1082 Description: shutdown vs evict race
1083 Details : client_disconnect_export vs connect request race.
1084 if client will evicted at this time - we start invalidate
1085 thread without referece to import and import can be freed
1089 Frequency : rare, need acl's on inode.
1091 Description: client can't handle ost additional correctly
1092 Details : if ost was added after client connected to mds client can have
1093 hit lnet_try_match_md ... to big messages to wide striped files.
1094 in this case need teach client to handle config events about add
1095 lov target and update client max ea size at that event.
1097 Severity : enhancement
1099 Description: Changelogs
1100 Details : Changelogs are a lightweight mechanism to track filesystem
1101 metadata and namespace changes. The changelog is recorded
1102 permanently on the MDTs, and is periodically "consumed" / purged
1103 when records are no longer needed.
1105 Severity : enhancement
1107 Description: compact fld format with extents
1108 Details : Store range of seq rather than every seq in FLD. Seq
1109 controller update FLD rather than clients. In Case of CMD, mdt0
1110 has FLD, all other metadata server act as non persistent proxy
1111 for FLD queries and cache fld entries in fld cache.
1116 Description: don't skip ost target if they assigned to file
1117 Details : Drop slow OSCs if we can, but not for requested start idx.
1118 This means "if OSC is slow and it is not the requested
1119 start OST, then it can be skipped, otherwise skip it only
1120 if it is inactive/recovering/out-of-space.
1124 Description: more cleanup in mds_lov
1125 Details : not send LOV EA under replay, we can't know about they size at this
1126 time. Don't allow client connect to mds before any ost connected,
1127 for avoid problems with LOV EA size and returning EIO to client.
1129 Severity : enhancement
1131 Description: Interoperability at server side (Disk interoperability)
1133 Severity : enhancement
1135 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1137 Severity : enhancement
1139 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1141 Severity : enhancement
1143 Description: New client IO stack (CLIO).
1145 Severity : enhancement
1147 Description: Commit on sharing. Eliminate inter-client dependencies between
1148 uncommitted transactions by doing transaction commits.
1149 Thereby clients may recovery independently.
1152 Frequency : Create a symlink file with a very long name
1154 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1155 Details : If there is no extra space in the request for early cancels,
1156 ldlm_req_handles_avail() returns 0 instead of a negative value.
1158 Severity : enhancement
1160 Description: Add /proc entry for import status
1161 Details : The mdc, osc, and mgc import directories now have
1162 an import directory that contains useful import data for debugging
1163 connection problems.
1165 Severity : enhancement
1167 Description: Re-disable certain /proc logging
1168 Details : Enable and disable client's offset_stats, extents_stats and
1169 extents_stats_per_process stats logging on the fly.
1172 Frequency : Only on FC kernels 2.6.22+
1174 Description: oops in statahead
1175 Details : Do not drop reference count for the dentry from VFS when lookup,
1176 VFS will do that by itself.
1178 Severity : enhancement
1180 Description: Generic /proc file permissions
1181 Details : Set /Proc file permissions in a more generic way to enable non-
1182 root users operate on some /proc files.
1186 Description: Hitting mdc_commit_close() ASSERTION
1187 Details : Properly handle request reference release in
1188 ll_release_openhandle().
1192 Description: quota recovery deadlock during mds failover
1193 Details : This patch includes att18982, att18236, att18237 in bz14840.
1195 1. fix osts hang when mds does failover with quotaon
1196 2. prevent watchdog storm when osts threads wait for the
1201 Frequency : only patchless client
1202 Description: add workaround for race between add/remove dentry from hash
1204 Severity : enhancement
1206 Description: Allow OST glimpses to return PW locks
1210 Description: LBUG when llog conf file is full
1211 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1215 Description: Prevent import from entering FULL state when server in recovery
1219 Description: service mount cannot take device name with ":"
1220 Details : Only when device name contains ":/" will mount treat it as
1226 Description: replace ptlrpcd with the statahead thread to interpret the async
1227 statahead RPC callback
1231 Frequency : on recovery
1232 Description: I/O failures after umount during fail back
1233 Details : if client reconnected to restarted server we need join to recovery
1234 instead of find server handler is changed and process self
1235 eviction with cancel all locks.
1237 Severity : enhancement
1239 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1241 Severity : enhancement
1243 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1245 Severity : enhancement
1247 Description: Upcall on Lustre log has been dumped
1248 Details : Allow for a user mode script to be called once a Lustre log has
1249 been dumped. It passes the filename of the dumped log to the
1250 script, the location of the script can be specified via
1251 /proc/sys/lnet/debug_log_upcall.
1256 Description: avoid idr_remove called for id which is not allocated.
1257 Details : Move assigment s_dev for clustered nfs to end of initialization,
1258 to avoid problem with error handling.
1263 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1264 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1265 into CONN_USED_HASH and this prodice warning when put connection
1266 again in unused hash.
1269 Severity : enhancement
1271 Description: File striping can now be set to use an arbitrary pool of OSTs.
1273 Severity : enhancement
1275 Description: Export bytes_read/bytes_write count on OSC/OST.
1279 Description: Early reply size mismatch, MGC loses connection
1280 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1281 the connect flags are properly negotiated.
1286 Description: quotas are not honored with O_DIRECT
1287 Details : all writes with the flag O_DIRECT will use grants which leads to
1288 this problem. Now using OBD_BRW_SYNC to guard this.
1292 Description: add quota statistics
1293 Details : 1. sort out quota proc entries and proc code.
1294 2. add quota statistics
1296 Severity : enhancement
1298 Description: enable quota support for HEAD.
1302 Description: Properly propagate oinfo flags from lov to osc for statfs
1303 Details : restore missing copy oi_flags to lov requests.
1305 Severity : enhancement
1307 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1308 : l_getgroups(8), lst(8), routerstat(8)
1309 Details : included man pages for llobdstat(8), llstat(8),
1310 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1312 Severity : enhancement
1314 Description: configure's --enable-quota should check the
1315 : kernel .config for CONFIG_QUOTA
1316 Details : configure is terminated if --enable-quota is passed but
1317 : no quota support is in kernel
1321 Description: Remove portals compatibility
1322 Details : Remove portals compatibility, not interoperable with releases
1327 Description: Resolve device initialization race
1328 Details : Prevent proc handler from accessing devices added to the
1329 obd_devs array but yet be intialized.
1331 Severity : enhancement
1333 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1335 Severity : enhancement
1337 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1342 Description: Using wrong pointer in osc_brw_prep_request
1343 Details : Access to array[-1] can produce panic if kernel compiled with
1344 CONFIG_PAGE_ALLOC enabled
1348 Description: Client runs out of low memory
1349 Details : Consider only lowmem when counting initial number of llap pages
1353 Description: *optional* service tags registration
1354 Details : if the "service tags" package is installed on a Lustre node
1355 When the filesystem is mounted, a local-node service tag will
1356 be created. See http://inventory.sun.com/ for more information
1357 about the Service Tags asset management system.
1361 Description: Kernel BUG tries to release flock
1362 Details : Lustre does not destroy flock lock before last reference goes
1363 away. So always drop flock locks when client is evicted and
1364 perform unlock regardless of successfulness of speaking to MDS.
1368 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1372 Description: sanity test 65a fails if stripecount of -1 is set
1373 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1378 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1379 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1382 Severity : enhancement
1384 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1388 Description: do not process already freed flock
1389 Details : flock can possibly be freed by another thread before it reaches
1390 to ldlm_flock_completion_ast.
1394 Description: LBUG during stress test
1395 Details : Need properly lock accesses the flock deadlock detection list.
1399 Description: oops in page fault handler
1400 Details : kernel page fault handler can return two special 'pages' in error
1401 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1405 Description: timeout with invalidate import.
1406 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1407 should be handled by ptlrpcd. This produce long age waiting and
1408 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1410 Severity : enhancement
1412 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1416 Description: enable MGS and MDT services start separately
1417 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1418 a co-located MGS without starting the MGS, which is a complement
1419 to 'nosvc' mount option.
1423 Frequency : after recovery
1424 Description: precreate to many object's after del orphan.
1425 Details : del orphan st in oscc last_id == next_id and this triger growing
1426 count of precreated objects. Set flag LOW to skip increase count
1427 of precreated objects.
1431 Frequency : rare, on clear nid stats
1432 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1433 Details : when clean nid stats sometimes try destroy live entry,
1434 and this produce panic in free.
1438 Description: Stack overflow during MDS log replay
1439 ease stack pressure by using a thread dealing llog_process.
1443 Description: wait until IO finished before start new when do lock cancel.
1444 Details : VM protocol want old IO finished before start new, in this case
1445 need wait until PG_writeback is cleared until check dirty flag and
1446 call writepages in lock cancel callback.
1448 Severity : enhancement
1450 Description: using special macro for print time and cleanup in includes.
1454 Description: mds_mfd_close() ASSERTION(rc == 0)
1455 Details : In mds_mfd_close(), we need protect inode's writecount change
1456 within its orphan write semaphore to prevent possible races.
1460 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1464 Frequency : rare, on shutdown ost
1465 Description: don't hit live lock with umount ost.
1466 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1467 use shrink_dcache_sb instead.
1471 Description: don't panic with use echo client
1472 Details : echo client pass NULL as client nid pointer and this produce null
1473 pointer dereference.
1477 Description: fix build on ppc32
1478 Details : compile code with -m64 flag produce wrong object file for ppc32.
1482 Description: add message levels for liblustreapi
1486 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1487 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1488 will result in an -EAGAIN and in the exit call path we call
1489 alloc_rr() which will with increasing aggressiveness attempt to
1490 aquire precreated objects on the minimum number of required OSCs.
1494 Description: Use old size assignment to avoid deadlock
1495 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1496 the scheduling while holding a spinlock. We do not need locking
1497 for size in ll_update_inode() because size is only updated from
1498 the MDS for directories or files without objects, so there is no
1499 other place to do the update, and concurrent access to such inodes
1500 are protected by the inode lock.
1504 Description: resolve "_IOWR redefined" build error on SLES10
1508 Description: dump the memory debugging after all modules are unloaded to
1509 suppress false negative in conf_sanity test 39
1511 Severity : enhancement
1513 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1518 Description: fix SLES kernel versioning
1519 Details : the kernel version for our SLES 10 kernel did not include a "-"
1520 before the "smp" at the end. while this was not a problem in
1521 general, it did mean that software trying to use the kernel
1522 version to try to detect a vendor specific kernel would fail.
1523 this was most evident by the OFED build scripts.
1527 Description: Don't update lov_desc members until making sure they are valid
1528 Details : When updating lov_desc members via proc fs, need fix their
1529 validities before doing the real update.
1533 Description: don't put request into delay list while invalidate in flight.
1534 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1535 import in flight. this produce timeout for invalidate and sometimes
1536 can cause stale data.
1540 Frequency : on ppc only
1541 Description: not convert ost objects for directory because it's not exist.
1542 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1546 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1548 Severity : enhancement
1550 Description: Update to support 2.6.22.14 vanilla kernel.
1554 Frequency : rare, on recovery
1555 Description: read procfs can produce deadlock in some situation
1556 Details : Holding lprocfs lock which send rpc can produce block for destroy
1557 obd objects and this also block reconnect with -EALREADY. This
1558 isn't fix all lprocfs bugs - but make it rare.
1560 Severity : enhancement
1562 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1565 Frequency : frequent on X2 node
1567 Description: mdc_set_open_replay_data LBUG
1568 Details : Set replay data for requests that are eligible for replay.
1572 Description: lustre_mgs: operation 101 on unconnected MGS
1573 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1574 MGC, and late on MGC cannot successfully connect to MGS and a lot
1575 of the error messages complaining that MGS is not connected.
1579 Frequency : on network error
1580 Description: panic with double free request if network error
1581 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1582 but it's true only for synchronus enqueue, for async enqueue
1583 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1586 Severity : enhancement
1588 Description: client-side metadata stat-ahead during readdir(directory readahead)
1589 Details : perform client-side metadata stat-ahead when the client detects
1590 readdir and sequential stat of dir entries therein
1593 Frequency : on start mds
1595 Description: Implement get_info(last_id) in obdfilter.
1598 Frequency : occasional
1600 Description: Correctly check stale fid, not start epoch if ost not support SOM
1601 Details : open with flag O_CREATE need set old fid in op_fid3 because
1602 op_fid2 was overwritten with new generated fid, but mds can answer
1603 with one of these two fids and both is not stale. Setattr
1604 incorrectly started an epoch and assume will be called
1605 done_writing, but without SOM done_writing ever being called.
1608 Frequency : rare, depends on device drivers and load
1610 Description: MDS or OSS nodes crash due to stack overflow
1611 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1612 In some cases, in conjunction with device drivers that use a lot
1613 of stack the MDS (or possibly OSS) service threads could overflow
1614 the stack. One change which was identified to consume additional
1615 stack has been reworked to avoid the extra stack usage.
1618 Frequency : occasional
1620 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1621 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1622 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1623 oscc_has_objects and see OSC has no objests and call
1624 oscc_internal_create to resend create request.
1626 Severity : enhancement
1628 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1630 Severity : enhancement
1632 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1635 Frequency : very rare
1637 Description: Fix replay if there is an un-replied request and open
1638 Details : In some cases, older replay request will revert the
1639 mcd->mcd_last_xid on MDS which is used to record the client's
1640 latest sent request.
1642 Severity : enhancement
1644 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1646 Severity : enhancement
1648 Description: Add rhel5 support to HEAD.
1650 Serverity : enhancement
1652 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1657 Description: Don't allow skipping OSTs if index has been specified.
1658 Details : Don't allow skipping OSTs if index has been specified, make
1659 locking in internal create lots better.
1663 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1664 Details : access to bitfield in structure is always rounded to long
1665 and this produce problem with not atomic change any bit.
1669 Description: Lustre make rpms failed.
1670 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1671 builds Lustre rpms from tarball.
1674 Frequency : rare on shutdown ost
1676 Description: If llog cancel was not send before clean_exports phase, this can
1677 produce deadlock in llog code.
1678 Details : If llog thread has last reference to obd and call class_import_put
1679 this produce deadlock because llog_cleanup_commit_master wait when
1680 last llog_commit_thread exited, but this never success because was
1681 called from llog_commit_thread.
1685 Description: allow userland application know is lost one of stripes.
1686 Details : fill lvb_blocks with error code on ost and return it to
1687 application if error flag found.
1691 Description: NULL lov_tgts causing MDS oops
1692 Details : more safe checks for NULL lov_tgts for avoid oops.
1694 Severity : enhancement
1696 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1700 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1701 Details : If an OST is down the MDS will hang indefinitely in
1702 obd_statfs_rqset() waiting for the statfs data. While for
1703 MDS QOS usage of statfs, it should not stuck in waiting.
1705 Severity : enhancement
1707 Description: remote_acl support
1708 Details : Support ACL-based permission check for remote user.
1709 Support setfacl/getfacl for remote user with the utils
1710 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1711 the system "{s,g}etfacl" utils.
1713 Severity : enhancement
1715 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1717 Severity : enhancement
1719 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1723 Description: make mgs_setparam() handle fsname containing dash
1724 Details : fsname containing a dash does not work with lctl conf_param
1726 Severity : enhancement
1728 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1730 Severity : enhancement
1732 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1736 Frequency : rare, at shutdown
1737 Description: access already free / zero obd_namespace.
1738 Details : if client_disconnect_export was called without force flag set,
1739 and exist connect request in flight, this can produce access to
1740 NULL pointer (or already free pointer) when connect_interpret
1741 store ocd flags in obd_namespace.
1745 Frequency : only at startup
1746 Description: not alloc memory with spinlock held.
1747 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1748 if any spinlock held.
1750 Severity : enhancement
1752 Description: make lustre randomly fail allocating memory
1753 Details : Make lustre randomly failed allocating memory for testing purpose.
1755 Severity : enhancement
1757 Description: lost problems with lov objid file
1758 Details : Fixes some scability and access to not inited memory problems
1759 in work with lov objdid file.
1764 Description: lfs find does not continue on file error
1765 Details : Continue other files processing when a file/dir is absent.
1769 Description: Inconsistent usage of lustre_pack_reply()
1770 Details : Standardize the usage of lustre_pack_reply() such that it
1771 always generate a CERROR on failure.
1774 Frequency : occasional
1776 Description: MDS hang or stay in waiting lock
1777 Details : If client receive lock with CBPENDING flag ldlm need send lock
1778 cancel as separate rpc, to avoid situation when cancel request
1779 can't processed due all i/o threads stay in wait lock.
1783 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1787 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1791 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1795 Description: Update to RHEL4 latest kernel.
1797 Severity : enhancement
1799 Description: Build SLES10 patchless client fails
1800 Details : The configure was broken by run ./configure with
1801 --with-linux-obj=.... argument for patchless client. When the
1802 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1803 can't search header adequately. Use absolute path such as
1804 -I($LINUX)/include instead.
1808 Description: interrupt oig_wait produce painc on resend.
1809 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1810 private set, and this produce situation when rq_ptlrpcd_data not
1811 copyed to new allocated request and triggered LBUG on assert
1812 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1813 ptlrpcd set if request is changed and can be safety changed to use
1818 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1819 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1820 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1821 these macros and replace them with their difinitions instead.
1825 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1826 Details : Update which_patch & target file for SLES10 latest kernel.
1829 Description: Remove client patches from SLES 10 kernel.
1830 Details : This causes SLES 10 clients to behave as patchless clients
1831 even on a Lustre-patched (server) kernel.
1833 Severity : enhancement
1835 Description: self-adjustable client's lru lists
1836 Details : use adaptive algorithm for managing client cached locks lru
1837 lists according to current server load, other client's work
1838 pattern, memory activities, etc. Both, server and client
1839 side namespaces provide number of proc tunables for controlling
1842 Severity : enhancement
1844 Description: light-weight GSS support
1845 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1846 and reduce performance overhead.
1848 Severity : enhancement
1850 Description: Linux keyring support
1851 Details : Support using service of Linux keyring for Lustre GSS internal
1852 context refresh/cache mechanism.
1856 Description: Fix errors in lfs documentation
1857 Details : Fixes man pages
1861 Description: don't use GFP_* in generic Lustre code.
1862 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1866 Description: obdclass is limited by single OBD_ALLOC(idarray)
1867 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1871 Description: warnings with build patchless client with vanila 2.6.19 and up
1872 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1873 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1877 Frequency : Only for SLES
1878 Description: sanity_quota fail test_1
1879 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1880 file, which makes TSTID[2] inunique.
1884 Frequency : Only for SLES9
1885 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1886 Details : Update target/ChangeLog/which_patch .
1890 Frequency : Only for test_10 in sanity-quota.sh
1891 Description: a bug in quota test code
1892 Details : checking a test flag in a wrong place
1896 Frequency : Only for test_13 in sanity-quota.sh
1897 Description: a bug in quota test script
1898 Details : shouldn't check overfull allocated quota
1902 Description: jbd statistics
1903 Details : Port older jbd statistics patch for sles10
1905 Severity : enhancement
1907 Description: Add printing inode info into message about error in writepage.
1912 Description: reply_lock_interpret crash due to race with it and lock cancel.
1913 Details : Do not replay locks that are being cancelled. Do not reference
1914 locks by their address during replay, just by their handle.
1918 Frequency : When flocks are used.
1919 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1920 Details : Flock locks might destroy just granted lock if it could be merged
1921 with another existing flock, this is done in completion handler,
1922 so teach ldlm_cli_enquque_fini that this is a valid case for
1928 Description: Oops in read and write path when failing to allocate lock.
1929 Details : Check if lock allocation failed and return error back.
1933 Description: lstripe command fails for valid OST index
1934 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1935 instead of lov->desc.ld_active_tgt_count.
1939 Description: sanity.sh failed test 103
1940 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1945 Description: df doesn't work properly if diskfs blocksize != 4K
1946 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1950 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1951 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1953 Severity : enhancement
1955 Description: Adaptive timeouts
1956 Details : RPC timeouts adapt to changing server load and network
1957 conditions to reduce resend attempts and improve recovery time.
1961 Description: llapi_file_create() does not allow some changes
1962 Details : add llapi_file_open() that allows specifying the mode and
1963 open flags, and also returns an open file handle.
1967 Description: merge and cleanup kernel patches.
1968 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1972 Description: Add journal checksum support.(Kernel part)
1973 Details : The journal checksum feature adds two new flags i.e
1974 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1975 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1976 indicates that the commit block contains the checksum for
1977 the blocks described by the descriptor blocks. Now commit
1978 record can be sent to disk without waiting for descriptor
1979 blocks to be written to disk. This behavior is controlled
1980 using JBD2_FEATURE_ASYNC_COMMIT flag.
1984 Description: OSS needs mutliple precreate threads
1985 Details : Add ability to start more than one create thread per OSS.
1989 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1990 Details : execute lfs setstripe on client
1994 Description: mds_obd_create error creating tmp object
1995 Details : When the user sets quota on root, llog will be affected and can't
1996 create files and write files.
