3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.10.1.el7 (RHEL7.6)
9 4.4.120-92.70 (SLES12 SP2)
10 4.4.162-94.69 (SLES12 SP3)
11 4.4.0-142 (Ubuntu 16.04)
12 4.15.0-45 (Ubuntu 18.04)
13 vanilla linux 4.6.7 (ZFS only)
14 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
15 * Client known to build on unpatched kernels (others may also work):
16 2.6.32-696.18.7.el6 (RHEL6.9)
17 2.6.32-754.6.3.el6 (RHEL6.10)
18 3.10.0-862.14.4.el7 (RHEL7.5)
19 3.10.0-957.10.1.el7 (RHEL7.6)
20 4.14.0-49.13.1.el7a (RHEL7.5)
21 4.14.0-115.2.2.el7a (RHEL7.6)
22 4.18.0-80.el8 (RHEL8.0)
23 4.4.120-92.70 (SLES12 SP2)
24 4.4.162-94.69 (SLES12 SP3)
25 4.12.14-95.13 (SLES12 SP4)
26 4.4.0-142 (Ubuntu 16.04)
27 4.15.0-45 (Ubuntu 18.04)
28 vanilla linux 4.15.0 (ZFS only)
29 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
30 * Recommended ZFS / SPL version: 0.7.13
31 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
32 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
33 stack overflow. For more information, please refer to bugzilla 17630.
34 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
35 the re-exporting nodes to be mounted with "32bitapi" mount option
36 * Note for Ubuntu 14.04.5:
37 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
38 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
42 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
43 for currently supported client and server kernel versions.
44 * Server known to build on the following kernels (others may also work):
45 2.6.32-696.18.7.el6 (RHEL6.9)
46 3.10.0-862.14.4.el7 (RHEL7.5)
47 3.10.0-957.10.1.el7 (RHEL7.6)
48 4.4.120-92.70 (SLES12 SP2)
49 4.4.162-94.69 (SLES12 SP3)
50 4.4.0-142 (Ubuntu 16.04)
51 4.15.0-45 (Ubuntu 18.04)
52 vanilla linux 4.6.7 (ZFS only)
53 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
54 * Client known to build on unpatched kernels (others may also work):
55 2.6.32-696.18.7.el6 (RHEL6.9)
56 2.6.32-754.6.3.el6 (RHEL6.10)
57 3.10.0-862.14.4.el7 (RHEL7.5)
58 3.10.0-957.10.1.el7 (RHEL7.6)
59 4.14.0-49.13.1.el7a (RHEL7.5)
60 4.14.0-115.2.2.el7a (RHEL7.6)
61 4.4.120-92.70 (SLES12 SP2)
62 4.4.162-94.69 (SLES12 SP3)
63 4.12.14-95.13 (SLES12 SP4)
64 4.4.0-142 (Ubuntu 16.04)
65 4.15.0-45 (Ubuntu 18.04)
66 vanilla linux 4.15.0 (ZFS only)
67 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
68 * Recommended ZFS / SPL version: 0.7.13
69 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
70 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
71 stack overflow. For more information, please refer to bugzilla 17630.
72 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
73 the re-exporting nodes to be mounted with "32bitapi" mount option
74 * Note for Ubuntu 14.04.5:
75 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
76 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
81 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
82 for currently supported client and server kernel versions.