1999 Frequency : Always on ia64 patchless client, and possibly others.
2001 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2002 Details : This allows the patchless client to be loaded on architectures
2003 without this export.
2007 Description: RedHat Update kernel for RHEL5
2008 Details : Add the kernel config file for RHEL5.
2012 Description: RedHat Update kernel for RHEL5
2013 Details : Modify the kernel config file more closer RHEL5.
2017 Description: Build failure against Centos5 (RHEL5)
2018 Details : Define PAGE_SIZE when it isn't present.
2022 Description: test_7 on sanity-quota.sh can't be used on separate machines
2023 Details : add facet to handle it
2027 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2028 Details : with NFS, the anon dentry's parent was set to itself in
2029 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2030 corresponding dentry other than use rec->ur_name.
2032 Severity : enhancement
2034 Description: lfs setstripe enhancement
2035 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2039 Description: enable data checksumming by default
2040 Details : enable checksum by default, allow --disable-checksum
2041 configure option and "-o nochecksum" mount option. Checksums
2042 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2046 Description: lustre support for RHEL5
2047 Details : Add support for RHEL5.
2051 Description: Client eviction due to failover config
2052 Details : after a connection loss, the lustre client should attempt to
2053 reconnect to the last active server first before trying the
2054 other potential connections.
2058 Description: when mds and osts use different quota unit(32bit and 64bit),
2059 quota will be released repeatly.
2060 Details : void sending multiple quota reqs to mds, which will keep the
2061 status between the reqs.
2065 Description: osts not allocated evenly to files
2066 Details : change the condition to increase offset_idx
2069 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2071 Description: Data corruption for OSTs that are formatted larger than 2TB
2073 Details : When generating the bio request for lustre file writes the
2074 sector number would overflow a temporary variable before being
2075 used for the IO. The data reads correctly from Lustre (which
2076 will overflow in a similar manner) but other file data or
2077 filesystem metadata may be corrupted in some cases.
2081 Description: Tune the kernel for good SCSI performance.
2082 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2083 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2088 Description: rewrite ext2-derived code in obdclass/uuid.c
2089 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2090 from scratch preserving functionality.
2094 Description: "lfs find -obd UUID" prints directories
2095 Details : "lfs find -obd UUID" will return all directory names instead
2096 of just file names. It is incorrect because the directories
2097 do not reside on the OSTs.
2101 Description: use i_size_read and i_size_write in 2.6 port
2102 Details : replace inode->i_size access with i_size_read/write()
2105 Frequency : when removing large files
2107 Description: scheduling issue during removal of large Lustre files
2108 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2109 It causes scheduling issues when removing large files (17TB in the
2113 Frequency : only with liblustre clients on XT3
2115 Description: evictions taking too long
2116 Details : allow llrd to evict clients directly on OSTs
2119 Frequency : only on ppc
2121 Description: /proc/fs/lustre/devices broken on ppc
2122 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2123 We should make sure the type of 'index' is loff_t and then cast
2124 explicitly as needed below. Do not assign an explicitly cast
2128 Frequency : only for rhel5
2130 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2131 Details : Modify the target file & which_kernel.
2133 Severity : enhancement
2135 Description: omit set fsid for export NFS
2136 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2137 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2141 Frequency : Always, for kernels after 2.6.16
2142 Description: Fix warning idr_remove called for id=.. which is not allocated.
2143 Details : Last kernels save old s_dev before kill super and not allow
2144 to restore from callback - restore it before call kill_anon_super.
2148 Description: buffer overruns could theoretically occur
2149 Details : llapi_semantic_traverse() modifies the "path" argument by
2150 appending values to the end of the origin string, and a
2151 overrun may occur. Adding buffer overrun check in liblustreapi.
2155 Description: Fix error on 'ls .' at the top of the Lustre mount.
2156 Details : Don't revalidate dentry if it is a root dentry.
2160 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2161 Details : Modify vm-tunables-rhel4.patch.
2165 Description: Kernel config for 2.6.18-vanilla.
2166 Details : Modify targets/2.6-vanilla.target.in.
2167 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2168 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2169 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2170 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2174 Description: improve handling recoverable errors
2175 Details : if request processig with error which can be recoverable on server
2176 request should be resend, otherwise page released from cache and
2181 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2182 Details : Modify target file & which_patch.
2183 A flaw was found in the IA32 system call emulation provided
2184 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2185 value could be stored in the %RAX register, which could trigger an
2186 out-of-bounds system call table access. An untrusted local user
2187 could exploit this flaw to run code in the kernel
2188 (ie a root privilege escalation). (CVE-2007-4573).
2192 Description: O_DIRECT bypasses client statistics.
2193 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2194 read_ahead_stats were updated. copied the stats section from
2195 osc_send_oap_rpc() into async_internal().
2199 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2203 Description: change order of libsysio includes
2204 Details : '#include sysio.h' should always come before '#include xtio.h'
2208 Description: handle "serious error: objid * already exists" more gracefully
2209 Details : If LAST_ID value on disk is smaller than the objects existing in
2210 the O/0/d* directories, it indicates disk corruption and causes an
2211 LBUG(). If the object is 0-length, then we should use the existing
2212 object. This will help to avoid a full fsck in most cases.
2214 Severity : enhancement
2216 Description: adapt the lustre_config script to support the upgrade case
2217 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2218 server targets to 1.6 in parallel.
2222 Description: To avoid grant space > avaible space when the disk is almost
2223 full. Without this patch you might see the error "grant XXXX >
2224 available" or some LBUG about grant, when the disk is almost
2226 Details : In filter_check_grant, for non_grant cache write, we should
2227 check the left space by if (*left > ungranted + bytes), instead
2228 of (*left > ungranted), because only we are sure the left space
2229 is enough for another "bytes", then the ungrant space should be
2230 increase. In client, we should update cl_avail_grant only there
2231 is OBD_MD_FLGRANT in the reply.
2235 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2236 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2237 described in CVE-2007-4573. This problem could allow untrusted
2238 local users to gain root access.
2241 Frequency : when using O_DIRECT and quotas
2243 Description: Incorrect file ownership on O_DIRECT output files
2244 Details : block usage reported by 'lfs quota' does not take into account
2245 files that have been written with O_DIRECT.
2250 Description: touch file failed when fs is not full
2251 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2252 otherwise we can get ENOSP while fs is not full.
2256 Description: parallel lock callbacks
2257 Details : Instead of sending blocking and completion callbacks as separated
2258 requests, adding them to a set and sending in parallel.
2261 Frequency : only for Cray XT3
2262 Bugzilla : 12829/13455
2263 Description: Changing primary group doesn't change the group lustre assigns to
2265 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2266 overridden with the primary group provided by the group upcall,
2267 whereas the supplied fsgid can be trusted if it is in the list of
2268 supplementary groups returned by the group upcall.
2270 Severity : enhancement
2272 Description: Allow masking D_WARNING, D_ERROR messages from console
2273 Details : Console messages can now be disabled via lnet.printk.
2277 Description: User code with malformed file open parameter crashes client node
2278 Details : Before packing join_file req, all the related reference should be
2279 checked carefully in case some malformed flags cause fake
2280 join_file req on client.
2284 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2285 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2286 mode is equal to the granted mode. However, it should also check
2287 that LDLM_FL_CP_REQD is not set.
2291 Description: Heavy nfs access might result in deadlocks
2292 Details : After ELC code landed, it is now improper to enqueue any mds
2293 locks under och_sem, because enqueue might want to decide to
2294 cancel open locks for same inode we are holding och_sem for.
2298 Description: Client eviction while running blogbench
2299 Details : A lot of unlink operations with concurrent I/O can lead to a
2300 deadlock causing evictions. To address the problem, the number of
2301 oustanding OST_DESTROY requests is now throttled to
2302 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2303 callbacks are processed in priority.
2307 Description: enable ACLs on MDS by default
2308 Details : ACLs must be enabled on MDS by default.
2311 Frequency : PPC/PPC64 only
2313 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2314 Details : fix duplicated definitions between asm-ppc64/types.h and
2315 lustre_types.h on PPC.
2318 Frequency : PPC/PPC64 only
2320 Description: asm-ppc/segment.h does not exist
2321 Details : fix compile issue on PPC.
2325 Description: better handle error messages in extents code
2328 Frequency : RHEL4 only
2330 Description: mkfs is very slow on IA64/RHEL4
2331 Details : A performance regression has been discovered in the MPT Fusion
2332 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2333 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2334 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2336 Severity : enhancement
2338 Description: SNMP support enhancement
2339 Details : Adding total number of sampled request for an MDS node in snmp
2342 Severity : enhancement
2344 Description: Optimize ldlm waiting list processing for PR extent locks
2345 Details : When processing waiting list for read extent lock and meeting read
2346 lock that is same or wider to it that is not contended, skip
2347 processing rest of the list and immediatelly return current
2348 status of conflictness, since we are guaranteed there are no
2349 conflicting locks in the rest of the list.
2353 Description: Time out and refuse to reconnect
2354 Details : When the failover node is the primary node, it is possible
2355 to have two identical connections in imp_conn_list. We must
2356 compare not conn's pointers but NIDs, otherwise we can defeat
2357 connection throttling.
2361 Description: port llog fixes from b1_6 into HEAD
2362 Details : Port llog reference couting and some llog cleanups from b1_6
2363 (bug 10800) into HEAD, for protect from panic and access to already
2364 free llog structures.
2368 Description: Detect stride IO mode in read-ahead
2369 Details : When a client does stride read, read-ahead should detect that and
2370 read-ahead pages according to the detected stride pattern.
2374 Description: data checksumming impacts single node performance
2375 Details : add support for several checksum algorithm. Currently, only CRC32
2376 and Adler-32 are supported. The checksum type can be changed on
2377 the fly via /proc/fs/lustre/osc/*/checksum_type.
2381 Description: use adler32 for page checksums
2382 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2387 Description: build for x2 fails
2388 Details : fix compile issue on Cray systems.
2392 Description: Properly match for duplicate locks
2393 Details : Due to different lock order from skiplists code, we need to
2394 traverse entire list for now
2397 Frequency : only on PPC/SLES10
2399 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2400 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2401 be defined. Add a hack in mkfs_lustre.c to work around this compile
2406 Description: LASSERT on MDS when client holding flock lock dies
2407 Details : ldlm pool logic depends on number of granted locks equal to
2408 number of released locks which is not true for flock locks, so
2409 just exclude such locks from consideration.
2413 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2414 Details : Use fsfilt_sync() for both the whole filesystem sync and
2415 individual file sync to eliminate dangerous inode locking
2416 with I_LOCK that can lead to a deadlock.
2420 Description: performance in 1.6.3
2421 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2422 sure that 1MB requests are not fragmented by the block layer.
2424 Severity : enhancement
2426 Description: organize the server-side client stats on per-nid basis
2427 Details : Change the structure of stats under obdfilter and mds to
2435 The "uuid"s file would list the uuids of _active_ exports.
2436 And the clear entry is to clear all stats and stale nids.
2438 Severity : enhancement
2440 Description: eliminate client locks in face of contention
2441 Details : file contention detection and lockless i/o implementation
2442 for contended files.
2446 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2449 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2451 Description: MDS may lose file striping (and hence file data) in some cases
2452 Details : If there are additional extended attributes stored on the MDS,
2453 in particular ACLs, SELinux, or user attributes (if user_xattr
2454 is specified for the client mount options) then there is a risk
2455 of attribute loss. Additionally, the Lustre file striping
2456 needs to be larger than default (e.g. striped over all OSTs),
2457 and an additional attribute must be stored initially in the
2458 inode and then increase in size enough to be moved to the
2459 external attribute block (e.g. ACL growing in size) for file
2464 Description: skiplist implementation simplification
2465 Details : skiplists are used to group compatible locks on granted list
2466 that was implemented as tracking first and last lock of each
2467 lock group the patch changes that to using doubly linked lists
2471 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2472 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2473 because LASSERT on both the data supplied by a client, and the
2474 data on disk is dangerous and incorrect.
2476 Severity : enhancement
2478 Description: Slow truncate/writes to huge files at high offsets.
2479 Details : Directly associate cached pages to lock that protect those pages,
2480 this allows us to quickly find what pages to write and remove
2481 once lock callback is received.
2485 Description: more ldlm soft lockups
2486 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2487 starve other threads from the resource lock for a long time in
2488 case of long waiting queue, so change the debug level from
2489 D_OTHER to the less frequently used D_INFO.
2491 Severity : enhancement
2493 Description: add -gid, -group, -uid, -user options to lfs find
2497 Description: Hung threads in invalidate_inode_pages2_range
2498 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2499 once one is completed. As a result, some RPCs are stuck in the
2500 queue and are never sent.
2504 Description: filter threads hungs on waiting journal commit
2505 Details : Cleanup filter group llog code, then only filter group llog will
2506 be only created in the MDS/OST syncing process.
2510 Description: Procfs and llog threads access destoryed import sometimes.
2511 Details : Sync the import destoryed process with procfs and llog threads by
2512 the import refcount and semaphore.
2514 Severity : enhancement
2516 Description: openlock cache of b1_6 port to HEAD
2521 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2522 Details : If insertion of an extent fails, then discard the inode
2523 preallocation and free data blocks else it can lead to duplicate
2528 Description: don't always update ctime in ext3_xattr_set_handle()
2529 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2530 In some cases the ctime should not be updated, for example for
2531 2.0->1.8 compatibility it is necessary to delete an xattr and it
2532 should not update the ctime.
2536 Bugzilla : 15713/16362
2537 Description: Assertion in iopen_connect_dentry in 1.6.3
2538 Details : looking up an inode via iopen with the wrong generation number can
2539 populate the dcache with a disconneced dentry while the inode
2540 number is in the process of being reallocated. This causes an
2541 assertion failure in iopen since the inode's dentry list contains
2542 both a connected and disconnected dentry.
2546 Description: assertion failure in ldlm_handle2lock()
2547 Details : fix a race between class_handle_unhash() and class_handle2object()
2548 introduced in lustre 1.6.5 by bug 13622.
2553 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2554 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2559 Description: oops in sd_iostats_seq_show()
2560 Details : unloading/reloading the scsi low level driver triggers a kernel
2561 bug when trying to access the sd iostat file.
2566 Description: Kernel panics during QLogic driver reload
2567 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2568 patch, causing memory corruption.
2573 Description: journal_dev option does not work in b1_6
2574 Details : pass mount option during pre-mount.
2576 Severity : enhancement
2578 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2579 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2580 extent information of a file. It can be used to map logical blocks
2581 in a file to physical blocks in the block device.
2585 Description: LDLM soft lockups - improvement
2586 Details : It is be possible to send the lock handle along with each read
2587 or write request because the client is already doing a lock match
2588 itself so there isn't any reason the OST should have to re-do that
2592 Frequency : only on Cray X2
2594 Description: X2 build failures
2595 Details : fix build failures on Cray X2.
2599 Description: xid & resent requests
2600 Details : Initialize RPC XID from clock at startup (randomly if clock is
2603 Severity : enhancement
2605 Description: Add lustre_start utility to start or stop multiple Lustre servers
2610 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2611 Details : In case of memory pressure, list_del() can be called twice on
2612 req->rq_history_list, causing a kernel oops.
2616 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2617 Details : fix stack overflow in the distributed lock manager by defering
2618 export eviction after a failed AST to the elt thread instead of
2619 handling it in the dlm interpret routine.
2623 Description: Convert some comments to new format.
2624 Details : Update documenting comments to match doxygen conventions.
2628 Description: Grammar fixes.
2629 Details : A couple of trivial spelling fixes.
2633 Description: OSD_COUNTERS-mandatory
2634 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2635 to simplify the code.
2639 Description: simplify cmm_device freeing logic.
2640 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2644 Description: Add lockdep support to dt_object_operations locking interface.
2645 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2646 parameter indicating lock ordering. Update mdd code to use new
2651 Description: Introduce failloc constants for lockless IO tests.
2652 Details : Add two new failloc constants to test lockless IO. Only one of
2653 them in implemented---another is checked in yet to be landed
2658 Description: Add lockdep support for inode mutex.
2659 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2660 in the situations where Lustre has to lock more than one inode
2665 Description: Add optional invariants checking support.
2666 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2667 configure switch. This macro is to be used for consistency and
2668 sanity checks that are too expensive to be left in `production'
2673 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2674 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2679 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2680 Details : Use spin_lock_nested() in (the only) situation where more than
2681 one ldlm_lock is locked simultaneously. Also, fix possible
2682 dead-lock in ldlm_lock_change_resource() by enforcing particular
2687 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2688 Details : Instead of specifying each ldlm_lock call-back through separate
2689 parameter, wrap them into struct ldlm_callback_suite.
2693 Description: Kill join_lru obd method and its callers.
2694 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2695 in memory---a requirement implemented through ->o_join_lru() obd
2696 method in HEAD. Remove this method and its users.
2700 Description: Add asynchronous ldlm ENQUEUE completion handler.
2701 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2702 case that stock ldlm_completion_ast() cannot handle as it waits
2703 until lock is granted. Introduce new ldlm_completion_ast_async()
2704 for this. Also comment ldlm_completion_ast().
2708 Description: ldlm_error <-> errno conversion.
2709 Details : Add functions to map (rather arbitrary) between LDLM error codes
2710 and standard errno values. CLIO needs this to prevent LDLM specific
2711 constants from escaping ldlm and osc.
2715 Description: Kill unused ldlm_handle2lock_ns() function.
2716 Details : Kill unused ldlm_handle2lock_ns() function.
2720 Description: Add lu_ref support to ldlm_lock
2721 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2722 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2723 and ldlm_resource. LDLM interface has to be changed, because code that
2724 releases a reference on a lock, has to "know" what reference this is.
2725 In the most frequent case
2727 lock = ldlm_handle2lock(handle);
2729 LDLM_LOCK_PUT(lock);
2731 no changes are required. When any other reference (received _not_ from
2732 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2733 instead of LDLM_LOCK_PUT().
2735 Arguably, changes are pervasive, and interface requires some discipline
2736 for proper use. On the other hand, it was very instrumental in finding
2737 a few leaked lock references.
2741 Description: Add ldlm_lock_addref_try().
2742 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2743 attempts to addref a lock that might be being canceled
2748 Description: Add ldlm_weigh_callback().
2749 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2750 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2751 hard-coded `number of pages' logic.
2755 Description: Add lockdep annotations to llog code.
2756 Details : Use appropriately tagged _nested() locking calls in the places
2757 where llog takes more than one ->lgh_lock lock.
2761 Description: Add loi_kms_set().
2762 Details : Wrap kms updates into a helper function.
2766 Description: Constify instances of struct lsm_operations.
2767 Details : Constify instances of struct lsm_operations.
2771 Description: lu_conf support.
2772 Details : On a server, a file system object is uniquely identified
2773 by a fid, which is sufficient to locate and load all object
2774 state (inode). On a client, on the other hand, more data are
2775 necessary instantiate an object. Change lu_object_find() and
2776 friends to take additional `lu_conf' argument describing object.
2777 Typically this includes layout information.
2781 Description: lu_context fixes.
2782 Details : Introduce new lu_context functions that are needed on the client
2783 side, where some system threads (ptlrpcd) are shared by multiple
2784 modules, and so cannot be stopped during module shutdown.
2788 Description: Add start and stop methods to lu_device_type_operations.
2789 Details : Introduce two new methods in lu_device_type_operations, that are
2790 invoked when first instance of a given type is created and last one
2791 is destroyed respectively. This is need by CLIO.
2795 Description: Add lu_ref support to struct lu_device.
2796 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2797 track leaked references.
2801 Description: Introduce lu_kmem_descr.
2802 Details : lu_kmem_descr and its companion interface allow to create
2803 and destroy a number of kmem caches at once.
2807 Description: Fix lu_object finalization race.
2808 Details : Fix a race between lu_object_find() finding an object and its
2809 concurrent finalization. This race is (most likely) not possible
2810 on the server, but might happen on the client.
2814 Description: Introduce lu_ref interface.
2815 Details : lu_ref is a debugging module allowing to track references to
2816 a given object. It is quite cpu expensive, and has to be
2817 explicitly enabled with --enable-lu_ref. See usage description
2822 Description: Factor lu_site procfs stats into a separate function.
2823 Details : Separate lu_site stats printing code into a separate function
2824 that can be reused on a client.
2828 Description: Constify instances of struct {lu,dt,md}_device_operations.
2829 Details : Constify instances of struct {lu,dt,md}_device_operations.
2833 Description: Introduce struct md_site and move meta-data specific parts of
2834 struct lu_site here.
2835 Details : Move md-specific fields out of struct lu_site into special struct
2836 md_site, so that lu_site can be used on a client.
2840 Description: Kill mdd_lov_destroy().
2841 Details : Remove unused mdd code.
2845 Description: Add st_block checking to multistat.c.
2846 Details : Add st_block checking to multistat.c.
2850 Description: Add lu_ref support to struct obd_device.
2851 Details : Add lu_ref tracking to obd_device.
2855 Description: Kill obd_set_fail_loc().
2856 Details : Remove unused code.
2860 Description: Add special type for ptlrpc_request interpret functions.
2861 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2862 there. Actual usage will be in CLIO.
2866 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2867 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2868 doesn't work with lockdep.
2872 Description: Add rwv.c test program.
2873 Details : New testing program exercising readv(2) and writev(2) (Qian).
2877 Description: Add sendfile.c test program.
2878 Details : New testing program exercising sendfile(2) (Jay).
2882 Description: Ratelimit a message that can be very frequent.
2883 Details : Ratelimit a memory allocation failure message that can
2888 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2889 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2893 Description: Liblustre build fixes.
2894 Details : Liblustre build fixes.
2898 Description: libcfs: add cfs_{need,cond}_resched() interface.
2899 Details : libcfs: add cfs_{need,cond}_resched() definition and
2900 implementations for Linux, NT, and liblustre.
2902 Severity : enhancement
2904 Description: More exported tunables for mballoc
2905 Details : Add support for tunable preallocation window and new tunables for
2906 large/small requests
2910 Description: Detect corruption of block bitmap and checking for preallocations
2911 Details : Checks validity of on-disk block bitmap. Also it does better
2912 checking of number of applied preallocations. When corruption is
2913 found, it turns filesystem readonly to prevent further corruptions.
2917 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2918 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2919 window to 0. If it does hit cache miss, and read-pattern is still
2920 stride-io mode, does not reset the stride window, but also does not
2921 increase the stride window length in this case.
2925 Frequency : only for big-endian servers
2926 Description: Check if system is big-endian while mounting fs with extents feature
2927 Details : Mounting a filesystem with extents feature will fail on big-endian
2928 systems since ext3-based ldiskfs is not supported on big-endian
2929 systems. This can be over-riden with "bigendian_extents" mount option.
2931 Severity : enhancement
2933 Description: The root squash functionality
2934 Details : A security feature, which is to prevent users from being able
2935 to mount lustre on their desktop, run as root, and delete
2936 all of the files in the filesystem. The goal is accomplished by
2937 remapping user id (UID) and group id (GID) of the root user to
2938 a UID and GID specified by the system administartor via Lustre
2939 configuration management server (MGS). The functionality also
2940 allows to specify sets of clients for which the remapping does
2945 Description: Excessive recovery window
2946 Details : With AT enabled, the recovery window can be excessively long (6000+
2947 seconds). To address this problem, we no longer use
2948 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2949 timeout no longer depends on the service time, it is set to
2950 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2951 time via pb_service_time.
2955 Description: Watchdog triggered on MDS failover
2956 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2957 the OSTs know that the MDS "UUID" can be reused for the same export
2958 from a different NID, so we do not need to wait for the export to be
2962 Frequency : rare, only if using MMP with Linux RAID
2964 Description: MMP doesn't work with Linux RAID
2965 Details : While using HA for Lustre servers with Linux RAID, it is possible
2966 that MMP will not detect multiple mounts. To make this work we
2967 need to unplug the device queue in RAID when the MMP block is being
2968 written. Also while reading the MMP block, we should read it from
2969 disk and not the cached one.
2971 Severity : enhancement
2973 Description: open file using fid
2974 Details : A file can be opened using just its fid, like
2975 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2978 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2981 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2982 Details : add proper locking for creation of jbd_4k slab cache
2986 Description: MMP check in ext3_remount() fails without displaying any error
2987 Details : When multiple mount protection fails during remount, proper error
2990 Severity : enhancement
2992 Description: Allow stripe size to be up to 4G-64k
2993 Details : Fix math logic to allow large stripe sizes.
2997 Description: add check for >8TB ldiskfs filesystems
2998 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2999 Don't allow >8TB ldiskfs filesystems to be mounted without
3000 force_over_8tb mount option
3005 Descriptoin: OSS complains "too many blocks" and crashes
3006 Details : In filter_commitrw_write():
3007 1. lquota_chkquota();
3008 2. obdo_from_inode();
3009 3. lquota_pending_commit();
3010 "2." has changed uid and gid in this case, then caused
3011 lquota_chkquota() and lquota_pending_commit() to use
3012 different uid and gid. That is the root of the bug.
3016 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3017 Details : Detect and print out the 'stripe count = -1' case properly.
3021 Description: Out or order replies might be lost on replay
3022 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3023 request with tid smaller than one currently being added, add it
3024 to the start, not end of the list.
3026 --------------------------------------------------------------------------------
3028 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3030 * Support for kernels:
3031 2.6.5-7.283 (SLES 9),
3032 2.6.9-55.EL (RHEL 4),
3033 2.6.16.46-0.14 (SLES 10),
3034 2.6.18.8 vanilla (kernel.org)
3035 * Client support for unpatched kernels:
3036 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3037 2.6.16 - 2.6.22 vanilla (kernel.org)
3038 * Due to recently discovered recovery problems, we do not recommend
3039 using patchless RHEL 4 clients with this or any earlier release.
3040 * Recommended e2fsprogs version: 1.39.cfs8
3041 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3042 * Starting with this release, the ldiskfs backing filesystem required
3043 by Lustre is now in its own package, lustre-ldiskfs. This package
3044 should be installed. It is versioned separately from Lustre and
3045 may be released separately in future.
3049 Description: block reactivating mgc import until all deactivates complete
3050 Details : Fix race when failing back MDT/MGS to itself (testing)
3053 Frequency : at statup only
3055 Description: mds_lov_synchronize race leads to various problems
3056 Details : simultaneous MDT->OST connections at startup can cause the
3057 sync to abort, leaving the OSC in a bad state.
3059 Severity : enhancement
3061 Description: add optional extra BUILD_VERSION info
3062 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3063 to override the lustre version.
3066 Frequency : 2.6.18 servers only
3068 Description: ll_kern_mount() doesn't release the module reference
3069 Details : The ldiskfs module reference count never drops down to 0
3070 because ll_kern_mount() doesn't release the module reference.
3075 Description: server LBUG when using old ost_num_threads parameter
3076 Details : Accept the old ost_num_threads parameter but warn that it
3077 is deprecated, and fix an off-by-one error that caused an LBUG.
3082 Description: Transient SCSI error results in persistent IO issue
3083 Details : iobuf->dr_error is not reinitialized to 0 between two
3087 Frequency : sometimes when underlying device returns I/O errors
3089 Description: OSTs not going read-only during write failures
3090 Details : OSTs are not remounted read-only when the journal commit threads
3091 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3092 instead of the ext3 wrappers.
3096 Description: poor connect scaling with increasing client count
3097 Details : Don't run filter_grant_sanity_check for more than 100 exports
3098 to improve scaling for large numbers of clients.
3101 Frequency : SLES10 only
3103 Description: sanity-quota.sh quotacheck failed: rc = -22
3104 Details : Quotas cannot be enabled on SLES10.
3107 Frequency : liblustre clients only
3109 Description: getdirentries does not give error when run on compute nodes
3110 Details : getdirentries does not fail when the size specified as an argument
3111 is too small to contain at least one entry
3113 Severity : enhancement
3115 Description: Add LNET router traceability for debug purposes
3116 Details : If a checksum failure occurs with a router as part of the
3117 IO path, the NID of the last router that forwarded the bulk data
3118 is printed so it can be identified.
3123 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3124 Details : Due to a race condition, liblustre clients were occasionally
3125 evicted incorrectly.
3127 Severity : enhancement
3129 Description: lfs setstripe use optional parameters instead of postional
3132 Severity : enhancement
3134 Description: Nanosecond timestamp support for ldiskfs
3135 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3136 resolution timestamps. There is not yet support for this at
3137 the Lustre filesystem level.
3140 Frequency : during server recovery
3142 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3143 Details : request with rq_no_resend flag not awake l_wait_event if they get
3147 Frequency : nfs export on patchless client
3149 Description: connectathon hang when test nfs export over patchless client
3150 Details : Disconnected dentry cannot be found with lookup, so we do not need
3151 to unhash it or make it invalid
3154 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3155 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3156 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3161 Description: open req refcounting wrong on reconnect
3162 Details : If reconnect happened between getting open reply from server and
3163 call to mdc_set_replay_data in ll_file_open, we will schedule
3164 replay for unreferenced request that we are about to free.
3165 Subsequent close will crash in variety of ways.
3166 Check that request is still eligible for replay in
3167 mdc_set_replay_data().
3172 Description: disable writes to filesystem when reading health_check file
3173 Details : the default for reading the health_check proc file has changed
3174 to NOT do a journal transaction and write to disk, because this
3175 can cause reads of the /proc file to hang and block HA state
3176 checking on a healthy but otherwise heavily loaded system. It
3177 is possible to return to the previous behaviour during configure
3178 with --enable-health-write.
3180 Severity : enhancement
3182 Description: 64-bit inode version
3183 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3184 to the inode. This will be required for version-based recovery.
3189 Description: MDS fails to start if a duplicate client export is detected
3190 Details : in some rare cases it was possible for a client to connect to
3191 an MDS multiple times. Upon recovery the MDS would detect this
3192 and fail during startup. Handle this more gracefully.
3194 Severity : enhancement
3196 Description: Add -o localflock option to simulate old noflock
3198 Details : This will achieve local-only flock/fcntl locks
3204 Description: log_commit_thread vs filter_destroy race leads to crash
3205 Details : Take import reference before releasing llog record semaphore
3210 Description: Wrong request locking in request set processing
3211 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3212 imp_delayed_list, in this place should be used imp_lock.
3215 Frequency : when reconnection
3217 Description: Grant Leak when osc reconnect to OST
3218 Details : When osc reconnect ost, OST(filter) should check whether it
3219 should grant more space to client by comparing fed_grant and
3220 cl_avail_grant, and return the granted space to client instead
3221 of "new granted" space, because client will call osc_init_grant
3222 to update the client grant space info.
3225 Frequency : when client reconnect to OST
3227 Description: Grant Leak when osc do resend and replay bulk write
3228 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3229 bulk write request, because the grant info while be sync between
3230 OSC and OST when reconnect, and we should ignore the grant info
3231 these of resend/replay write req.
3236 Description: Grant space more than available left space sometimes.
3237 Details : When then OST is about to be full, if two bulk writing from
3238 different clients came to OST. Accord the avaliable space of the
3239 OST, the first req should be permitted, and the second one
3240 should be denied by ENOSPC. But if the seconde arrived before
3241 the first one is committed. The OST might wrongly permit second
3242 writing, which will cause grant space > available space.
3245 Frequency : when client is evicted
3247 Description: Grant might be wrongly erased when osc is evicted by OST
3248 Details : when the import is evicted by server, it will fork another
3249 thread ptlrpc_invalidate_import_thread to invalidate the
3250 import, where the grant will be set to 0. While the original
3251 thread will update the grant it got when connecting. So if
3252 the former happened latter, the grant will be wrongly errased
3253 because of this race.
3258 Description: Checking Stale with correct fid
3259 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3260 to check whether it is stale, because sometimes, we want the
3261 enqueue happened anyway, and op_data.fid2 will not be initialized.
3263 Severity : enhancement
3265 Description: update patchless client
3266 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3269 Frequency : only with 2.4 kernel
3271 Description: random memory corruption
3272 Details : size of struct ll_inode_info is to big for union inode.u and this
3273 can be cause of random memory corruption.
3278 Description: Memory leak in recovery
3279 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3280 It should also check obd_fail before fsfilt_start, otherwise if
3281 fsfilt_start return -EROFS,(failover mds during mds recovery).
3282 then the req will return with repmsg->transno = 0 and rc = EROFS.
3283 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3284 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3285 be freed no matter whether fsfilt_commit success or not.
3288 Frequency : only with huge count clients
3290 Description: Prevents from taking the superblock lock in llap_from_page for
3292 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3293 for llap_from_page call in ll_removepage prevents from taking the
3294 superblock lock for a soon died page.
3299 Description: Not check open intent error before release open handle
3300 Details : in some rare cases, the open intent error is not checked before
3301 release open handle, which may cause
3302 ASSERTION(open_req->rq_transno != 0), because it tries to release
3303 the failed open handle.
3308 Description: Set cat log bitmap only after create log success.
3309 Details : in some rare cases, the cat log bitmap is set too early. and it
3310 should be set only after create log success.
3314 Description: Accessing a block bevice can re-enable I/O when Lustre is
3315 tearing down a device.
3316 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3321 Description: service threads may hog cpus when there are a lot of requests
3322 Details : Insert cond_resched to give other threads a chance to use some CPU
3327 Description: the cat log was not initialized in recovery
3328 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3329 unlink log on mds will not be initialized until mds post
3330 recovery. And also in mds post recovery, the unlink log will
3331 initialization will be done asynchronausly, so there will be race
3332 between add unlink log and unlink log initialization.
3336 Description: brw_stats were being printed incorrectly
3337 Details : brw_stats were being printed as log2 but all of them were not
3338 recorded as log2. Also remove some code duplication arising from
3339 filter_tally_{read,write}.
3343 Frequency : rare, only in recovery.
3344 Description: ASSERTION(req->rq_type != LI_POISON) failed
3345 Details : imp_lock should be held while iterating over imp_sending_list for
3346 prevent destroy request after get timeout in ptlrpc_queue_wait.
3350 Description: replay-single.sh test 52 fails
3351 Details : A lock's skiplist need to be cleanup when it being unlinked
3352 from its resource list.
3356 Description: Short directio read returns full requested size rather than
3358 Details : Direct I/O operations should return actual amount of bytes
3359 transferred rather than requested size.
3361 Severity : enhancement
3363 Description: metadata RPC reduction (e.g. for rm performance)
3364 Details : decrease the amount of synchronous RPC between clients and servers
3365 by canceling conflicing lock before the operation on the client
3366 and packing thier handles into the main operation RPC to server.
3368 Severity : enhancement
3370 Description: Async OSC create to avoid the blocking unnecessarily.
3371 Details : If an OST has no remaining object, system will block on the
3372 creation when it needs to create a new object on this OST. Now,
3373 ways use pre-created objects when available, instead of blocking on
3374 an empty osc while others are not empty. If we must block, we
3375 block for the shortest possible period of time.
3379 Description: improve handling recoverable errors
3380 Details : if request processig with error which can be recoverable on server
3381 request should be resend, otherwise page released from cache and
3384 Severity : enhancement
3386 Description: refine locking for avoid write wrong info into lov_objid file
3387 Details : fix possible races with add new target and write/update data in
3390 --------------------------------------------------------------------------------
3392 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3397 Frequency : on some architectures
3399 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3400 Details : Uninitialized flags sometimes cause configuration commands to
3404 Frequency : patchless clients only
3406 Description: missing __iget() symbol export
3407 Details : The __iget() symbol export is missing. To avoid the need for
3408 this on patchless clients the deathrow inode reaper is turned
3409 off, and we depend on the VM to clean up old inodes. This
3410 dependency was during via the fix for bug 12181.
3412 --------------------------------------------------------------------------------
3414 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3416 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3417 INTEROPERATE with older versions automatically. In many cases a
3418 special upgrade step is needed. Please read the
3419 user documentation before upgrading any part of a live system.
3420 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3421 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3422 * WARNING: Lustre configuration and startup changes are required with
3423 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3427 Severity : enhancement
3429 Description: MountConf
3430 Details : Lustre configuration is now managed via mkfs and mount
3431 commands instead of lmc and lconf. New obd types (MGS, MGC)
3432 are added for dynamic configuration management. See
3433 https://mail.clusterfs.com/wikis/lustre/MountConf for
3436 Severity : enhancement
3438 Description: dynamic OST addition
3439 Details : OSTs can now be added to a live filesystem
3441 Severity : enhancement
3443 Description: startup order invariance
3444 Details : MDTs and OSTs can be started in any order. Clients only
3445 require the MDT to complete startup.
3447 Severity : enhancement
3449 Description: parallel, asynchronous orphan cleanup
3450 Details : orphan cleanup is now performed in separate threads for each
3451 OST, allowing parallel non-blocking operation.
3453 Severity : enhancement
3455 Description: optimized stripe assignment
3456 Details : stripe assignments are now made based on ost space available,
3457 ost previous usage, and OSS previous usage, in order to try
3458 to optimize storage space and networking resources.
3460 Severity : enhancement
3462 Description: Permanently set tunables
3463 Details : All writable /proc/fs/lustre tunables can now be permanently
3464 set on a per-server basis, at mkfs time or on a live system.
3466 Severity : enhancement
3468 Description: Lustre message v2
3469 Details : Add lustre message format v2.
3471 Severity : enhancement
3473 Description: client OST exclusion list
3474 Details : Clients can be started with a list of OSTs that should be
3475 declared "inactive" for known non-responsive OSTs.
3479 Description: SPEC SFS validation failure on NFS v2 over lustre.
3480 Details : Changes the blocksize for regular files to be 2x RPC size,
3481 and not depend on stripe size.
3483 Severity : enhancement
3485 Description: Multiple MD RPCs in flight.
3486 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3487 To avoid overly-overloading MDS, introduce a limit on number of
3488 MDS RPCs in flight for a single client and add /proc controls
3489 to adjust this limit.
3491 Severity : enhancement
3493 Description: client read/write statistics
3494 Details : Add client read/write call usage stats for performance
3495 analysis of user processes.
3496 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3497 file access. extents_stats shows chunk size distribution.
3498 extents_stats_per_process show chunk size distribution per
3501 Severity : enhancement
3503 Description: mds statistics
3504 Details : Add detailed mds operations statistics in
3505 /proc/fs/lustre/mds/*/stats.
3509 Description: Failure of copying files with lustre special EAs.
3510 Details : Client side always return success for setxattr call for lustre
3511 special xattr (currently only "trusted.lov").
3516 Description: Refcount LNET uuids
3517 Details : The global LNET uuid list grew linearly with every startup;
3518 refcount repeated list entries instead of always adding to
3521 Severity : enhancement
3523 Description: Dynamic service threads
3524 Details : Within a small range, start extra service threads
3525 automatically when the request queue builds up.
3528 Frequency : mixed-endian client/server environments
3530 Description: mixed-endian crashes
3531 Details : The new msg_v2 system had some failures in mixed-endian
3534 Severity : enhancement
3536 Description: Easy OST removal
3537 Details : OSTs can be permanently deactivated with e.g. 'lctl
3538 conf_param lustre-OST0001.osc.active=0'
3540 Severity : enhancement
3542 Description: MGS proc entries
3543 Details : Added basic proc entries for the MGS showing what filesystems
3546 Severity : enhancement
3548 Description: provide MGS failover
3549 Details : Added config lock reacquisition after MGS server failover.
3551 Severity : enhancement
3553 Description: add Linux 2.4 support
3554 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3558 Description: a large application tries to do I/O to the same resource and dies
3559 in the middle of it.
3560 Details : Check the req->rq_arrival time after the call to
3561 ost_brw_lock_get(), but before we do anything about
3562 processing it & sending the BULK transfer request. This
3563 should help move old stale pending locks off the queue as
3564 quickly as obd_timeout.
3567 Frequency : when an incorrect nid is specified during startup
3569 Description: ptlrpc connect to non-existant node causes kernel crash
3570 Details : LNET can't be re-entered from an event callback, which
3571 happened when we expire a message after the export has been
3572 cleaned up. Instead, hand the zombie cleanup off to another
3575 Severity : enhancement
3577 Description: plain/inodebits lock performance improvement
3578 Details : Grouping plain/inodebits in granted list by their request modes
3579 and bits policy, thus improving the performance of search through
3583 Frequency : only if OST filesystem is corrupted
3585 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3586 Details : for a short time RPCs with bulk IO are in the replay list,
3587 but replay of bulk IOs is unimplemented. If the OST filesystem
3588 is corrupted due to disk cache incoherency and then replay is
3589 started it is possible to trip an assertion. Avoid putting
3590 committed RPCs into the replay list at all to avoid this issue.
3593 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3595 Description: System hang on startup
3596 Details : This bug allowed the liblustre (e.g. catamount) client to
3597 return to the app before handling all startup RPCs. This
3598 could leave the node unresponsive to lustre network traffic
3599 and manifested as a server ptllnd timeout.
3601 Severity : enhancement
3603 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3604 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3605 boolean to control whether to print peer debug info when a
3606 client's RPC times out.
3609 Frequency : only for kernels with patches from Lustre below 1.4.3
3611 Description: Remove old rdonly API
3612 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3615 Frequency : only for devices with external journals
3617 Description: Set external device read-only also
3618 Details : During a commanded failover stop, we set the disk device
3619 read-only while the server shuts down. We now also set any
3620 external journal device read-only at the same time.
3623 Frequency : when upgrading from 1.4 while trying to change parameters
3625 Description: The wrong (new) MDC name was used when setting parameters for
3626 upgraded MDT's. Also allows changing of OSC (and MDC)
3627 parameters if --writeconf is specified at tunefs upgrade time.
3630 Frequency : when setting specific ost indicies
3632 Description: QOS code breaks on skipped indicies
3633 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3634 created with --index need not be sequential.
3636 Severity : enhancement
3638 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3639 Details : The uninit_groups feature works in conjunction with the kernel
3640 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3641 up the pass1 processing of e2fsck. This is a read-only feature
3642 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3643 mount filesystems that have had this feature enabled.