83 * Server known to build on the following kernels (others may also work):
84 2.6.32-696.18.7.el6 (RHEL6.9)
85 3.10.0-862.14.4.el7 (RHEL7.5)
86 3.10.0-957.10.1.el7 (RHEL7.6)
87 4.4.120-92.70 (SLES12 SP2)
88 4.4.162-94.69 (SLES12 SP3)
89 4.4.0-142 (Ubuntu 16.04)
90 4.15.0-45 (Ubuntu 18.04)
91 vanilla linux 4.6.7 (ZFS only)
92 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
93 * Client known to build on unpatched kernels (others may also work):
94 2.6.32-696.18.7.el6 (RHEL6.9)
95 2.6.32-754.6.3.el6 (RHEL6.10)
96 3.10.0-862.14.4.el7 (RHEL7.5)
97 3.10.0-957.10.1.el7 (RHEL7.6)
98 4.14.0-49.13.1.el7a (RHEL7.5)
99 4.14.0-115.2.2.el7a (RHEL7.6)
100 4.4.120-92.70 (SLES12 SP2)
101 4.4.162-94.69 (SLES12 SP3)
102 4.12.14-95.13 (SLES12 SP4)
103 4.4.0-142 (Ubuntu 16.04)
104 4.15.0-45 (Ubuntu 18.04)
105 vanilla linux 4.15.0 (ZFS only)
106 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
107 * Recommended ZFS / SPL version: 0.7.13
108 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
109 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
110 stack overflow. For more information, please refer to bugzilla 17630.
111 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
112 the re-exporting nodes to be mounted with "32bitapi" mount option
113 * Note for Ubuntu 14.04.5:
114 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
115 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
119 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
120 for currently supported client and server kernel versions.
121 * Server known to build on the following kernels (others may also work):
122 2.6.32-696.18.7.el6 (RHEL6.9)
123 3.10.0-862.14.4.el7 (RHEL7.5)
124 3.10.0-957.el7 (RHEL7.6)
125 4.4.120-92.70 (SLES12 SP2)
126 4.4.162-94.69 (SLES12 SP3)
127 4.4.0-131 (Ubuntu 16.04)
128 4.15.0-32 (Ubuntu 18.04)
129 vanilla linux 4.6.7 (ZFS only)
130 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
131 * Client known to build on unpatched kernels (others may also work):
132 2.6.32-696.18.7.el6 (RHEL6.9)
133 2.6.32-754.6.3.el6 (RHEL6.10)
134 3.10.0-862.14.4.el7 (RHEL7.5)
135 3.10.0-957.el7 (RHEL7.6)
136 4.14.0-49.13.1.el7a (RHEL7.5)
137 4.14.0-115.2.2.el7a (RHEL7.6)
138 4.4.120-92.70 (SLES12 SP2)
139 4.4.162-94.69 (SLES12 SP3)
140 4.4.0-131 (Ubuntu 16.04)
141 4.15.0-32 (Ubuntu 18.04)
142 vanilla linux 4.15.0 (ZFS only)
143 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
144 * Recommended ZFS / SPL version: 0.7.9
145 * Tested with ZFS / SPL version: 0.7.9
146 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
147 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
148 stack overflow. For more information, please refer to bugzilla 17630.
149 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
150 the re-exporting nodes to be mounted with "32bitapi" mount option
151 * Note for Ubuntu 14.04.5:
152 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
153 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
155 --------------------------------------------------------------------------------
158 2018-04-03 Intel Corporation
160 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
161 for currently supported client and server kernel versions.
162 * Server known to build on patched kernels:
163 2.6.32-431.29.2.el6 (RHEL6.5)
164 2.6.32-504.30.3.el6 (RHEL6.6)
165 2.6.32-573.26.1.el6 (RHEL6.7)
166 2.6.32-642.15.1.el6 (RHEL6.8)
167 2.6.32-696.18.7.el6 (RHEL6.9)
168 3.10.0-693.21.1.el7 (RHEL7.4)
169 3.0.101-0.47.71 (SLES11 SP3)
170 3.0.101-107 (SLES11 SP4)
171 3.12.74-60.64.40 (SLES12 SP1)
172 4.4.114-92.64 (SLES12 SP2)
173 4.4.114-94.11 (SLES12 SP3)
174 3.13.0-101 (Ubuntu 14.04, ZFS only)
175 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
176 4.4.0-116 (Ubuntu 16.04)
177 vanilla linux 4.6.7 (ZFS only)
178 * Client known to build on unpatched kernels:
179 2.6.32-431.29.2.el6 (RHEL6.5)
180 2.6.32-504.30.3.el6 (RHEL6.6)
181 2.6.32-573.26.1.el6 (RHEL6.7)
182 2.6.32-642.15.1.el6 (RHEL6.8)
183 2.6.32-696.18.7.el6 (RHEL6.9)
184 3.10.0-693.21.1.el7 (RHEL7.4)
185 3.0.101-0.47.71 (SLES11 SP3)
186 3.0.101-107 (SLES11 SP4)
187 3.12.74-60.64.40 (SLES12 SP1)
188 4.4.114-92.64 (SLES12 SP2)
189 4.4.114-94.11 (SLES12 SP3)
190 3.13.0-101 (Ubuntu 14.04)
191 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
192 4.4.0-116 (Ubuntu 16.04)
194 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
195 * Recommended ZFS / SPL version: 0.7.6
196 * Tested with ZFS / SPL version: 0.7.6
197 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
198 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
199 stack overflow. For more information, please refer to bugzilla 17630.