3645 Severity : enhancement
3647 Description: Improve multi-block allocation algorithm to avoid fragmentation
3648 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3649 allocation locality and avoid filesystem fragmentation.
3651 ------------------------------------------------------------------------------
3653 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3655 * Support for kernels:
3656 2.4.21-47.0.1.EL (RHEL 3)
3657 2.6.5-7.283 (SLES 9)
3658 2.6.9-42.0.10.EL (RHEL 4)
3659 2.6.12.6 vanilla (kernel.org)
3660 2.6.16.27-0.9 (SLES 10)
3661 * Recommended e2fsprogs version: 1.39.cfs5
3663 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3667 Frequency : occasional, depends on client load and configuration
3668 Bugzilla : 12181, 12203
3669 Description: data loss for recently-modified files
3671 Details : In some cases it is possible that recently written or created
3672 files may not be written to disk in a timely manner (this should
3673 normally be within 30s unless client IO load is very high).
3674 The problem appears as zero-length files or files that are a
3675 multiple of 1MB in size after a client crash or client eviction
3676 that are missing data at the end of the file.
3678 This problem is more likely to be hit on clients where files are
3679 repeatedly created and unlinked in the same directory, clients
3680 have a large amount of RAM, have many CPUs, the filesystem has
3681 many OSTs, the clients are rebooted frequently, and/or the files
3682 are not accessed by other nodes after being written.
3684 The presence of the problem can be detected by looking at
3685 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3686 smaller than the second (nr_unused) then dirty files will not
3687 be flushed automatically to disk. "sync; sleep 10" should be
3688 run several times on the node before unmounting it to update
3689 Lustre (this is also safe to run on nodes without this problem).
3691 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3692 that can cause this same problem, so customers using that kernel
3693 also need to update the kernel in addition to Lustre. In order
3694 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3696 It is normal that files written just before a client crash (less
3697 than 30s) may not yet have been flushed to disk, even for local
3701 Frequency : frequent on thin XT3 nodes
3703 Description: UUID collision on thin XT3 Linux nodes
3704 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3705 randomly, since we relied on an insufficiently-seeded PRNG.
3710 Description: OSS hangs after "All ost request buffers busy"
3711 Details : A deadlock between quota and journal operations caused OSS
3712 hangs after printing "All ost request buffers busy."
3715 Frequency : always on liblustre builds
3717 Description: Cleanup compiler warnings on liblustre
3720 Frequency : always on liblustre builds on XT3
3722 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3726 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3727 > 32000 subdirectories
3728 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3729 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3730 subdirectory count crosses 32000. This will aid e2fsck to
3731 correctly handle more than 32000 subdirectories.
3734 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3736 Description: System hang on startup
3737 Details : This bug allowed the liblustre (e.g. catamount) client to
3738 return to the app before handling all startup RPCs. This
3739 could leave the node unresponsive to lustre network traffic
3740 and manifested as a server ptllnd timeout.
3742 Severity : enhancement
3744 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3745 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3746 boolean to control whether to print peer debug info when a
3747 client's RPC times out.
3752 Description: make O_SYNC working on 2.6 kernels
3753 Details : 2.6 kernels use different method for mark pages for write,
3754 so need add a code to lustre for O_SYNC work.
3759 Description: Failure to close file and release space on NFS
3760 Details : Put inode details into lock acquired in ll_intent_file_open.
3761 Use mdc_intent_lock in ll_intent_open to properly
3762 detect all kind of errors unhandled by mdc_enqueue.
3767 Description: proc file read during shutdown sometimes raced obd removal,
3769 Details : Add lock to prevent obd access after proc file removal.
3772 Frequency : Only for files larger than 4GB on 32-bit clients.
3774 Description: improperly doing page alignment of locks
3775 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3776 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3781 Description: RPCs being resent when they shouldn't be
3782 Details : Some RPCs that should not be resent are being resent. This
3783 can cause inconsistencies in the RPC state machine. Do not
3784 resend such requests.
3787 Frequency : rare, only with NFS export
3789 Description: Crash on NFS re-export node
3790 Details : under very unusual load conditions an assertion is hit in
3791 ll_intent_file_open()
3794 Frequency : only if OST filesystem is corrupted
3796 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3797 Details : for a short time RPCs with bulk IO are in the replay list,
3798 but replay of bulk IOs is unimplemented. If the OST filesystem
3799 is corrupted due to disk cache incoherency and then replay is
3800 started it is possible to trip an assertion. Avoid putting
3801 committed RPCs into the replay list at all to avoid this issue.
3806 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3807 Details : Large single O_DIRECT read and write calls can fail to allocate
3808 a sufficiently large buffer to process the request. In case of
3809 allocation failure the allocation is retried with a smaller
3810 buffer and broken into smaller requests.
3812 Severity : enhancement
3814 Description: Add -o localflock option to simulate old noflock behaviour.
3815 Details : This will achieve local-only flock/fcntl locks coherentness.
3820 Description: versioning check is incomplete
3821 Details : Checking the version difference of client vs. server, report
3822 error if the gap is too big.
3827 Description: add support PG_writeback bit
3828 Details : add support for PG_writeback bit for Lustre, for more carefull
3829 work with page cache in 2.6 kernel. This also fix some deadlocks
3830 and remove hack for work O_SYNC with 2.6 kernel.
3832 Severity : enhancement
3834 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3835 Details : The uninit_groups feature works in conjunction with the kernel
3836 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3837 up the pass1 processing of e2fsck. This is a read-only feature
3838 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3839 mount filesystems that have had this feature enabled.
3841 Severity : enhancement
3843 Description: Improve multi-block allocation algorithm to avoid fragmentation
3844 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3845 allocation locality and avoid filesystem fragmentation.
3847 ------------------------------------------------------------------------------
3849 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3851 * Support for kernels:
3852 2.6.9-42.0.3.EL (RHEL 4)
3853 2.6.5-7.276 (SLES 9)
3854 2.4.21-47.0.1.EL (RHEL 3)
3855 2.6.12.6 vanilla (kernel.org)
3856 2.6.16.21-0.8 (SLES10)
3857 * Recommended e2fsprogs version: 1.39.cfs2-0
3859 * The backwards-compatible /proc/sys/portals symlink has been removed
3860 in this release. Before upgrading, please ensure that you change
3861 any configuration scripts or /etc/sysctl.conf files that access
3862 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3863 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3864 in advance of the upgrade on any system running Lustre 1.4.6 or
3865 newer, since /proc/sys/lnet was added in that version.
3866 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3870 Frequency : only when quota is used
3872 Description: avoid scanning export list for quota master
3873 Details : Change the algorithms to avoid scanning export list in order
3874 to improve the efficiency.
3877 Frequency : MDS failover only, very rarely
3879 Description: "went back in time" messages on mds failover
3880 Details : The greatest transno may be lost when the current operation
3881 finishes with an error (transno==0) and the client's last_rcvd
3882 record is over-written. Save the greatest transno in the
3883 mds_last_transno for this case.
3886 Frequency : always for specific kernels and striping counts
3888 Description: client may get "Matching packet too big" without ACL support
3889 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3890 when trying to access files in certain configurations. The
3891 clients should in fact be denied when mounting because they do
3892 not understand ACLs.
3895 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3897 Description: many clients connecting with IO in progress causes connect timeouts
3898 Details : Avoid synchronous journal commits to avoid delays caused by many
3899 clients connecting/disconnecting when bulk IO is in progress.
3900 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3901 OST_IO_PORTAL to avoid delays behind potentially many pending
3905 Frequency : occasionally with multiple writers to a single file
3907 Description: shared writes to file may result in wrong size reported by stat()
3908 Details : Allow growing of kms when extent lock is cancelled
3911 Frequency : always with random mmap IO to multi-striped file
3913 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3914 Details : Only if the hole is at the end of OST object so that kms is too
3915 small. Fix is to increase kms accordingly in ll_nopage.
3918 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3920 Description: writes to a missing object would leak memory on the OST
3921 Details : If there is an inconsistency between the MDS and OST filesystems,
3922 such that the MDS references an object that doesn't exist, writes
3923 to that object will leak memory due to incorrect cleanup in the
3924 error handling path, eventually running out of memory on the OST.
3929 Description: Creating too long symlink causes lustre errors
3930 Details : Check symlink and name lengths before sending requests to MDS.
3933 Frequency : only if flock is enabled (not on by default)
3935 Description: posix locks not released on fd closure on 2.6.9+
3936 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3937 kernels, this caused such locks not to be released on fd close and
3938 then assertions on fs unmount about still used locks.
3941 Frequency : MDS failover only, very rarely
3943 Description: clients may get ASSERTION(granted_lock != NULL)
3944 Details : When request was taking a long time, and a client was resending
3945 a getattr by name lock request. The were multiple lock requests
3946 with the same client lock handle and
3947 mds_getattr_name->fixup_handle_for_resent_request found one of the
3948 lock handles but later failed with ASSERTION(granted_lock != NULL).
3953 Description: handle->h_buffer_credits > 0, assertion failure
3954 Details : h_buffer_credits is zero after truncate, causing assertion
3955 failure. This patch extends the transaction or creates a new
3959 Frequency : NFS re-export or patchless client
3960 Bugzilla : 11179, 10796
3961 Description: Crash on NFS re-export node (__d_move)
3962 Details : We do not want to hash the dentry if we don't have a lock.
3963 But if this dentry is later used in d_move, we'd hit uninitialised
3964 list head d_hash, so we just do this to init d_hash field but
3965 leave dentry unhashed.
3968 Frequency : NFS re-export or patchless client
3970 Description: NFS exports has problem with symbolic link
3971 Details : lustre client didn't properly install dentry when re-exported
3972 to NFS or running patchless client.
3975 Frequency : NFS re-export or patchless client
3977 Description: Various nfs/patchless fixes.
3978 Details : fixes reuse disconected alias for lookup process - this fixes
3979 warning "find_exported_dentry: npd != pd",
3980 fix permission error with open files at nfs.
3981 fix apply umask when do revalidate.
3984 Frequency : occasional
3986 Description: Crash on NFS re-export node
3987 Details : calling clear_page() on the wrong pointer triggered oops in
3988 generic_mapping_read().
3991 Frequency : rarely, using O_DIRECT IO
3993 Description: unaligned directio crashes client with LASSERT
3994 Details : check for unaligned buffers before trying any requests.
3997 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3999 Description: stale data returned from RAID cache
4000 Details : If only a small amount of IO is done to the RAID device before
4001 reading it again it is possible to get stale data from the RAID
4002 cache instead of reading it from disk.
4005 Frequency : always for sles10 kernel
4007 Description: sles10 support
4008 Details : ll_follow_link: compile fixes and using of nd_set_link
4009 under newer kernels.
4012 Frequency : depends on arch, kernel and compiler version, always on sles10
4015 Description: recursive or deep enough symlinks cause stack overflow
4016 Details : getting rid of large stack-allocated variable in
4020 Frequency : depends on hardware
4022 Description: lustre write performance loss in the SLES10 kernel
4023 Details : the performance loss is caused by using of write barriers in the
4024 ext3 code. The SLES10 kernel turns barrier support on by
4025 default. The fix is to undo that change for ldiskfs.
4027 ------------------------------------------------------------------------------
4029 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4031 * Support for kernels:
4032 2.6.9-42.0.3EL (RHEL 4)
4033 2.6.5-7.276 (SLES 9)
4034 2.4.21-47.0.1.EL (RHEL 3)
4035 2.6.12.6 vanilla (kernel.org)
4039 Frequency : quota enabled and large files being deleted
4041 Description: releasing more than 4GB of quota at once hangs OST
4042 Details : If a user deletes more than 4GB of files on a single OST it
4043 will cause the OST to spin in an infinite loop. Release
4044 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4049 Description: statfs data retrieved from /proc may be stale or zero
4050 Details : When reading per-device statfs data from /proc, in the
4051 {kbytes,files}_{total,free,avail} files, it may appear
4052 as zero or be out of date.
4055 Frequency : systems with MD RAID1 external journal devices
4057 Description: lconf's call to blkid is confused by RAID1 journal devices
4058 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4059 of returning all block devices that match the journal UUID.
4062 Frequency : always, for aggregate stripe size over 4GB
4064 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4065 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4066 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4069 Frequency : always if "lfs find" used on a local file/directory
4071 Description: "lfs find" segfaults if used on a local file/directory
4072 Details : The case where a directory component was not specified wasn't
4073 handled correctly. Handle this properly.
4076 Frequency : always on ppc64
4078 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4079 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4080 perform endianness conversion.
4083 Frequency : rarely (truncate to non-zero file size after write under load)
4084 Bugzilla : 10730, 10687
4085 Description: Files padded with zeros to next 4K multiple
4086 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4087 it is possible that files that are truncated to a non-zero size
4088 immediately after being written are filled with zero bytes beyond
4089 the truncated size. No file data is lost.
4091 Severity : enhancement
4093 Description: Allow recovery/failover for liblustre clients.
4094 Details : liblustre clients were unaware of failover configurations until
4097 Severity : enhancement
4099 Description: user file locks should fail when not mounting with flock option
4100 Details : Set up an error-returning stub in ll_file_operations.lock field
4101 to prevent incorrect behaviour when client is mounted without
4102 flock option. Also, set up properly f_op->flock field for
4106 Frequency : always on ia64
4108 Description: "lfs df" loops on printing out MDS statfs information
4109 Details : The obd_ioctl_data was not initialized and in some systems
4110 this caused a failure during the ioctl that did not return
4111 an error. Initialize the struct and return an error on failure.
4114 Frequency : SLES 9 only
4116 Description: Error of copying files with lustre special EAs as root
4117 Details : Client side always return success for setxattr call for lustre
4118 special xattr (currently only "trusted.lov").
4121 Frequency : rarely on clusters with both ia64+i386 clients
4123 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4124 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4125 clients are doing concurrent writes to the same file it is
4126 possible that the ia64 clients may LASSERT because the OST
4127 extent locks are not PAGE_SIZE aligned. Ensure that grown
4128 locks are always aligned on the request boundary.
4131 Frequency : specific use, occasional
4133 Description: Overwriting in use executable truncates on-disk binary image
4134 Details : If one node attempts to overwrite an executable in use by
4135 another node, we now correctly return ETXTBSY instead of
4136 truncating the file.
4138 Severity : enhancement
4140 Description: Async OSC create to avoid the blocking unnecessarily.
4141 Details : If an OST has no remaining object, system will block on the creating
4142 when need to create a new object on this OST. Now, ways use
4143 pre-created objects when available, instead of blocking on an
4144 empty osc while others are not empty. If we must block, we block
4145 for the shortest possible period of time.
4150 Description: chmod on Lustre root is propagated to other clients
4151 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4152 invalid by the follow_mount time.
4155 Frequency : liblustre clients only
4157 Description: Race in 'instant cancel' lock handling could lead to such locks
4158 never to be granted in case of SMP MDS
4159 Details : Do not destroy not yet granted but cbpending locks in
4163 Frequency : replay/resend of open
4165 Description: non null lock assetion failure in mds_intent_policy
4166 Details : Trying to replay/resend lockless open requests resulted in
4167 mds_open() returning 0 with no lock. Now it sets a flag if
4168 a lock is going to be returned.
4170 Severity : enhancement
4172 Description: Checksum enhancements
4173 Details : New checksum enhancements allow for resending RPCs that failed
4176 Severity : enhancement
4178 Description: Tunables on number of dirty pages in cacche
4179 Details : Allow to set limit on number of dirty pages cached.
4184 Description: client crash on unmount - lock still has references
4185 Details : In some error handling cases it was possible to leak a lock
4186 reference on a client while accessing a file. This was not
4187 harmful to the client during operation, but would cause the
4188 client to crash when the filesystem is unmounted.
4191 Frequency : specific case, rare
4193 Description: ETXTBSY on mds though file not in use by client
4194 Details : ETXTBSY is no longer incorrectly returned when attempting to
4195 chmod or chown a directory that the user previously tried to
4196 execute or a currently-executing binary.
4199 Frequency : extremely rare except on liblustre-based clients
4201 Description: Lustre space not freed when files are deleted
4202 Details : Clean up open-unlinked files after client eviction. Previously
4203 the unlink was skipped and the files remained as orphans.
4208 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4209 Details : In some cases it was possible to send a blocking callback to
4210 a client doing a glimpse, even though that client didn't get
4211 a lock granted. When the glimpse lock is cancelled on the OST
4212 the freed lock is left on the waiting list and corrupted the list.
4215 Frequency : all core dumps
4217 Description: Broke core dumps to lustre
4218 Details : Negative dentry may be unhashed if parent does not have UPDATE
4219 lock, but some callers, e.g. do_coredump, expect dentry to be
4220 hashed after successful create, hash it in ll_create_it.
4222 ------------------------------------------------------------------------------
4224 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4226 * Support for kernels:
4227 2.6.9-42.0.2.EL (RHEL 4)
4228 2.6.5-7.276 (SLES 9)
4229 2.4.21-40.EL (RHEL 3)
4230 2.6.12.6 vanilla (kernel.org)
4234 Frequency : always on RHEL 3
4236 Description: Number of open files grows over time
4237 Details : The number of open files grows over time, whether or not
4238 Lustre is started. This was due to a filp leak introduced
4239 by one of our kernel patches.
4241 ------------------------------------------------------------------------------
4243 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4245 * Support for kernels:
4246 2.6.9-42.EL (RHEL 4)
4247 2.6.5-7.276 (SLES 9)
4248 2.4.21-40.EL (RHEL 3)
4249 2.6.12.6 vanilla (kernel.org)
4254 Bugzilla : 5719, 9635, 9792, 9684
4255 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4256 Details : If a server is under heavy load and cannot reply to new
4257 connection requests before the client resends the (re)connect,
4258 the connection handling code can behave badly if two service
4259 threads are concurrently handing separate (re)connections from
4260 the same client. Add better locking to the connection handling
4261 code, and ensure that only a single connection will be processed
4262 for a given client UUID, even if the lock is dropped.
4264 Severity : enhancement
4266 Description: add TCP zero-copy support to kernel
4267 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4268 sends if the hardware supports scatter-gather and checksumming.
4269 This allows socklnd to do client-write and server-read more
4270 efficiently and reduce CPU utilization from skbuf copying.
4273 Frequency : only if NFS exporting from client
4275 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4276 Details : If setting attributes on a file created under NFS that had
4277 never been opened it would be possible to oops the client
4278 if the file had no objects.
4281 Frequency : always for liblustre
4283 Description: liblustre client does MDS+OSTs setattr RPC for each write
4284 Details : When doing a write from a liblustre client, the client
4285 incorrectly issued an RPC to the MDS and each OST the file was
4286 striped over in order to update the timestamps. When writing
4287 with small chunks and many clients this could overwhelm the MDS
4288 with RPCs. In all cases it would slow down the write because
4289 these RPCs are unnecessary.
4291 Severity : enhancement
4293 Description: allow number of MDS service threads to be changed at module load
4294 Details : It is now possible to change the number of MDS service threads
4295 running. Adding "options mds mds_num_threads={N}" to the MDS's
4296 /etc/modprobe.conf will set the number of threads for the next
4297 time Lustre is restarted (assuming the "mds" module is also
4298 reloaded at that time). The default number of threads will
4299 stay the same, 32 for most systems.
4304 Description: OST crash if filesystem is unformatted or corrupt
4305 Details : If an OST is started on a device that has never been formatted
4306 or if the filesystem is corrupt and cannot even mount then the
4307 error handling cleanup routines would dereference a NULL pointer.
4312 Description: NULL pointer deref in llap_from_page.
4313 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4314 incorrect) mapping if the page was truncated/reclaimed while it was
4315 searched for. Check for this condition and skip such pages when
4316 doing readahead. Introduce extra check to llap_from_page() to
4317 verify page->mapping->host is non-NULL (so page is not anonymous).
4320 Frequency : Sometimes when using sys_sendfile
4322 Description: "page not covered by a lock" warnings from ll_readpage
4323 Details : sendfile called ll_readpage without right page locks present.
4324 Now we introduced ll_file_sendfile that does necessary locking
4325 around call to generic_file_sendfile() much like we do in
4329 Frequency : with certain MDS communication failures at client mount time
4331 Description: NULL pointer deref after failed client mount
4332 Details : a client connection request may delayed by the network layer
4333 and not be sent until after the PTLRPC layer has timed out the
4334 request. If the client fails the mount immediately it will try
4335 to clean up before the network times out the request. Add a
4336 reference from the request import to the obd device and delay
4337 the cleanup until the network drops the request.
4340 Frequency : occasionally during client (re)connect
4342 Description: assertion failure during client (re)connect
4343 Details : processing a client connection request may be delayed by the
4344 client or server longer than the client connect timeout. This
4345 causes the client to resend the connection request. If the
4346 original connection request is replied in this interval, the
4347 client may trip an assertion failure in ptlrpc_connect_interpret()
4348 which thought it would be the only running connect process.
4351 Frequency : only with obd_echo servers and clients that are rebooted
4353 Description: kernel BUG accessing uninitialized data structure
4354 Details : When running an obd_echo server it did not start the ping_evictor
4355 thread, and when a client was evicted an uninitialized data
4356 structure was accessed. Start the ping_evictor in the RPC
4357 service startup instead of the OBD startup.
4359 Severity : enhancement
4360 Bugzilla : 10193 (patchless)
4361 Description: Remove dependency on various unexported kernel interfaces.
4362 Details : No longer need reparent_to_init, exit_mm, exit_files,
4363 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4366 Frequency : rare (only users of deprecated and unsupported LDAP config)
4368 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4369 Details : LDAP apparently contains 'lustreName' attributes instead of
4370 'name'. A simple remapping of the name is sufficient.
4373 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4375 Description: waiting_locks_callback trips kernel BUG if client is evicted
4376 Details : Running with the dump_on_timeout debug flag turned on makes
4377 it possible that the waiting_locks_callback() can try to dump
4378 the Lustre kernel debug logs from an interrupt handler. Defer
4379 this log dumping to the expired_lock_main() thread.
4381 Severity : enhancement
4383 Description: Support NFS exporting on 2.6 kernels.
4384 Details : Implement non-rawops metadata methods for NFS server to use without
4385 changing NFS server code.
4388 Frequency : very rare (synthetic metadata workload only)
4390 Description: two racing renames might cause an MDS thread to deadlock
4391 Details : Running the "racer" program may cause one MDS thread to rename
4392 a file from being the source of a rename to being the target of
4393 a rename at exactly the same time that another thread is doing
4394 so, and the second thread has already enqueued these locks after
4395 doing a lookup of the target and is trying to relock them in
4396 order. Ensure that we don't try to re-lock the same resource.
4399 Frequency : only very large systems with liblustre clients
4401 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4402 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4403 all OSTs in parallel. This allows the eviction of stale liblustre
4404 clients to proceed much faster than if they were done in series,
4405 and also offers similar improvements for other set_info RPCs.
4410 Description: excessive CPU usage during initial read phase on client
4411 Details : During the initial read phase on a client, it would agressively
4412 retry readahead on the file, consuming too much CPU and impacting
4413 performance (since 1.4.5.8). Improve the readahead algorithm
4414 to avoid this, and also improve some other common cases (read
4415 of small files in particular, where "small" is files smaller than
4416 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4421 Description: MDS crash when receiving packet with unknown intent.
4422 Details : Do not LBUG in unknown intent case, just return -EFAULT
4424 Severity : enhancement
4425 Bugzilla : 9293, 9385
4426 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4427 Details : Do not serialize getattr (non-intent version) and statfs.
4430 Frequency : occasional, when OST network is overloaded/intermittent
4432 Description: client evicted by OST after bulk IO timeout
4433 Details : If a client sends a bulk IO request (read or write) the OST
4434 may evict the client if it is unresposive to its data GET/PUT
4435 request. This is incorrect if the network is overloaded (takes
4436 too long to transfer the RPC data) or dropped the OST GET/PUT
4437 request. There is no need to evict the client at all, since
4438 the pinger and/or lock callbacks will handle this, and the
4439 client can restart the bulk request.
4442 Frequency : Always when mmapping file with no objects
4444 Description: client crashes when mmapping file with no objects
4445 Details : Check that we actually have objects in a file before doing any
4446 operations on objects in ll_vm_open, ll_vm_close and
4452 Description: Request leak when working with deleted CWD
4453 Details : Introduce advanced request refcount tracking for requests
4454 referenced from lustre intent.
4456 Severity : Enhancement
4458 Description: Cache open file handles on client.
4459 Details : MDS now will return special lock along with openhandle, if
4460 requested and client is allowed to hold openhandle, even if unused,
4461 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4462 closing files for every read/write openration.
4464 Severity : Enhancement
4466 Description: Cache open negative dentries on client when possible.
4467 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4468 negative dentries on lock revocation.
4473 Description: Remounting a client read-only wasn't possible with a zconf mount
4474 Details : It wasn't possible to remount a client read-only with llmount.
4476 Severity : enhancement
4477 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4478 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4479 MPICH is included to add Lustre-specific ADIO interfaces.
4480 This is based closely on the UFS ADIO layer and only differs
4481 in file creation, in order to allow the OST striping to be set.
4482 This is user-contributed code and not supported by CFS.
4487 Description: extended inode attributes (immutable, append-only) work improperly
4488 when 2.4 and 2.6 kernels are used on client/server or vice versa
4489 Details : Introduce kernel-independent values for these flags.
4491 Severity : enhancement
4494 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4495 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4496 granularity. Now, Fractional MB granularity is be supported,
4497 this is very useful for low memory system.
4499 Severity : enhancement
4501 Description: Getattr by fid
4502 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4503 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4504 have locks and avoids getattr rpc requests for it on every stat.
4507 Frequency : Always, for filesystems larger than 2TB
4509 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4510 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4511 filesystems larger than 2TB to be mounted with mballoc enabled.
4514 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4516 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4517 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4518 filesystems larger than 2TB is mounted then the kernel will
4519 silently corrupt the start of the filesystem. CONFIG_LBD is
4520 enabled for all CFS-supported kernels, but the possibility of
4521 this happening with a modified kernel config exists.
4523 Severity : enhancement
4525 Description: add client O_DIRECT support for 2.6 kernels
4526 Details : It is now possible to do O_DIRECT reads and writes to files
4527 in the Lustre client mountpoint on 2.6 kernel clients.
4529 Severity : enhancement
4531 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4532 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4533 parallel, not waiting for response from every OST before sending
4534 a rpc to the next OST.
4539 Description: setattr vs write race when updating file timestamps
4540 Details : Client processes that update a file timestamp into the past
4541 right after writing to the file (e.g. tar) it is possible that
4542 the updated file modification time can be reset to the current
4543 time due to a race between processing the setattr and write RPC.
4545 Severity : enhancement
4547 Description: Bring 'lfs find' closer in line with regular Linux find.
4548 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4549 -print0 options and obtains all the needed info through the lustre
4552 Severity : enhancement
4554 Description: support up to 1024 configured devices on one node
4555 Details : change obd_dev array from statically allocated to dynamically
4556 allocated structs as they are first used to reduce memory usage
4561 Description: Flush dirty partially truncated pages during truncate
4562 Details : Immediatelly flush partially truncated pages in filter_setattr,
4563 this way we completely avoid having any pages in page cache on OST
4564 and can retire ugly workarounds during writes to flush such pages.
4569 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4570 Details : protect inode from truncation within vfs_unlink() context
4571 just take a reference before calling vfs_unlink() and release it
4572 when parent's i_sem is free.
4577 Description: last_id value checked outside lock on OST caused LASSERT failure
4578 Details : If there were multiple MDS->OST object precreate requests in
4579 flight, it was possible that the OST's last object id was checked
4580 outside a lock and incorrectly tripped an assertion. Move checks
4581 inside locks, and discard old precreate requests.
4584 Frequency : always, if extents are used on OSTs
4586 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4587 Details : OSTs using the extents format would not zero the high 16 bits of
4588 the index physical block number. This is not a problem for any
4589 OST filesystems smaller than 16TB, and no kernels support ext3
4590 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4591 new/modified files) and can be fixed for existing filesystems
4592 with e2fsprogs-1.39-cfs1.
4597 Description: import connection selection may be incorrect if timer wraps
4598 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4599 import connections to be ignored if the 32-bit jiffies counter
4600 wraps. Use a 64-bit jiffies counter.
4603 Frequency : very large clusters immediately after boot
4605 Description: LNET request buffers exhausted under heavy short-term load
4606 Details : If a large number of client requests are generated on a service
4607 that has previously never seen so many requests it is possible
4608 that the request buffer growth cannot keep up with the spike in
4609 demand. Instead of dropping incoming requests, they are held in
4610 the LND until the RPC service can accept more requests.
4613 Frequency : Sometimes during replay
4615 Description: Assertion failure in ll_local_open after replay.
4616 Details : If replay happened on an open request reply before we were able
4617 to set replay handler, reply will become not swabbed tripping the
4618 assertion in ll_local_open. Now we set the handler right after
4619 recognising of open request
4622 Frequency : very rare
4624 Description: kernel reports "badness in vsnprintf"
4625 Details : Reading from the "recovery_status" /proc file in small chunks
4626 may cause a negative length in lprocfs_obd_rd_recovery_status()
4627 call to vsnprintf() (which is otherwise harmless). Exit early
4628 if there is no more space in the output buffer.
4630 Severity : enhancement
4632 Description: clear OBD RPC statistics by writing to them
4633 Details : It is now possible to clear the OBD RPC statistics by writing
4634 to the "stats" file.
4639 Description: Client mtime is not the same on different clients after utimes
4640 Details : In some cases, the client was using the utimes() syscall on
4641 a file cached on another node. The clients now validate the
4642 ctime from the MDS + OSTs to determine which one is right.
4647 Description: Inability to activate failout mode
4648 Details : lconf script incorrectly assumed that in pythong string's numeric
4649 value is used in comparisons.
4652 Frequency : always with multiple stripes per file
4654 Description: Inefficient object allocation for mutli-stripe files
4655 Details : When selecting which OSTs to stripe files over, for files with
4656 a stripe count that divides evenly into the number of OSTs,
4657 the MDS is always picking the same starting OST for each file.
4658 Return the OST selection heuristic to the original design.
4663 Description: mount failures may take full timeout to return an error
4664 Details : Under some heavy load conditions it is possible that a
4665 failed mount can wait for the full obd_timeout interval,
4666 possibly several minutes, before reporting an error.
4667 Instead return an error as soon as the status is known.
4669 Frequency : quota enabled and large files being deleted
4671 Description: releasing more than 4GB of quota at once hangs OST
4672 Details : If a user deletes more than 4GB of files on a single OST it
4673 will cause the OST to spin in an infinite loop. Release
4674 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4679 Description: statfs data retrieved from /proc may be stale or zero
4680 Details : When reading per-device statfs data from /proc, in the
4681 {kbytes,files}_{total,free,avail} files, it may appear
4682 as zero or be out of date.
4685 Frequency : systems with MD RAID1 external journal devices
4687 Description: lconf's call to blkid is confused by RAID1 journal devices
4688 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4689 of returning all block devices that match the journal UUID.
4692 Frequency : always, for aggregate stripe size over 4GB
4694 Description: assertion fails when trying to use 4GB stripe size
4695 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4696 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4699 Frequency : always on ppc64
4701 Description: the first write on an ext3 filesystem with mballoc got stuck
4702 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4703 perform endianness conversion.
4705 ------------------------------------------------------------------------------
4707 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4709 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4710 INTEROPERATE with older versions automatically. Please read the
4711 user documentation before upgrading any part of a live system.
4712 * WARNING: Lustre networking configuration changes are required with
4713 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4716 * Support for kernels:
4717 2.6.9-22.0.2.EL (RHEL 4)
4718 2.6.5-7.244 (SLES 9)
4719 2.6.12.6 vanilla (kernel.org)
4722 Severity : enhancement
4723 Bugzilla : 7981/8208
4724 Description: Introduced Lustre Networking (LNET)
4725 Details : LNET is new networking infrastructure for Lustre, it includes
4726 a reorganized network configuration mode (see the user
4727 documentation for full details) as well as support for routing
4728 between different network fabrics. Lustre Networking Devices
4729 (LNDs) for the supported network fabrics have also been
4730 created for this new infrastructure.
4732 Severity : enhancement
4733 Description: Introduced Access control lists
4734 Details : clients can set ACLs on files and directories in order to have
4735 more fine-grained permissions than the standard Unix UGO+RWX.
4736 The MDS must be started with the "-o acl" mount option.
4738 Severity : enhancement
4739 Description: Introduced filesystem quotas
4740 Details : Administrators may now establish per-user quotas on the
4743 Severity : enhancement
4745 Description: Configuration change for the XT3
4746 The PTLLND is now used to run Lustre over Portals on the XT3
4747 The configure option(s) --with-cray-portals are no longer used.
4748 Rather --with-portals=<path-to-portals-includes> is used to
4749 enable building on the XT3. In addition to enable XT3 specific
4750 features the option --enable-cray-xt3 must be used.
4755 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4756 Details : A race between allocating a new llcd and re-getting the llcd_lock
4757 allowed another thread to grab newly-allocated llcd.
4759 Severity : enhancement
4761 Description: 2.6 OST async journal commit and locking fix to improve performance
4762 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4763 2.6 kernels are now async as they already were in 2.4 kernels so
4764 that they can commit concurrently with the network bulk transfer.
4765 For block-allocated files the filter allocation semaphore is held
4766 to avoid filesystem fragmentation during allocation. BKL lock
4767 removed for 2.6 xattr operations where it is no longer needed.
4772 Description: lconf incorrectly determined whether two IP networks could talk
4773 Details : In some more complicated routing and multiple-network
4774 configurations, lconf will avoid trying to make a network
4775 connection to a disjoint part of the IP space. It was doing the
4776 math incorrectly for one set of cases.
4781 Description: Fix for potential infinite loop processing records in an llog.
4782 Details : If an llog record is corrupted/zeroed, it is possible to loop
4783 forever in llog_process(). Validate the llog record length
4784 and skip the remainder of the block on error.
4787 Frequency : occasional (liblustre only)
4789 Description: liblustre could not open files whose last component is a symlink
4790 Details : sysio_path_walk() would incorrectly pass the open intent to
4791 intermediate path components.
4794 Frequency : rare (liblustre only with non-standard tuning)
4795 Bugzilla : 7201 (7350)
4796 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4797 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4798 after it has been released, only do this for referenced locks
4800 Severity : enhancement
4802 Description: specifying an (invalid) directory default stripe_size of -1
4803 would reset the directory default striping
4804 Details : stripe_size -1 was used internally to signal directory stripe
4805 removal, now use "all default" to signal dir stripe removal
4806 as a directory striping of "all default" is not useful
4809 Frequency : common for large clusters running liblustre clients
4811 Description: doing an ls when liblustre clients are running is slow
4812 Details : sending a glimpse AST to a liblustre client waits for every AST
4813 to time out, as liblustre clients will not respond. Since they
4814 cannot cache data we refresh the OST lock LVB from disk instead.
4816 Severity : enhancement
4818 Description: doing an ls at the same time as file IO can be slow
4819 Details : enqueue and other "small" requests can be blocked behind many
4820 large IO requests. Create a new OST IO portal for non-IO
4821 requests so they can be processed faster.
4824 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4826 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4827 created file with mode 000
4828 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4829 allow an ftruncate() as a non-root user to a file with mode 000.
4830 Lustre now respects this flag to disable mode checks when
4831 truncating a file owned by the user
4834 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4837 Description: Revoking locks from clients that went dead or catatonic might take
4839 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4840 cancellation of such locks instant on servers without waiting for
4841 any reply from clients. Clients drops these locks when cancel
4842 notification from server is received without replying.
4845 Frequency : liblustre-only, when liblustre client dies or becomes busy
4847 Description: Doing ls on Linux clients can take a long time with active
4849 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4850 granting such locks to it in the first place if possible. Locks
4851 are taken by proxy on the OST during the read or write and
4852 dropped immediately afterward. Add connect flags handling, do
4853 not grant locks to liblustre clients for glimpse ASTs.
4855 Severity : enhancement
4857 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4858 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4859 sequential reads and behaves badly with applications doing random
4860 reads. Improve it by reading ahead at least read region, and
4861 avoiding excessive large RPC for small reads.
4863 Severity : enhancement
4865 Description: Creating more than 1000 files for a single job may cause a load
4866 imbalance on the OSTs if there are also a large number of OSTs.
4867 Details : qos_prep_create() uses an OST index reseed value that is an
4868 even multiple of the number of available OSTs so that if the
4869 reseed happens in the middle of the object allocation it will
4870 still utilize the OSTs as uniformly as possible.
4875 Description: OST or MDS may oops in ping_evictor_main()
4876 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4877 but doesn't restart at beginning of obd_exports_timed list
4880 Severity : enhancement
4882 Description: improve by-nid export eviction on the MDS and OST
4883 Details : allow multiple exports with the same NID to be evicted at one
4884 time without re-searching the exports list.
4887 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4889 Description: MDS may oops in groups_free()
4890 Details : in rare race conditions a newly allocated group_info struct is
4891 freed again, and this can be NULL. The 2.4 compatibility code
4892 for groups_free() checked for a NULL pointer, but 2.6 did not.
4895 Frequency : common for liblustre clients doing little filesystem IO
4896 Bugzilla : 9352, 7313
4897 Description: server may evict liblustre clients accessing contended locks
4898 Details : if a client is granted a lock or receives a completion AST
4899 with a blocking AST already set it would not reply to the AST
4900 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4903 Frequency : lfs setstripe, only systems with more than 160 OSTs
4905 Description: unable to set striping with a starting offset beyond OST 160
4906 Details : llapi_create_file() incorrectly limited the starting stripe
4907 index to the maximum single-file stripe count.
4910 Frequency : LDAP users only
4912 Description: lconf did not handle in-kernel recovery with LDAP properly
4913 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4915 Severity : enhancement
4917 Description: bind OST threads to NUMA nodes to improve performance
4918 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4919 node and do their allocations there to localize memory access
4921 Severity : enhancement
4923 Description: llmount can determine client NID directly from Myrinet (GM)
4924 Details : the client NID code from gmnalnid was moved directly into
4925 llmount, removing the need to use this or specifying the
4926 client NID explicitly when mounting GM clients with zeroconf
4929 Frequency : if client is started with down MDS
4931 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4932 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4933 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4934 ptlrpc_import_delayed_req() because we didn't check if the
4935 request was interrupted, and we also didn't break out of the
4936 event loop if there was no timeout
4941 Description: data loss during non-page-aligned writes to a single file from
4942 both multiple nodes and multiple threads on one node at same time
4943 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4944 inconsistency led to false short-reads, that were cached and later
4945 used by ->prepare_write() to fill in partially written page,
4946 leading to data loss.
4949 Frequency : always, if lconf --abort_recovery used
4951 Description: lconf --abort_recovery fails with 'Operation not supported'
4952 Details : lconf was attempting to abort recovery on the MDT device and not
4955 Severity : enhancement
4957 Description: remove cleanup logs
4958 Details : replace lconf-generated cleanup logs with lustre internal
4959 cleanup routines. Eliminates the need for client-cleanup and
4962 Severity : enhancement
4964 Description: add support for EAs (user and system) on lustre filesystems
4965 Details : it is now possible to store extended attributes in the Lustre
4966 client filesystem, and with the user_xattr mount option it
4967 is possible to allow users to store EAs on their files also
4969 Severity : enhancement
4971 Description: Add possibility (config option) to show minimal available OST free
4973 Details : When compiled with --enable-mindf configure option, statfs(2)
4974 (and so, df) will return least minimal free space available from
4975 all OSTs as amount of free space on FS, instead of summary of
4976 free spaces of all OSTs.
4978 Severity : enhancement
4980 Description: do not expand extent locks acquired on OST-side
4981 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4982 by server: they are not cached anyway.
4985 Frequency : when mmap is used/binaries executed from Lustre
4987 Description: Unmmap pages before throwing them away from read cache.
4988 Details : llap_shrink cache now attempts to unmap pages before discarding
4989 them (if unmapping failed - do not discard). SLES9 kernel has
4990 extra checks that trigger if this unmapping is not done first.
4995 Description: lconf didn't resolve symlinks before checking to see whether a
4996 given mountpoint was already in use
4999 Frequency : when migrating failover services
5000 Bugzilla : 6395, 9514
5001 Description: When migrating a subset of services from a node (e.g. failback
5002 from a failover service node) the remaining services would
5003 time out and evict clients.
5004 Details : lconf --force (implied by --failover) sets the global obd_timeout
5005 to 5 seconds in order to quickly disconnect, but this caused
5006 other RPCs to time out too quickly. Do not change the global
5007 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5009 Severity : enhancement
5010 Frequency : if MDS is started with down OST
5011 Bugzilla : 9439,5706
5012 Description: Allow startup/shutdown of an MDS without depending on the
5013 availability of the OSTs.
5014 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5015 Add appropriate locking and lov-osc refcounts for safe
5016 cleaning. Add osc abort_inflight calls in case the
5017 synchronize never started.
5020 Frequency : occasional (Cray XT3 only)
5022 Description: root not authorized to access files in CRAY_PORTALS environment
5023 Details : The client process capabilities were not honoured on the MDS in
5024 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5025 been accessed by an authorized user then root was able to access
5026 the file on the local client also. The root user capabilities
5027 are now allowed on the MDS, as this environment has secure UID.
5030 Frequency : occasional
5032 Description: ldiskfs "too long searching" message happens too often
5033 Details : A debugging message (otherwise harmless) prints too often on
5034 the OST console. This has been reduced to only happen when
5035 there are fragmentation problems on the filesystem.
5040 Description: Division by zero in statfs when all OSCs are inactive
5041 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5042 lov_statfs divides by value returned by lov_get_stripecnt().