200 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
201 the re-exporting nodes to be mounted with "32bitapi" mount option
202 * Note for Ubuntu 14.04.5:
203 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
204 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
206 --------------------------------------------------------------------------------
208 2017-07-05 Intel Corporation
210 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
211 for currently supported client and server kernel versions.
212 * Server known to build on patched kernels:
213 2.6.32-431.29.2.el6 (RHEL6.5)
214 2.6.32-504.30.3.el6 (RHEL6.6)
215 2.6.32-573.26.1.el6 (RHEL6.7)
216 2.6.32-642.15.1.el6 (RHEL6.8)
217 2.6.32-696.3.1.el6 (RHEL6.9)
218 3.10.0-514.21.1.el7 (RHEL7.3)
219 3.0.101-0.47.71 (SLES11 SP3)
220 3.0.101-100 (SLES11 SP4)
221 3.12.74-60.64.40 (SLES12 SP1)
222 4.4.59-92.17 (SLES12 SP2)
223 vanilla linux 4.6.7 (ZFS only)
224 * Client known to build on unpatched kernels:
225 2.6.32-431.29.2.el6 (RHEL6.5)
226 2.6.32-504.30.3.el6 (RHEL6.6)
227 2.6.32-573.26.1.el6 (RHEL6.7)
228 2.6.32-642.15.1.el6 (RHEL6.8)
229 2.6.32-696.3.1.el6 (RHEL6.9)
230 3.10.0-514.21.1.el7 (RHEL7.3)
231 3.0.101-0.47.71 (SLES11 SP3)
232 3.0.101-100 (SLES11 SP4)
233 3.12.74-60.64.40 (SLES12 SP1)
234 4.4.59-92.17 (SLES12 SP2)
236 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
237 * Recommended ZFS / SPL version: 0.6.5.9
238 * Tested with ZFS / SPL version: 0.6.5.9
239 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
240 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
241 stack overflow. For more information, please refer to bugzilla 17630.
242 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
243 the re-exporting nodes to be mounted with "32bitapi" mount option
245 --------------------------------------------------------------------------------
247 12-07-2016 Intel Corporation
249 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
250 for currently supported client and server kernel versions.
251 * Server known to build on patched kernels:
252 2.6.32-431.29.2.el6 (RHEL6.5)
253 2.6.32-504.30.3.el6 (RHEL6.6)
254 2.6.32-573.26.1.el6 (RHEL6.7)
255 2.6.32-642.11.1.el6 (RHEL6.8)
256 3.10.0-514.el7 (RHEL7.3)
257 3.0.101-0.47.71 (SLES11 SP3)
258 3.0.101-80 (SLES11 SP4)
259 3.12.62-60.64.8 (SLES12 SP1)
260 vanilla linux 4.5.7 (ZFS only)
261 * Client known to build on unpatched kernels:
262 2.6.32-431.29.2.el6 (RHEL6.5)
263 2.6.32-504.30.3.el6 (RHEL6.6)
264 2.6.32-573.26.1.el6 (RHEL6.7)
265 2.6.32-642.11.1.el6 (RHEL6.8)
266 3.10.0-514.el7 (RHEL7.3)
267 3.0.101-0.47.71 (SLES11 SP3)
268 3.0.101-80 (SLES11 SP4)
269 3.12.62-60.64.8 (SLES12 SP1)
271 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
272 * Recommended ZFS / SPL version: 0.6.5.8
273 * Tested with ZFS / SPL version: 0.6.5.8
274 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
275 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
276 stack overflow. For more information, please refer to bugzilla 17630.