5046 Bugzilla : 9489, 3273
5047 Description: First write from each client to each OST was only 4kB in size,
5048 to initialize client writeback cache, which caused sub-optimal
5049 RPCs and poor layout on disk for the first written file.
5050 Details : Clients now request an initial cache grant at (re)connect time
5051 and so that they can start streaming writes to the cache right
5052 away and always do full-sized RPCs if there is enough data.
5053 If the OST is rebooted the client also re-establishes its grant
5054 so that client cached writes will be honoured under the grant.
5059 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5060 Details : Now I/O RPCs go to different portal number and (presumably) fast
5061 lock requests (and glimses) and other RPCs get their own service
5062 threads pool that should be able to service those RPCs
5065 Severity : enhancement
5067 Description: Ability to exchange lustre version between client and servers and
5068 issue warnings at client side if client is too old. Also for
5069 liblustre clients there is ability to refuse connection of too old
5071 Details : New 'version' field is added to connect data structure that is
5072 filled with version info. That info is later checked by server and
5076 Frequency : rare, liblustre only.
5077 Bugzilla : 9296, 9581
5078 Description: Two simultaneous writes from liblustre at offset within same page
5079 might proceed at the same time overwriting eachother with stale
5081 Details : I/O lock withing llu_file_prwv was released too early, before data
5082 actually was hitting the wire. Extended lock-holding time until
5083 server acknowledges receiving data.
5086 Frequency : extremely rare. Never observed in practice.
5088 Description: avoid generating lustre_handle cookie of 0.
5089 Details : class_handle_hash() generates handle cookies by incrementing
5090 global counter, and can hit 0 occasionaly (this is unlikely, but
5091 not impossible, because initial value of cookie counter is
5092 selected randonly). Value of 0 is used as a sentinel meaning
5093 "unassigned handle" --- avoid it. Also coalesce two critical
5094 sections in this function into one.
5096 Severity : enhancement
5098 Description: allow liblustre clients to delegate truncate locking to OST
5099 Details : To avoid overhead of locking, liblustre client instructs OST to
5100 take extent lock in ost_punch() on client's behalf. New connection
5101 flag is added to handle backward compatibility.
5103 Severity : enhancement
5104 Bugzilla : 4928, 7341, 9758
5105 Description: allow number of OST service threads to be specified
5106 Details : a module parameter allows the number of OST service threads
5107 to be specified via "options ost ost_num_threads={N}" in the
5108 OSS's /etc/modules.conf or /etc/modprobe.conf.
5112 Bugzilla : 6146, 9635, 9895
5113 Description: servers crash with bad pointer in target_handle_connect()
5114 Details : In rare cases when a client is reconnecting it was possible that
5115 the connection request was the last reference for that export.
5116 We would temporarily drop the export reference and get a new
5117 one, but this may have been the last reference and the export
5118 was just destroyed. Get new reference before dropping old one.
5120 Severity : enhancement
5121 Frequency : if client is started with failover MDS
5123 Description: Allow multiple MDS hostnames in the mount command
5124 Details : Try to read the configuration from all specified MDS
5125 hostnames during a client mount in case the "primary"
5128 Severity : enhancement
5130 Description: Stop sending data to evicted clients as soon as possible.
5131 Details : Check if the client we are about to send or are sending data to
5132 was evicted already. (Check is done every second of waiting,
5133 for which l_wait_event interface was extended to allow checking
5134 of exit condition at specified intervals).
5137 Frequency : rare, normally only when NFS exporting is done from client
5139 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5141 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5144 Frequency : always, if srand() is called before liblustre initialization
5146 Description: Liblustre uses system PRNG disturbing its usage by user application
5147 Details : Introduce internal to lustre fast and high-quality PRNG for
5148 lustre usage and make liblustre and some other places in generic
5149 lustre code to use it.
5151 Severity : enhancement
5152 Bugzilla : 9477, 9557, 9870
5153 Description: Verify that the MDS configuration logs are updated when xml is
5154 Details : Check if the .xml configuration logs are newer than the config
5155 logs stored on the MDS and report an error if this is the case.
5156 Request --write-conf, or allow starting with --old_conf.
5158 Severity : enhancement
5160 Description: Handle symlinks in the path when checking if Lustre is mounted.
5161 Details : Resolve intermediate symlinks when checking if a client has
5162 mounted a filesystem to avoid duplicate client mounts.
5167 Description: lconf can hit an error exception but still return success.
5168 Details : The lconf command catches the Command error exception at the top
5169 level script context and will exit with the associated exit
5170 status, but doesn't ensure that this exit status is non-zero.
5175 Description: failure of ptlrpc thread startup can cause oops
5176 Details : Starting a ptlrpc service thread can fail if there are a large
5177 number of threads or the server memory is very fragmented.
5178 Handle this without oopsing.
5181 Frequency : always, only if liblustre and non-default acceptor port was used
5183 Description: liblustre cannot connect to servers with non-default acceptor port
5184 Details : tcpnal_set_default_params() was not called and was therefore
5185 ignoring the environment varaible TCPNAL_PORT, as well as other
5186 TCPNAL_ environment variables
5191 Description: two objects could be created on the same OST for a single file
5192 Details : If an OST is down, in some cases it was possible to create two
5193 objects on a single OST for a single file. No problems other
5194 than potential performance impact and spurious error messages.
5198 Bugzilla : 5681, 9562
5199 Description: Client may oops in ll_unhash_aliases
5200 Details : Client dcache may become inconsistent in race condition.
5201 In some cases "getcwd" can fail if the current directory is
5207 Description: Inode refcounting problems in NFS export code
5208 Details : link_raw functions used to call d_instantiate without obtaining
5209 extra inode reference first.
5213 Bugzilla : 9942, 9903
5214 Description: Referencing freed requests leading to crash, memleaks with NFS.
5215 Details : We used to require that call to ll_revalidate_it was always
5216 followed by ll_lookup_it. Also with revalidate_special() it is
5217 possible to call ll_revalidate_it() twice for the same dentry
5218 even if first occurence returned success. This fix changes semantic
5219 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5220 reference on a request referred from the intent.
5221 ll_intent_release() then releases such a request.
5224 Frequency : rare, normally benchmark loads only
5226 Description: unlinked inodes were kept in memory on the client
5227 Details : If a client is repeatedly creating and unlinking files it
5228 can accumulate a lot of stale inodes in the inode slab cache.
5229 If there is no other client load running this can cause the
5230 client node to run out of memory. Instead flush old inodes
5231 from client cache that have the same inode number as a new inode.
5234 Frequency : SLES9 2.6.5 kernel and long filenames only
5235 Bugzilla : 9969, 10379
5236 Description: utime reports stale NFS file handle
5237 Details : SLES9 uses out-of-dentry names in some cases, which confused
5238 the lustre dentry revalidation. Change it to always use the
5242 Frequency : rare, unless heavy write-truncate concurrency is continuous
5243 Bugzilla : 4180, 6984, 7171, 9963, 9331
5244 Description: OST becomes very slow and/or deadlocked during object unlink
5245 Details : filter_destroy() was holding onto the parent directory lock
5246 while truncating+unlinking objects. For very large objects this
5247 may block other threads for a long time and slow overall OST
5248 responsiveness. It may also be possible to get a lock ordering
5249 deadlock in this case, or run out of journal credits because of
5250 the combined truncate+unlink. Solution is to do object truncate
5251 first in one transaction without parent lock, and then do the
5252 final unlink in a new transaction with the parent lock. This
5253 reduces the lock hold time dramatically.
5256 Frequency : rare, 2.4 kernels only
5258 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5259 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5260 before unmounting their filesystem in order to avoid deadlock.
5261 The kernel_locked() function in 2.4 kernels only checks whether
5262 the kernel lock is held, not whether it is this process that is
5263 holding it as 2.6 kernels do.
5268 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5269 Details : The client ptlrpc code may be trying to reconnect to a down
5270 server before a previous connection attempt has timed out.
5271 Increase the reconnect interval to be longer than the connection
5272 timeout interval to avoid sending duplicate connections to
5276 Frequency : echo_client brw_test command
5278 Description: fix echo_client to work with OST preallocated code
5279 Details : OST preallocation code (5137) didn't take echo_client IO path
5280 into account: echo_client calls filter methods outside of any
5281 OST thread and, hence, there is no per-thread preallocated
5282 pages and buffers to use. Solution: hijack pga pages for IO. As
5283 a byproduct, this avoids unnecessary data copying.
5287 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5288 Description: Client can oops in mdc_commit_close() after open replay
5289 Details : It was possible for the MDS to return an open request with no
5290 transaction number in mds_finish_transno() if the client was
5291 evicted, but without actually returning an error. Clients
5292 would later try to replay that open and may trip an assertion
5293 Simplify the client close codepath, and always return an error
5294 from the MDS in case the open is not successful.
5297 Frequency : rare, 2.6 OSTs only
5299 Description: OST may deadlock under high load on fragmented files
5300 Details : If there was a heavy load and highly-fragmented OST filesystems
5301 it was possible to have all the OST threads deadlock waiting on
5302 allocation of biovecs, because the biovecs were not released
5303 until the entire RPC IO was completed. Instead, release biovecs
5304 as soon as they are complete to ensure forward IO progress.
5306 Severity : enhancement
5308 Description: Support for specifying external journal device at mount
5309 Details : If an OST or MDS device is formatted with an external journal
5310 device, this device major/minor is stored in the ext3 superblock
5311 and may not be valid for failover. Allow detecting and
5312 specifying the external journal at mount time.
5317 Description: Mounting an MDS with pending unlinked files may cause oops
5318 Details : target_finish_recovery() calls mds_postrecov() which returned
5319 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5320 considers this an error and immediately begins cleaning up the
5321 lov, just after starting the mds_lov process
5323 Severity : enhancement
5325 Description: Implement 'lfs df' to report actual free space on per-OST basis
5326 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5327 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5328 usage of objects; '-h' to report in human readable format.
5330 ------------------------------------------------------------------------------
5332 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5339 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5340 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5341 at startup during orphan unlinking. Instead, skip buddy update
5342 before setup, as it will be regenerated after recovery is complete.
5347 Description: If an OST is inactive, its locks might reference stale inodes.
5348 Details : lov_change_cbdata() must iterate over all namespaces, even if
5349 they are inactive to clear inode references from the lock.
5351 Severity : enhancement
5352 Frequency : occasional, if non-standard max_dirty_mb used
5354 Description: Client will block write RPCs if not enough grant
5355 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5356 then the client will block writes while waiting for another RPC
5357 to complete instead of consuming its dirty limit. With change
5358 we get improved performance when max_dirty_mb is small.
5360 Severity : enhancement
5361 Bugzilla : 3389, 6253
5362 Description: Add support for supplementary groups on the MDS.
5363 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5364 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5365 lookups for user supplementary groups into a cache.
5369 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5370 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5371 for liblustre O_EXCL is enforced by MDS.
5375 Description: readdir never returns NULL in liblustre.
5376 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5377 dentry in struct dirent.
5381 Frequency : liblustre only, depends on application IO pattern
5382 Description: liblustre clients evicted if not contacting servers
5383 Details : Don't put liblustre clients into the ping_evictor list, so
5384 they will not be evicted by the pinger ever.
5386 Severity : enhancement
5388 Description: Add ability to evict clients by NID from MDS.
5389 Details : By echoing "nid:$NID" string into
5390 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5391 $NID would be instantly evicted from this MDS and from all active
5392 OSTs connected to it.
5396 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5397 Details : lookup_it_finish() used to query file size from OSTs that was not
5402 Description: service threads change working directory to that of init
5403 Details : Starting lustre service threads may pin the working directory
5404 of the parent thread, making that filesystem busy. Threads
5405 now change to the working directory of init to avoid this.
5409 Frequency : during shutdown only
5410 Description: shutdown with a failed MDS or OST can cause unmount to hang
5411 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5416 Frequency : 2.6 only
5417 Description: chmod/chown may include an extra supplementary group
5418 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5419 supplementary group and if none is specified this is used.
5422 Bugzilla : 5479 (6816)
5424 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5425 Details : If lookup is in progress on a file that is unlinked we might try
5426 to revalidate the inode and fail in revalidate after lookup is
5427 complete and ll_file_open() enqueues the open again but
5428 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5431 Frequency : always, if lconf --abort_recovery used
5433 Description: lconf --abort_recovery fails with 'Operation not supported'
5434 Details : lconf was attempting to abort recovery on the MDT device and not
5437 ------------------------------------------------------------------------------
5439 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5444 Frequency : rare (only unsupported configurations with a node running as an
5446 Bugzilla : 6514, 5137
5447 Description: Mounting a Lustre file system on a node running as an OST could
5449 Details : OSTs now preallocates memory needed to write out data at
5450 startup, instead of when needed, to avoid having to
5451 allocate memory in possibly low memory situations.
5452 Specifically, if the file system is mounted on on OST,
5453 memory pressure could force it to try to write out data,
5454 which it needed to allocate memory to do. Due to the low
5455 memory, it would be unable to do so and the node would
5456 become unresponsive.
5458 Severity : enhancement
5460 Description: Addition of lconf --service command line option
5461 Details : lconf now accepts a '--service <arg>' option, which is
5462 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5464 Severity : enhancement
5466 Description: Failover mode is now the default for OSTs.
5467 Details : By default, OSTs will now run in failover mode. To return to
5468 the old behaviour, add '--failout' to the lmc line for OSTs.
5470 Severity : enhancement
5472 Description: Health checks are now provided for MDS and OSTs
5473 Details : Additional detailed health check information on MSD and OSTs
5474 is now provided through the procfs health_check value.
5477 Frequency : occasional, depends on IO load
5479 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5480 after numerous create/delete cycles
5481 Details : The ext3 inode allocation policy would not allocate new inodes
5482 very well on the OSTs because there are no new directories
5483 being created. Instead we look for groups with free space if
5484 the parent directories are nearly full.
5489 Description: Network or server problems during mount may cause partially
5490 mounted clients instead of returning an error.
5491 Details : The config llog parsing code may overwrite the error return
5492 code during mount error handling, returning success instead
5498 Description: MDS can fail to allocate large reply buffers
5499 Details : After long uptimes the MDS can fail to allocate large reply
5500 buffers (e.g. zconf client mount config records) due to memory
5501 fragmentation or consumption by the buffer cache. Preallocate
5502 some large reply buffers so that these replies can be sent even
5503 under memory pressure.
5507 Frequency : rare (liblustre)
5508 Description: fsx running with liblustre complained that using truncate() to
5509 extend the file doesn't work. This patch corrects that issue.
5510 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5511 ATTR_SIZE and lsm use in llu_setattr_raw.
5515 Frequency : rare, only 2.6 kernels
5516 Description: Unusual file access patterns on the MDS may result in inode
5517 data being lost in very rare circumstances.
5518 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5519 code in the 2.6 kernel caused the inode and/or EA data not to
5520 be read from disk, causing single-file corruption.
5524 Frequency : rare, only 2.6 filesystems using extents
5525 Description: Heavy concurrent write and delete load may cause data corruption.
5526 Details : It was possible under high-load situations to have an extent
5527 metadata block in the block device cache from a just-unlinked
5528 file overwrite a newly-allocated data block. We now unmap any
5529 metadata buffers that alias just-allocated data blocks.
5533 Frequency : filesystems with default stripe_count larger than 77
5534 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5535 Details : lconf specifies an inode size of 4096 bytes when the default
5536 stripe_count is larger than 77. This conflicts with the default
5537 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5538 this case to avoid pinning too much memory for large EAs.
5540 ------------------------------------------------------------------------------
5542 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5547 Frequency : rare (extremely heavy IO load with hundreds of clients)
5549 Description: Client is evicted, gets IO error writing to file
5550 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5551 caused clients to be evicted for AST timeouts. The fixes in
5552 bug 5192 mean we no longer need to have such short AST timeouts
5553 so ldlm_timeout has been increased.
5556 Frequency : occasional during --force or --failover shutdown under load
5557 Bugzilla : 5949, 4834
5558 Description: Server oops/LBUG if stopped with --force or --failover under load
5559 Details : a collection of import/export refcount and cleanup ordering
5560 issues fixed for safer force cleanup
5563 Frequency : only filesystems larger than 120 OSTs
5564 Bugzilla : 5990, 6223
5565 Description: lfs getstripe would oops on a very large filesystem
5566 Details : lov_getconfig used kfree on vmalloc'd memory
5569 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5570 Bugzilla : 6242, 6243
5571 Description: reading from files that had been truncated to a non-zero size
5572 but never opened returned no data
5573 Details : ll_file_read() reads zeros from no-object files to EOF
5578 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5579 Details : There's little harm in aborting MDS/OSS recovery and letting it
5580 try again, so I removed the LASSERT and return an error instead.
5582 Severity : enhancement
5584 Description: New debugging infrastructure for tracking down data corruption
5585 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5586 (b) cover more of the client-side code path, and (c) try to narrow
5587 down where problems occurred
5591 Bugzilla : 3819, 4364, 4397, 6313
5592 Description: Racing close and eviction MDS could cause assertion in mds_close
5593 Details : It was possible to get multiple mfd references during close and
5594 client eviction, leading to one thread referencing a freed mfd.
5596 Severity: : enhancement
5597 Bugzilla : 3262, 6359
5598 Description: Attempts to reconnect to servers are now more aggressive.
5599 Details : This builds on the enhanced upcall-less recovery that was added
5600 in 1.4.2. When trying to reconnect to servers, clients will
5601 now try each server in the failover group every 10 seconds. By
5602 default, clients would previously try one server every 25 seconds.
5607 Description: After recovery, certain operations trigger a failed
5608 assertion on a client.
5609 Details : Failing over an mds, using lconf -d --failover, while a
5610 client was doing a readdir() call would cause the client to
5611 LBUG after recovery completed and the readdir() was resent.
5613 Severity : enhancement
5615 Description: Default groups are now added by lconf
5616 Details : You can now run lconf --group <servicename> without having to
5617 manually add groups with lmc.
5620 Frequency : occasional
5622 Description: Nodes with an elan id of 0 trigger a failed assertion
5625 Frequency : always when accessing e.g. tty/console device nodes
5627 Description: tty and some other devices nodes cannot be used on lustre
5628 Details : file's private_data field is used by device data and lustre
5629 values in there got lost. New field was added to struct file to
5630 store fs-specific private data.
5633 Frequency : when exporting Lustre via NFS
5635 Description: NFSD failed occasionally when looking up a path component
5636 Details : NFSD is looking up ".." which was broken in ext3 directories
5637 that had grown large enough to become hashed.
5640 Frequency : Clusters with multiple interfaces not on the same subnet
5642 Description: Nodes will repeatedly try to reconnect to an interface which it
5643 cannot reach and report an error to the log.
5644 Details : Extra peer list entries will be created by lconf with some peers
5645 unreachable. lconf now validates the peer before adding it.
5648 Frequency : Only if a default stripe is set on the filesystem root.
5650 Description: Setting a default stripe on the filesystem root prevented the
5651 filesystem from being remounted.
5652 Details : The client was sending extra request flags in the root getattr
5653 request and did not allocate a reply buffer for the dir EA.
5656 Frequency : occasional, higher if lots of files are accessed by one client
5657 Bugzilla : 6159, 6097
5658 Description: Client trips assertion regarding lsm mismatch/magic
5659 Details : While revalidating inodes the VFS looks up inodes with ifind()
5660 and in rare cases can find an inode that is being freed.
5661 The ll_test_inode() code will free the lsm during ifind()
5662 when it finds an existing inode and then the VFS later attaches
5663 this free lsm to a new inode.
5667 Bugzilla : 6422, 7030
5668 Description: MDS deadlock between mkdir and client eviction
5669 Details : Creating a new file via mkdir or mknod (starting a transaction
5670 and getting the ns lock) can deadlock with client eviction
5671 (gets ns lock and trying to finish a synchronous transaction).
5674 Frequency : occasional
5675 Description: While starting a server, the fsfilt_ext3 module could not be
5677 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5678 kernels. Previously, lconf would still use the ext3 name
5679 when trying to load modules. Now, it will correctly use
5680 ext3 on 2.4 and ldiskfs on 2.6.
5682 Severity : enhancement
5683 Description: The default stripe count has been changed to 1
5684 Details : The interpretation of the default stripe count (0, to lfs
5685 or lmc) has been changed to mean striping across a single
5686 OST, rather than all available. For general usage we have
5687 found a stripe count of 1 or 2 works best.
5689 Severity : enhancement
5690 Description: Add support for compiling against Cray portals.
5691 Details : Conditional compiling for some areas that are different
5695 Frequency : occasional
5696 Bugzilla : 6409, 6834
5697 Description: Creating files with an explicit stripe count may lead to
5698 a failed assertion on the MDS
5699 Details : If some OSTs are full or unavailable, creating files may
5700 trigger a failed assertion on the MDS. Now, Lustre will
5701 try to use other servers or return an error to the
5705 Frequency : occasional
5707 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5708 caused lock timeout during readahead (since 1.4.2).
5709 Details : Processes doing readahead might match a lock that hasn't been
5710 granted yet if there are overlapping and conflicting lock
5711 requests. The readahead process waits on ungranted lock
5712 (original lock is CBPENDING), while OST waits for that process
5713 to cancel CBPENDING read lock and eventually evicts client.