277 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
278 the re-exporting nodes to be mounted with "32bitapi" mount option
280 --------------------------------------------------------------------------------
282 02-29-2016 Intel Corporation
284 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
285 for currently supported client and server kernel versions.
286 * Server known to build on patched kernels:
287 2.6.32-431.29.2.el6 (RHEL6.5)
288 2.6.32-504.30.3.el6 (RHEL6.6)
289 2.6.32-573.12.1.el6 (RHEL6.7)
290 3.10.0-327.3.1.el7 (RHEL7.2)
291 3.0.101-0.47.71 (SLES11 SP3)
292 3.0.101-68 (SLES11 SP4)
293 vanilla linux 4.2.1 (ZFS only)
294 * Client known to build on unpatched kernels:
295 2.6.32-431.29.2.el6 (RHEL6.5)
296 2.6.32-504.30.3.el6 (RHEL6.6)
297 2.6.32-573.12.1.el6 (RHEL6.7)
298 3.10.0-327.3.1.el7 (RHEL7.2)
299 3.0.101-0.47.71 (SLES11 SP3)
300 3.0.101-68 (SLES11 SP4)
303 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
304 * Recommended ZFS / SPL version: 0.6.4.2
305 * Tested with ZFS / SPL version: 0.6.4.2
306 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
307 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
308 stack overflow. For more information, please refer to bugzilla 17630.
309 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
310 the re-exporting nodes to be mounted with "32bitapi" mount option
312 --------------------------------------------------------------------------------
314 03-10-2015 Intel Corporation
316 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
317 for currently supported client and server kernel versions.
318 * Server known to build on patched kernels:
319 2.6.32-431.29.2.el6 (RHEL6.5)
320 2.6.32-504.8.1.el6 (RHEL6.6)
321 3.0.101-0.46 (SLES11 SP3)
322 * Client known to build on unpatched kernels:
323 2.6.32-431.29.2.el6 (RHEL6.5)
324 2.6.32-504.8.1.el6 (RHEL6.6)
325 3.10.0-123.20.1.el7 (RHEL7)
326 3.0.101-0.46 (SLES11 SP3)
327 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
328 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
329 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
330 stack overflow. For more information, please refer to bugzilla 17630.
331 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
332 the re-exporting nodes to be mounted with "32bitapi" mount option
334 Severity : enhancement
336 Description: control OST-index in IDIF via ROCOMPAT flag.
337 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
338 last_rcvd file. For new formatted OST device, it will be auto set;
339 for the case of upgrading from old OST device, you can enable it
340 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
341 enabled, for new created OST-object, its IDIF-in-LMA will contain
342 the OST-index; for the existing OST-object, the OSD will convert
343 old format IDIF as new format IDIF with OST-index stored in the
344 LMA EA when accessing such OST-object or via OI scrub. Once such
345 flag is enabled, it cannot be reverted back, so the system cannot
346 be downgraded to the orignal incompatible version.
348 --------------------------------------------------------------------------------
350 07-30-2014 Intel Corporation
352 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
353 for currently supported client and server kernel versions.
354 * Server known to build on patched kernels:
355 2.6.32-431.20.3.el6 (RHEL6.5)
356 3.0.101-0.31 (SLES11 SP3)
357 * Client known to build on unpatched kernels:
358 2.6.32-431.20.3.el6 (RHEL6.5)
359 3.0.101-0.31 (SLES11 SP3)
362 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
363 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
364 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
365 stack overflow. For more information, please refer to bugzilla 17630.