5715 Severity : enhancement
5717 Description: Initial enabling of flock support for clients
5718 Details : Implements fcntl advisory locking and file status functions.
5719 This feature is provided as an optional mount flag (default
5720 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5721 locking are implemented yet, and those that are are not guaranteed
5722 to be completely correct in production environments.
5723 mount -t lustre -o [flock|noflock] ...
5726 Frequency : occasional
5728 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5729 assertion in the ext3 JBD (journaling) layer.
5730 Details : The b_committed_data struct is protected by the big kernel lock
5731 in 2.4 kernels, serializing journal_commit_transaction() and
5732 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5733 access to this struct. In 2.6 kernels there is finer grained
5734 locking to improve SMP performance of the JBD layer.
5738 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5743 Description: merge and cleanup kernel patches.
5746 -----------------------------------------------------------------------------
5748 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5750 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5751 versions of Lustre. Please update all servers and clients to
5752 version 1.4.2 or later at the same time. You must also run
5753 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5754 to update the configuration logs.
5756 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5757 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5758 - swab llog records of type '0' so we get proper header size/idx (5861)
5759 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5760 - fix rename of one directory over another leaking an inode (5953)
5761 - avoid SetPageDirty on 2.6 (5981)
5762 - don't re-add just-being-destroyed locks to the waiting list (5653)
5763 - when creating new directories, inherit the parent's custom
5764 striping settings if present parent (3048)
5765 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5766 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5767 - don't hold client locks on temporary worklist from l_lru (5666)
5768 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5769 - automatically evict dead clients (5921)
5770 - Update file size properly in create+truncate+fstat case (6196)
5771 - Do not unhash mountpoint dentries, do not allow removal of
5773 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5774 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5775 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5776 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5777 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5778 - Correctly handle reads of files with no objects (6243)
5779 - lctl recover will also mark a device active if deactivate used (5933)
5781 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5782 - allow --write-conf on an MDS with different nettype than client (5619)
5783 - don't write config llogs to MDS for mounts not from that MDS (5617)
5784 - lconf should create multiple TCP connections from a client (5201)
5785 - init scripts are now turned off by default; run chkconfig --on
5786 lustre and chkconfig --on lustrefs to use them
5787 - upcalls are no longer needed for clients to recover to failover
5789 - add --abort-recovery option to lconf to abort recovery on device
5791 - add support for an arbitrary number of OSTs (3026)
5792 - Quota support protocol changes.
5793 - forward compatibility changes to wire structs (6007)
5794 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5795 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5796 - improved "lustre status" script
5797 - initialize blocksize for non-regular files (6062)
5798 - added --disable-server and --disable-client configure options (5782)
5799 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5800 - Vanilla 2.4.29 support
5801 - increase maximum number of obd devices to 520 (6242)
5802 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5803 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5804 SLES 9 2.6.5 kernels
5805 - limit stripes per file to 160 (the maximum EA size) (6093)
5807 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5810 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5811 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5812 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5813 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5814 - small ext3 extents cleanups and fixes (5733)
5815 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5816 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5817 - avoid needless client->OST connect, fix handle mismatch (5317)
5818 - fix DLM error path that led to out-of-sync client, long delays (5779)
5819 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5820 - fix several locking issues related to i_size (5492,5624,5654,5672)
5821 - don't move pending lock onto export if it is already evicted (5683)
5822 - fix kernel oops when creating .foo in unlinked directory (5548)
5823 - fix deadlock in obdfilter statistics vs. object create (5811)
5824 - use time_{before,after} to avoid timer jiffies wrap (5882)
5825 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5826 - Do not leak request if server was not able to process it (5154)
5827 - If mds_open unable to find parent dir, make that negative lookup(5154)
5828 - don't create new directories with extent-mapping (5909, 5936)
5830 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5831 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5832 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5833 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5834 - support for RHEL4 kernel on i686 (5773)
5835 - provide error messages when incompatible logs are encountered (5898)
5837 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5838 * version 1.4.0.10 (1.4.1 release candidate 1)
5840 - don't keep a lock reference when lock is not granted (4238)
5841 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5842 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5843 - drop import inflight refcount on signal_completed_replay error (5255)
5844 - unlock page after async write error during send (3677)
5845 - handle missing objects in filter_preprw_read properly (5265)
5846 - no transno return for symlink open, don't save no-trasno open (3440)
5847 - don't try to complete elan receive that already failed (4012)
5848 - free RPC server reply state on error (5406)
5849 - clean up thread from ptlrpc_start_thread() on error (5160)
5850 - readahead could read extra page into cache that wasn't ejected (5388)
5851 - prevent races in class_attach/setup/cleanup/detach (5260)
5852 - don't dereference de->d_inode after l_dput of de (5458)
5853 - use "int" for stripe value returned from lock_to_stripe (5544)
5854 - mballoc allocation and error-checking fixes in 2.6 (5504)
5855 - block device patches to fix I/O request sizes in 2.6 (5482)
5856 - look up hostnames for IB nals (5602)
5857 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5858 - don't start multiple acceptors for the same port (5277)
5859 - fix incorrect LASSERT in mds_getattr_name (5635)
5860 - export a proc file for general "ping" checking (5628)
5861 - fix "lfs check" to not block when the MDS is down (5628)
5863 - service request history (4965)
5864 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5865 - create an "evict_client" /proc entry on OSTs, like the MDS has
5866 - fix mount usage message, return errors per mount(8) (5168)
5867 - change grep [] to grep "[]" in tests so they work in more UMLs
5868 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5869 - remove ancient LOV_MAGIC_V0 EA support (5047)
5870 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5871 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5872 - allow readahead allocations to fail when low on memory (5383)
5873 - mmap locking landed again, after considerable improvement (2828)
5874 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5876 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5879 - send OST transaction number in read/write reply to free req (4966)
5880 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5881 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5882 - the watchdog thread now runs as interruptible (5246)
5883 - flock/lockf fixes (but it's still disabled, pending 5135)
5884 - don't use EXT3 constants in llite code (5094)
5885 - memory shortage at startup could cause assertion (5176)
5887 - reorganization of lov code
5888 - single portals codebase
5890 - add extents/mballoc support (5025)
5891 - direct I/O reads in the obdfilter (4048)
5892 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5894 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5897 - send OST transaction number in read/write reply to free req (4966)
5898 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5899 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5900 - the watchdog thread now runs as interruptible (5246)
5901 - handle missing objects in filter_preprw_read properly (5265)
5902 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5903 - drop import inflight refcount on signal_completed_replay error (5255)
5904 - unlock page after async write error during send (3677)
5905 - return original error code on reconstructed replies (3761)
5906 - no transno return for symlink open, don't save no-trasno open (3440)
5908 - add pid to ldlm debugging output (4922)
5909 - bump the watchdog timeouts -- we can't handle 30sec yet
5910 - extra debugging for orphan dentry/inode bug (5259)
5912 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5915 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5916 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5917 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5918 - fix NULL dereference in /proc/sys/portals/routes (4827)
5919 - allow failed mdc_close() operations to be interrupted (4561)
5920 - stop precreate on OST before MDS would time out on it (4778)
5921 - don't send partial-page writes before EOF from client (4410)
5922 - discard client grant for sub-page writes on large-page clients (4520)
5923 - don't free dentries not owned by NFS code, check generation (4806)
5924 - fix lsm leak if mds_create_objects() fails (4801)
5925 - limit debug_daemon file size, always print CERROR messages (4789)
5926 - use transno after validating reply (3892)
5927 - process timed out requests if import state changes (3754)
5928 - update mtime on OST during writes, return in glimpse (4829)
5929 - add mkfsoptions to LDAP (4679)
5930 - use ->max_readahead method instead of zapping global ra (5039)
5931 - don't interrupt __l_wait_event() during strace
5933 - add software watchdogs to catch hung threads quickly (4941)
5934 - make lustrefs init script start after nfs is mounted
5935 - fix CWARN/ERROR duplication (4930)
5936 - return async write errors to application if possible (2248)
5937 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5938 - print NAL number in %x format (4645)
5939 - update barely-supported suse-2.4.21-171 series (4842)
5940 - support for sles 9 %post scripts
5941 - support for building 2.6 kernel-source packages
5942 - support for sles km_* packages
5944 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5947 - ignore -ENOENT errors in osc_destroy (3639)
5948 - notify osc create thread that OSC is being cleaned up (4600)
5949 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5950 - reconstruct ost_handle() like mds_handle() (4657)
5951 - create a new thread to do import eviction to avoid deadlock (3969)
5952 - let lconf resolve symlinked-to devices (4629)
5953 - don't unlink "objects" from directory with default EA (4554)
5954 - hold socknal file ref over connect in case target is down (4394)
5955 - allow more than 32000 subdirectories in a single directory (3244)
5956 - fix blocks count for O_DIRECT writes (3751)
5957 - OST returns ENOSPC from object create when no space left (4539)
5958 - don't send truncate RPC if file size isn't changing (4410)
5959 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5960 - bind to privileged port in socknal and tcpnal (3689)
5962 - rate limit CERROR/CWARN console message to avoid overload (4519)
5963 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5964 - basic mmap support (3918)
5965 - kernel patch series update from b1_4 (4711)
5967 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5970 - avoid crash during MDS cleanup with OST shut down (2775)
5971 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5972 - don't use bad inodes on the MDS (3744)
5973 - dynamic object preallocation to improve recovery speed (4236)
5974 - don't hold spinlock over lock dumping or change debug flags (4401)
5975 - don't zero obd_dev when it is force cleaned (3651)
5976 - print grants to console if they go negative (4431)
5977 - "lctl deactivate" will stop automatic recovery attempts (3406)
5978 - look for existing locks in ldlm_handle_enqueue() (3764)
5979 - don't resolve lock handle twice in recovery avoiding race (4401)
5980 - revalidate should check working dir is a directory (4134)
5982 - don't always mark "slow" obdfilter messages as errors (4418)
5984 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5987 - don't close LustreDB during write_conf until it is done (3860)
5988 - fix typo in lconf for_each_profile (3821)
5989 - allow dumping logs from multiple threads at one time (3820)
5990 - don't allow multiple threads in OSC recovery (3812)
5991 - fix debug_size parameters (3864)
5992 - fix mds_postrecov to initialize import for llog ctxt (3121)
5993 - replace config semaphore with spinlock (3306)
5994 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5995 - don't allow enqueue to complete on a destroyed export (3822)
5996 - down write_lock before checking llog header bitmap (3825)
5997 - recover from lock replay timeout (3764)
5998 - up llog sem before sending rpc (3652)
5999 - reduce ns lock hold times when setting kms (3267)
6000 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6001 - fix NULL deref and obd_dev leak on setup error (3312)
6002 - replace some LBUG about llog ops with error handling (3841)
6003 - don't match INVALID dentries from d_lookup and spin (3784)
6004 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6005 - fix invalid assertion in ptlrpc_set_wait (3880)
6007 - add libwrap support for the TCP acceptor (3996)
6008 - add /proc/sys/portals/routes for non-root route listing (3994)
6009 - allow setting MDS UUID in .xml (2580)
6010 - print the stack of a process that LBUGs (4228)
6012 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6015 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6016 - make sure to unset replay flag from failed open requests (3440)
6017 - if default stripe count is 0, use OST count for inode size (3636)
6018 - update parent mtime/ctime on client for create/unlink (2611)
6019 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6020 - free recovery state on server during a forced cleanup (3571)
6021 - unregister_reply for resent reqs (3063)
6022 - loop back devices mounting and status check on 2.6 (3563)
6023 - fix resource-creation race that can provoke i_size == 0 (3513)
6024 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6025 - more debugging for page-accounting assertion (3746)
6026 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6027 - avoid deadlock after precreation failure (3758)
6028 - fix race and lock order deadlock in orphan handling (3450, 3750)
6029 - add validity checks when grabbing inodes from l_ast_data (3599)
6031 - add /proc/.../recovery_status to obdfilter (3428)
6032 - lightweight CDEBUG infrastructure, debug daemon (3668)
6033 - change default OSC RPC parameters to be better on small clusters
6034 - turn off OST read cache for files smaller than 32MB
6035 - install man pages and include them in rpms (3100)
6036 - add new init script for (un)mounting lustre filesystems (2593)
6037 - run chkconfig in %post for init scripts (3701)
6038 - drop scimac NAL (unmaintained)
6040 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6043 - clean kiobufs before and after use (3485)
6044 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6045 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6046 - change init script to fail more clearly if not run as root (1528)
6047 - allow clients to reconnect during replay (1742)
6048 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6049 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6050 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6051 - fix cond_resched (3554)
6052 - only evict unfinished clients after recovery (3515)
6053 - allow bulk resend, prevent data loss (3570)
6054 - dynamic ptlrpc request buffer allocation (2102)
6055 - don't allow unlinking open directory if it isn't empty (2904)
6056 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6057 - remove extraneous obd dereference causing LASSERT failure (3334)
6058 - don't use get_cycles() when creating temp. files on the mds (3156)
6059 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6060 - handle EEXIST for set-stripe, set proper directory name (3336)
6062 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6063 - fix ksocknal_fmb_callback() error messages (2918)
6065 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6068 - don't copy lvb into (possibly NULL) reply on error (2983)
6069 - don't deref dentry after dput, don't free lvb on error (2922)
6070 - use the kms to determine writeback rpc length (2947)
6071 - increment oti_logcookies when osc is inactive (2948)
6072 - update client's i_blocks count via lvb messages (2543)
6073 - handle intent open/close of special files properly (1557)
6074 - mount MDS with errors=remount-ro, like obdfilter (2009)
6075 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6076 - don't use cancelling-locks' kms values (2947)
6077 - use highest lock extent for kms, not last one (2925)
6078 - don't dereference ERR_PTR() dentry in error handling path (3107)
6079 - fix thread race in portals_debug_dumplog() (3122)
6080 - create lprocfs device entries at setup instead of at attach (1519)
6081 - common AST error handler, don't evict client on completion race (3145)
6082 - zero nameidata in detach_mnt in 2.6 (3118)
6083 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6084 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6085 - initialize RPC timeout timer earlier for 2.6 (3219)
6086 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6087 - print nal/nid for unknown nid (3258)
6088 - additional checks for oscc recovery before doing precreate (3284)
6089 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6090 - don't crash in mdc_close for bad permissions on open (3285)
6091 - zero i_rdev for non-device files (3147)
6092 - clear page->private before handing to FS, better assertion (3119)
6093 - tune the read pipeline (3236)
6094 - fix incorrect decref of invalidated dentry (2350)
6095 - provide read-ahead stats and refine rpc in flight stats (3328)
6096 - don't hold journal transaction open across create RPC (3313)
6097 - update atime on MDS at close time (3265)
6098 - close LDAP connection when recovering to avoid server load (3315)
6099 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6100 - don't leak open file on MDS after open resend (3325)
6101 - serialize filter_precreate and filter_destroy_precreated (3329)
6102 - loop device shouldn't call sync_dev() for nul device (3092)
6103 - clear page cache after eviction (2766)
6104 - resynchronize MDS->OST in background (2824)
6105 - refuse to mount the same filesystem twice on same mountpoint (3394)
6106 - allow llmount to create routes for mounting behind routers (3320)
6107 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6108 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6109 - fix rare problem with rename on htree directories (3417)
6111 - allow default OST striping configuration per directory (1414)
6112 - fix compilation for qswnal for 2.6 kernels (3125)
6113 - increase maximum number of MDS request buffers for large systems
6114 - change liblustreapi to be useful for external progs like lfsck (3098)
6115 - increase local configuration timeout for slow disks (3353)
6116 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6118 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6121 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6122 - don't overwrite extent policy data in reply if lock was blocked (2901)
6123 - drop filter export grants atomically with removal from device (2663)
6124 - del obd_self_export from work_list in class_disconnect_exports (2908)
6125 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6126 - swab reply message in mdc_close, other PPC fixes (2464)
6127 - fix destroying of named logs (2325)
6128 - overwrite old logs when running lconf --write_conf (2264)
6129 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6130 - fix race in target_handle_connect (2898)
6131 - mds_reint_create() should take same inode create lock (2926)
6132 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6133 - don't close files for self_export to avoid uninitialized obd (2936)
6134 - allow MDS with the same name as client node (2939)
6135 - hold dentry reference for closed log files for unlink (2325)
6136 - reserve space for all logs during transactions (2059)
6137 - don't evict page beyond end of stripe extent (2925)
6138 - don't oops on a deleted current working directory (2399)
6139 - handle hard links to targets without a parent properly (2517)
6140 - don't dereference NULL lock when racing during eviction (2867)
6141 - don't grow lock extents when lots of conflicting locks (2919)
6143 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6146 - account for cache space usage on clients to avoid data loss (974)
6147 - lfsck support in lustre kernel code (2349)
6148 - reduce journal credits needed for BRW writes (2370)
6149 - orphan handling to avoid losing space on client/server crashes
6150 - ptlrpcd can be blocked, stopping ALL progress (2477)
6151 - use lock value blocks to assist in proper KMS, faster stat (1021)
6152 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6153 - recovery for initial connections (2355)
6154 - fixes for mds_cleanup_orphans (1934)
6155 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6156 - block all file creations until orphan recovery completes (1901)
6157 - client remove rq_connection from request struct (2423)
6158 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6159 - recovery timer race (2670)
6160 - mdc_close recovey bug (2532)
6161 - ptlrpc cleanup bug (2710)
6162 - mds timeout on local locks (2588)
6163 - namespace lock held during RPCs (2431)
6164 - handle interrupted sync write properly (2503)
6165 - don't try to handle a message that hasn't been replied to (2699)
6166 - client assert failure during cleanup after abort recovery (2701)
6167 - leak mdc device after failed mount (2712)
6168 - ptlrpc_check_set allows timedout requests to complete (2714)
6169 - wait for inflight reqs when ptlrpcd finishes (2710)
6170 - make sure unregistered services are removed from the srv_list
6171 - reset bulk XID's when resending them (caught by 1138 test)
6172 - unregister_bulk after timeout
6173 - fix lconf error (2694)
6174 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6175 - readahead locks pages, leaves pending causing memory pressure (2673)
6176 - increase OST request buffers to 4096 on large machines (2729)
6177 - fix up permission of existing directories in simple_mkdir (2661)
6178 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6179 - don't assign transno to errored transactions (2742)
6180 - don't delete objects on OST if given a bogus objid from MDS (2751)
6181 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6182 - if rq_no_resend, then timeout request after recovery (2432)
6183 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6184 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6185 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6186 - MDS recovery shouldn't delete objects that it has given out (2730)
6187 - if enqueue arrives after completion, don't clobber LVB (2819)
6188 - don't unlock pages twice when trigger_group_io returns error (2814)
6189 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6190 - don't write pages to disk if there was an error (1450)
6191 - don't ping imports that have recovery disabled (2676)
6192 - take buffered bytes into account when balancing socknal conn (2817)
6193 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6194 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6195 - remove little-endian swabbing of llog records (1987)
6196 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6197 - retry reposting request buffers if they fail (1191)
6198 - grow extent at grant time to avoid granting a revoked lock (2809)
6199 - lock revoke doesn't evict page if covered by a second lock (2765)
6200 - disable VM readahead to avoid reading outside lock extents (2805)
6202 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6203 - updated kernel patches for hp-2.4.20 kernel (2681)
6205 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6208 - fix truncated write corruption (2366)
6209 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6211 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6212 - ensure len(uuid) < 37 in lmc (1171)
6213 - fix ia64 OOPS in llog_test (2255)
6214 - zero end of page at obdfilter for partial page writes (2648)
6215 - don't leave stale dentries around after renames (bug 2428)
6216 - fix timeouts when evicting a client with a single lock held (2642)
6217 - set deadline for the initial HELLO message to drain (2634)
6218 - print out dotted-quad IP addresses in the socknal (2302)
6220 - additional debugging for MDS client eviction problem (2443)
6221 - fix mkfsoptions support for osts (2603, 2604)
6223 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6226 - add series for the vanilla 2.6.0 kernel
6227 - add series for the vanilla 2.4.24 kernel
6228 - add series for a cray x86/64 UL kernel drop
6229 - fix xattr patches for the vanilla 2.4.19 series
6231 - generate true UUIDs in lmc (1171)
6232 - have portals stack dumping break in UML (2466)
6233 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6234 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6235 - wake callback waiting threads on client eviction (2460)
6236 - Add --ptldebug and --subsystem to lmc (1719)
6237 - update assertion to allow safe interrupt allocation
6238 - set rq_no_resend for cancel requests (2432)
6239 - recalculate ptlrpcd timeout after resend (2494)
6240 - call vfs_rmdir when removing pending directories (2368)
6241 - fix renaming a file to itself (2429)
6242 - lmc creates a default one-stripe lov (2454)
6243 - expand procfs space to handle large clusters (2326)
6244 - increase UML stack to avoid overflow
6245 - update lconf's list of debug and subsystem masks
6246 - fix lfs find --obd (2510)
6247 - /proc tunable for disabling filter read caching (2591)
6248 - stop rpm packages from altering slapd.conf (2301)
6249 - disable nagle in the socknal under 0conf (2578)
6250 - choose mds inode size based on stripe count (2572)
6251 - fix kernel-source rpm problems (2516)
6253 - add --disable-doc to avoid pdf generation (2421)
6254 - update documentation, tests, type-os, comments
6255 - avoid format warnings on ia64
6256 - remove the TOE NAL
6257 - tiny code cleanups by removing unused fields
6259 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6262 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6263 - remove the most blatant lies from BUILDING file (2371)
6264 - change default debug level to reasonable production setting
6265 - reduce client side cache size to reduce cache flush time
6266 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6267 - make TCP zerocopy and pinger support enabled by default (2476)
6268 - sync writes completed after process exits caused crashes (2319)
6269 - maintain correct mount count on the MDS (2356)
6270 - backout 1557, because 2316 wasn't really fixed
6271 - better file I/O statistics gathering in /proc
6272 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6273 - another kernel patch to fix zero-copy TCP function export
6274 - don't take duplicate lock when processing re-sent getattr (2420)
6275 - lctl uses obd_self_export instead of creating new conn (2353)
6276 - MDS/OST recovery case which requires object creation asserted (2425)
6277 - move lfs from /usr/sbin to /usr/bin in packages
6278 - fix race between mds_client_add and mds_client_free (2417)
6279 - use kmalloc instead of slabs in portals (2430)
6280 - don't create duplicate records when a failover MDS is present (2442)
6281 - remove unnecessary mount age check (2332)
6282 - don't remove directory inodes from locks prematurely (2451)
6283 - don't break if MDS service name is the same as hostname (2103)
6284 - fix races in client write RPC generation when cache full (2482)
6286 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6289 - remove now-unused request->rq_obd (278)
6290 - if an allocation fails, print out how much memory we've used (1933)
6291 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6292 - add the "configurable stack size" patch to most series files (1256)
6293 - ability to write large log records, for 100+ OST configs (2306)
6294 - fix NULL deref when filter_prep fails (2314)
6295 - fix operator precedence error in filter_sync
6296 - dynamic allocation of socknal TX descriptors (2315)
6297 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6298 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6299 - ensure that CWARN messages in Portals always get to the syslog
6300 - __init/__exit are not for prototype decls (ldlm_init/exit)
6301 - x86-64 compile warning fixes
6302 - fix gateway LMC keyword conflict (2318)
6303 - fix MDS lock inversions in getattr/reint paths (1844)
6304 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6305 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6306 - DLM race condition prevented some lock evictions (2328)
6307 - ENOMEM detection and retry on socknal sends (2230)
6308 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6309 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6310 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6311 - better error messages when a client is rejected during recovery (1505)
6312 - avoid cancelling locks which were never granted, after failure (2330)
6313 - fix i_sem/journal inversion in mds_client_add (2333)
6314 - fix truncate/getattr lock cycle deadlock (2334)
6315 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6316 - fix two rare exit paths which could leak an l_lock() ref (2321)
6317 - fencepost error in MDS/OST orphan recovery (2226)
6318 - make log record alignment 8 bytes (1988)
6319 - lstripe now fails when requested offset > ost_count (2237)
6320 - ensure that all kernel series have a complete list.h (1607)
6321 - fix crashes in special-file operations (2316)
6322 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6324 - allow configurable automake binary, for testing new versions
6325 - small update to the lfs documentation
6327 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6329 * fix negative export reference count in fsfilt_sync (2312)
6331 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6332 * release candidate 0.9.1
6334 - orphans are moved into the PENDING directory for possible recovery
6335 - replayed opens now open by fid for orphan/rename safety (1042)
6336 - last close of an orphan inode generates a transno (683)
6337 - chdir() and mount() now pin the directory entry (1020)
6338 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6339 - discard very old requests without processing them (1502)
6340 - remove some common, well-understood CERRORs (1505)
6341 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6342 - clear "grant" flags in OST replies until OST grant code lands (1644)
6343 - fix read performance by not clobbering i_blksize on client (1598)
6344 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6345 - make LRU size a /proc tunable, clears locks when reduced (707)
6346 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6347 - server locks take references on exports now (1558)
6348 - build fixes for 2.4.20-rh trees (1663)
6349 - return an error from lov_create if all OSCs are inactive (1751)
6350 - fix import levels when a reconnect happens without a timeout (1597)
6351 - exit early from mds_open if we get a lookup error (1749)
6352 - partial page read at EOF wouldn't wait for disk before sending (1642)
6353 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6354 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6355 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6356 - write/truncate lock inversion (1639)
6357 - Don't auto-load obdclass, portals modules during cleanup (1495)
6358 - fix timestamps from jumping to "now" (1763)
6359 - extra journal assertions (1648)
6360 - add an extra multiunlink test (1771)
6361 - fix read_record/write_record API (1776)
6362 - fix leak of offset_extent, possible incorrect i_size later (1772)
6363 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6364 - Debugging for the kqswnal_get_idle_tx problems (1820)
6365 - Allow recovery to be attempted multiple times (1536)
6366 - Write out MDS last_rcvd file after it is first created (1600)
6367 - Fix tx_descriptor leak in failed transmit situations (1827)
6368 - ext3 journaling fixes for assertion failure after IO error (1871)
6369 - class_export_put() on freed export after completion AST error (1896)
6370 - Fix revalidate looping in VFS (1322)
6371 - Don't access a freed export during MDS_REINT timeout (1521)
6372 - Add open-unlink recovery support on the MDS (1673,1764)
6373 - Return an error if no MDS data was read from last_rcvd (1946)
6374 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6375 - Don't setup a disk device that doesn't match exported UUID (317)
6376 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6377 - avoid committing NULL handle in force close
6378 - local.sh is now a one-stripe LOV configuration
6379 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6380 - fix ext3 htree duplicate directory entry corruption (1516)
6381 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6382 - update to new LOV EA format (2097)
6383 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6385 2003-06-15 Phil Schwan <phil@clusterfs.com>
6388 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6389 - per-import/export recovery handling (958, 931, 959)
6390 - multiple last-rcvd slots, for serving multiple FSes (949)
6391 - connections are again shared between multiple imp/exports (963, 964)
6392 - "umount -f" would hang if any requests needed to be sent (393, 978)
6393 - avoid pinning large req buffer by copying for queued messages (989)
6394 - add "uuid" to "lctl device" command to help upcalls (991)
6395 - "open" RPCs with transnos would confuse recovery counters (1037)
6396 - do proper endian conversion of all wire messages (288, 340, 891)
6397 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6398 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6399 - fix ptlrpc_connection leak in target_handle_connect (1174)
6400 - fix import refcounting bug in OST and MDS cleanup (1134)
6401 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6402 - fix very unlikely obd_types race condition (501)
6403 - remove osc_open hack for echo_client (1187)
6404 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6405 - a failure in read_inode2 leads to deadlock (1139)
6406 - cancel ack-locks as soon as transaction is committed (1072)
6407 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6408 - make sure to commitrw after any preprw to avoid deadlock (1162)
6409 - failing to execute a file in a lustre FS would lock inode (1203)
6410 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6411 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6412 - obd_uuid2tgt would walk off the end of the list (1255)
6413 - on IA64 the osc would give portals incorrect bulk size (1258)
6414 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6415 - fix lock inversion caused by new llite matching code (1282)
6416 - limit the number of dirty pages on a client to 10MB (1286)
6417 - timed out locks were not being corrected cancelled (1289)
6418 - fix O_DIRECT above 4GB on IA-32 (1292)
6419 * major user-visible changes
6420 - fail out/fail over policy now controlled by the upcall (993)
6422 - add OBD_PING to check server availability and failure (954)
6423 - lustre messages are now sent in sending host order (288, 340, 891)
6424 - add eadatalen to MDS getattr reply (340)
6425 - OST read replies may contain second buffer, with per-page status (593)
6427 2003-03-11 Phil Schwan <phil@clusterfs.com>
6430 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6431 - failed open()s could cause deadlock; fixed (867, 869)
6432 - stop cancelling OST locks when files are closed (481)
6433 - overlapping XID spaces caused network corruption (851, 853)
6434 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6435 - setattr_raw added, to do single-RPC, server-side setattrs
6436 - lmc/lconf syntax change for OST UUIDs
6437 - fix crashy race condition between ptlrpc_free_req and osc_close
6438 - don't use request in mdc_enqueue if we hit a timeout (889)
6439 - don't set the inode i_size for regular files from the MDS (896)
6440 - handle out of order completion AST (842)
6441 - don't LBUG if a lock request times out after receiving AST (913)
6442 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6443 - if a bad lock AST arrives, send an error instead of dropping entirely
6444 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6445 - fix leak in bulk IO when only partially completed (899, 900, 926)
6446 - fix O_DIRECT for ia64 (55)
6447 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6448 - C-z after timeout could hang a process forever; fixed (977)
6450 - client-side I/O cache (678, 924, 929, 941, 970)
6452 - READPAGE and SETATTRs which don't take server-side locks get
6455 2003-02-11 Phil Schwan <phil@clusterfs.com>
6458 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6459 - Fix fsfilt_extN_readpage() to read a full page of directory
6460 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6461 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6462 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6463 - Most intent operations were removed, in favour of a new RPC mode
6464 that does a single RPC to the server and bypasses most of the VFS
6465 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6466 - Aggressively cancel local locks on DLM servers
6467 - mds_reint_unlink sends EA to the client if it's the last nlink.
6468 client uses that EA to unlink OST objects.
6469 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6470 - recursive symlinks were fixed (439)
6471 - fixed NULL deref in DEBUG_REQ
6472 - filter_update_lastobjid no longer calls sync, which annoyed extN
6473 - fixed multi-client small-writes to a single file problem (445)
6474 - fixed mtime updates during file writes (607)
6475 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6476 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6477 - fixed Linux OST size reporting problem (444, 656)
6478 - OST now updates object mtime with writes or setattr (607, 619)
6479 - client verifies file size before zeroing page past EOF (445)
6480 - OST now writes last allocated objid to disk with allocation (108)
6481 - LOV on echo now works (409)
6483 - mds_reint_unlink sends a new buffer, with the EA included. this
6484 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6485 set if a regular file was being unlinked, and it was the last link
6486 - use PtlGet from the target for bulk writes (315)
6487 - OST now updates object mtime with writes or setattr (607, 619)
6488 - LDLM now has a grant-time callback to revalidate locked items, if
6490 - Many MDS operations were reorganized to combat race conditions
6492 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6493 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6495 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6498 - Fully reactivate OST imports after reconnection (512, others)
6499 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6500 - More graceful error handling for truncating on dead OST (515)
6501 - Don't error out unless we're actually accessing dead stripes (474)
6502 - Fix garbage sizes when stripes are missing (410)
6503 - LRU counters were broken, causing constant lock purge (433, 432)
6504 - garbage on read from stripes with failed OSTs (441)
6505 - mark OSCs as active before reconnecting during recovery (438)
6506 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6507 - lfind did not preserve OST order in output (443)
6508 - symlinks cause hung clients, incorrect data (439)
6509 - stop dereferencing request after dropping refcount (457)
6510 - don't LASSERT(spin_is_locked) on non-SMP (455)
6511 - fixes for many rename() bugs
6512 - fstat didn't correctly synchronize attributes (399)
6513 - server must handle lock cancellation during blocking AST prep (487)
6514 - bulk descriptors were free()d too soon (511)
6515 - fix paths in lconf, which would load incorrect modules (451, 507)
6516 - fix confusing lconf 'host not found' error message (386)
6517 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6518 - fix race condition in mdc_blocking_ast() for inode access (526)
6519 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6520 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6521 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6522 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6523 - fix lstripe in several ways (kernel side) (527)
6524 - fix request leak in ldlm_cli_enqueue (262)
6525 - incorrect OSC was marked inactive after OST failure
6526 - call mds_fs_cleanup before unmounting filesystem (524)
6527 - fix races between taking ns_lock and ldlm_lock_change_resource
6528 - fix races updating LOV export open file list
6529 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6530 - fix recovery NULL deref in ldlm_cli_cancel_unused
6531 - fix some DLM races by using new hash table for lock handles (419)
6532 - permit the client to specify desired inodes, at replay
6533 - duplicate requests when we queue them for replay reintegration
6534 - fix last_rcvd offset calculation
6535 - sync after each recovered transaction, so we always make progress
6536 - never, not always, ERESTART requests without transnos
6537 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6539 - skip replay if the MDS says that the client is already connected
6540 - don't check for a recovery-enabled export to match lctl's UUID
6541 - don't INC_USE_COUNT for phantom exports
6542 - don't crash when cleaning up phantom exports (567)
6543 - don't double-finish or set replay data for errored mdc_open requests
6544 - abort requests when they time out, so we don't get old replies
6545 - send/receive replies for AST messages again
6546 - if the client says that it doesn't have the lock, cancel it on the
6548 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6549 mark it as destroyed, it will all work out when decref is called
6550 - fix module use counts (22, 581)
6552 - ASTs now expect a reply (server cancels lock on error reply)
6554 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6557 - fix many simultaneous client startup (392)
6558 - fix dentry->d_it clobbering
6559 - credentials weren't being shipped for readdir/getattr operations
6560 - remove invalid assertions triggered during some concurrent MD
6562 - proper Lustre versions added (336, 389)
6563 - fix memory leak for create error case (398)
6564 - fix LOV locking bug that would get cli/srv out of sync
6565 - fix echo client over LOV (409)
6566 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6567 - fix double-O_EXCL intent crash (424)
6568 - avoid sending multiple lock CANCELs (352)
6570 - MDS can do multi-client recovery (modulo bugs in new code)
6572 - many updates, edits, cleanups
6574 2002-11-18 Phil Schwan <phil@clusterfs.com>
6577 - fix null d_it dereference (346)
6578 - fix full OST/dbench hang (333)
6579 - fix permission problem with file removal (286)
6580 - fix removal of OSCs from LOV when they fail
6581 - fix NULL deref during bulk timeout (214)
6582 - fix problems related to multiple filesystems on one MDS (241)
6583 - fixed serious subtle metadata locking bugs
6584 - free locks on clients when inodes are removed due to memory
6586 - fix inode pointer in lock data (285)
6587 - partial support for multiple MDS on a single host (241)
6588 - data locks weren't cancelled at clear_inode time (290, 311)
6589 - intent locks could lead to unbounded lock growth (205)
6590 - added a maximum lock count, an LRU list, and a flusher
6591 - fix multiple rename (365)
6592 - properly abstracted the echo client
6593 - OSC locked 1 byte too many; fixed
6594 - rewrote brw callback code:
6595 - fixed recovery bugs related to LOVs (306)
6596 - fixed too-many-pages-in-one-write crash (191)
6597 - fixed (again) crash in sync_io_timeout (214)
6598 - probably fixed callback-related race (385)
6600 - Add capability to MDS protocol
6601 - LDLM cancellations and callbacks on different portals
6603 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6606 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6607 - defer open object destruction to close time (601981, 138)
6608 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6609 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6610 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6611 - don't call obd_setattr in ll_file_release for destroyed objects
6612 * protocol change to lustre_msg: move |version| and add |flags|
6613 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6614 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6615 * added replay of create, unlink, link and rename operations during
6616 MDS failover; recovery should be much more robust now
6617 * remove failed OSCs from LOVs (only lov_create uses this so far)
6618 * the lustre-HOWTO was brought (more) up to date (582544)
6620 2002-10-23 Phil Schwan <phil@clusterfs.com>
6623 - in-use dentries weren't being reused properly (617851)
6624 - prevent multiple LDLM setup (599178)
6625 - fix LOV size calculations for truncate (617853)
6626 - fix client handling of MDS intent errors (POSIX)
6627 - fix permission bug in lovstripe.c test (624321)
6628 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6629 - truncate past end of file could corrupt data
6630 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6631 - a race in recovery could return ETIMEDOUT to apps (623947)
6632 - building outside the source directory was fixed
6633 * the lustre-HOWTO was brought (more) up to date (582544)
6634 * major progress was made on recovery functionality
6636 2002-10-10 Phil Schwan <phil@clusterfs.com>
6639 - recovery deadlock fix
6640 - rm -rf causes LBUG fix (617817)
6641 - file open by multiple tasks fix (618962)
6642 - directory permissions bugs (602707 and 620007)
6643 - journal_stop fixed with locking (611313)
6644 - O_APPEND failures resolved (618273, perhaps 614459)
6645 - lconf PATH fix (619770)
6646 - IA64 build fix (621450)
6647 - RPC buffer sizes scale with amount of memory
6649 2002-10-01 Phil Schwan <phil@clusterfs.com>
6652 - locks would be cancelled without throwing away data pages,
6653 resulting in inconsistent data (605627)
6654 - inode attributes were not always being refreshed (605627, 612449)
6655 - lconf now continues to cleanup after lctl reports an error
6656 - MDS now enforces user permissions (602707)
6657 - lprocfs cleanup fixed, but not yet enabled (614157)
6658 - fixed infinite server hang, should a client not respond to an AST
6659 - avoid going into recovery if user calls readlink() with a buffer
6660 that's too small (613941)
6661 - AST RPCs no longer require replies (614867) -- this may be changed
6662 - don't crash server if client sends an IOV that's too big (611336)
6663 - fixed lock conversion deadlock (611892)
6664 - fixed the following of symlinks (614622)
6665 * recovery: the server can remove locks from a client that dies, other
6666 clients can make progress
6667 * more extN patch fixes
6668 * compile-time configurable ptlrpc buffer allocations
6670 - collaborative read cache document
6671 - Lustre Lite Performance CDR document-in-progress
6673 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6676 - fix typo in patch-2.4.18
6678 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6681 - clear ptlrpc request each time in handle_incoming_request()
6682 - unlink of files now destroys the object on the OST
6684 2002-09-19 Peter Braam <braam@clusterfs.com>
6686 * add hard link support
6687 * change obdfile creation method
6688 * kernel patch changed
6690 2002-09-19 Peter Braam <braam@clusterfs.com>
6693 - stack overflow bug in extN fixed
6695 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6697 * documentation updates
6698 - add man pages for config tools
6699 - update tests/README to describe testing with new config tools
6700 - finish metadata API descriptions
6701 * bug fixes and cleanups
6702 - statfs workaround for 16TB limit
6703 - LOV stripe allocation improved, can stripe on subset of OSTs
6704 - LOV file size/IO offset was wrong for files > 4GB in size
6705 - object EA data was being dropped, caused files to be unreadable
6706 - memory overflow with non-LOV OST caused memory corruption
6707 - fixed regression tests to work with new config tools, obdfilter
6708 - fixed bug when directory size became larger than 1 block
6709 - fixed bug (for single client case) when PWD was deleted
6710 - invalidate local directory pages when doing intent-based ops
6711 - avoid LDLM oops when lock callback contained bad data
6713 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6715 * documentation updates
6716 * bug fixes and cleanups
6717 - configuration tools
6720 - 64-bit compile warnings
6721 - 64-bit internal statfs data
6723 * test_brw on persistent OST devices
6725 * lprocfs (disabled)
6727 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6729 * documentation updates
6730 * bug fixes and cleanups
6731 * configuration tools
6733 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6736 * many small fixes to 0_5_4
6737 * io/network handling
6738 * thinkos in MDS operations
6740 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6743 * crucial basic fixes to 0.5.3
6744 * IOR, Iozone work over Elan
6747 2002-08-07 Phil Schwan <phil@clusterfs.com>
6748 * version 0_5_3, our first alpha
6749 * we use the new Portals iovs
6750 * documentation updates
6751 * bug fixes and cleanups
6752 * small changes in the DLM wire protocol
6754 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6755 * version 0_5_1 with some initial stability,
6756 * locking on MD and file I/O.
6757 * documentation updates
6758 * several bug fixes since 0.5.0
6759 * small changes in wire protocol
6761 2002-07-18 Phil Schwan <phil@clusterfs.com>
6763 * delivered as Lustre Light Alpha
6764 * fixed a crash after handling invalid MDS requests
6765 * fixed directory pages for architectures with non-4k pages sizes
6767 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6768 * release version v0_4_4
6769 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6771 * Fixes a number of interruption problems with OST operations.
6772 * Update documentation for portals header changes
6773 * Move all wire protocol structs/defines to lustre_idl.h
6774 * Fixes symlink length bug.
6775 * Add tcpdump to repository.
6777 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6778 * release version v0_4_3
6779 * Fixes statfs for inodes on extN.
6780 * Fixes bug in runtests which would delete /etc/hosts.
6781 * Use 64-bit object IDs wherever possible (not into VFS though)
6782 Remove ost_get_info, which is unused by lustre, and out of date.
6784 2002-07-03 Peter Braam <braam@clusterfs.com>
6785 * release version v0_4_2 Fixes a lookup error (type not passed)
6786 * move forward to head of Portals
6787 * move forward to latest Lustre kernel
6789 2002-06-25 Peter Braam <braam@clusterfs.com>
6790 * release version v0_4_1. Hopefully stable on single node use.