366 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
367 the re-exporting nodes to be mounted with "32bitapi" mount option
369 --------------------------------------------------------------------------------
371 10-31-2013 Intel Corporation
373 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
374 * Server builds for kernels:
375 2.6.32-358.18.1.el6 (RHEL6.4)
376 * Client builds for unpatched kernels:
377 2.6.32-358.18.1.el6 (RHEL6.4)
378 3.0.92-0.8 (SLES11 SP3)
380 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
381 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
382 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
383 stack overflow. For more information, please refer to bugzilla 17630.
384 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
385 the re-exporting nodes to be mounted with "32bitapi" mount option
387 --------------------------------------------------------------------------------
389 05-30-2013 Intel Corporation
391 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
392 * Server builds for kernels:
393 2.6.32-358.6.2.el6 (RHEL6.4)
394 * Client builds for unpatched kernels:
395 2.6.32-358.6.2.el6 (RHEL6.4)
396 3.0.74-0.6.6 (SLES11 SP2)
398 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
399 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
400 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
401 stack overflow. For more information, please refer to bugzilla 17630.
402 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
403 the re-exporting nodes to be mounted with "32bitapi" mount option
405 --------------------------------------------------------------------------------
407 9-30-2012 Whamcloud, Inc.
409 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
410 * Server support for kernels:
411 2.6.32-279.5.1.el6 (RHEL6)
412 * Client support for unpatched kernels:
413 2.6.32-279.5.1.el6 (RHEL6)
415 2.6.18-238.19.1.el5 (RHEL5)
416 2.6.16.60-0.69.1 (SLES10)
417 * Recommended e2fsprogs version: 1.41.90.wc4
418 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
419 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
420 stack overflow. For more information, please refer to bugzilla 17630.
421 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
422 the re-exporting nodes to be mounted with "32bitapi" mount option
424 Severity : enhancement
426 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
428 Severity : enhancement
430 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
432 Severity : enhancement
434 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
436 Severity : enhancement
438 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
440 Severity : enhancement
442 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
444 --------------------------------------------------------------------------------
446 04-30-2012 Whamcloud, Inc.
448 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
449 * Server support for kernels:
450 2.6.32-220.4.2.el6 (RHEL6)
451 * Client support for unpatched kernels:
452 2.6.32-220.4.2.el6 (RHEL6)
454 2.6.18-238.19.1.el5 (RHEL5)
455 2.6.16.60-0.69.1 (SLES10)
456 * Recommended e2fsprogs version: 1.41.90.wc4
457 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
458 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
459 stack overflow. For more information, please refer to bugzilla 17630.
460 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
461 the re-exporting nodes to be mounted with "32bitapi" mount option
463 Severity : enhancement
465 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
467 Severity : enhancement
469 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
471 Severity : enhancement
473 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
475 Severity : enhancement
477 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
479 --------------------------------------------------------------------------------
481 09-30-2011 Whamcloud, Inc.
483 * Server support for kernels:
484 2.6.18-238.19.1.el5 (RHEL5)
485 2.6.32-131.6.1.el6 (RHEL6)
486 * Client support for unpatched kernels:
487 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
488 2.6.18-238.19.1.el5 (RHEL5)
489 2.6.32-131.6.1.el6 (RHEL6)
490 2.6.29.4-167.fc11 (FC11)
491 2.6.16.60-0.69.1 (SLES10)
493 * Recommended e2fsprogs version: 1.41.90.wc3
494 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
495 Lustre filesystem with 4K stack may cause a stack overflow. For more
496 information, please refer to bugzilla 17630.
497 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
498 mounted with "32bitapi" mount option
500 Severity : enhancement
501 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
503 Severity : enhancement
504 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
506 Severity : enhancement
507 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
510 Frequency : only with SLES10
512 Description: Use OFED "KMP" provided by Novell
513 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
514 Lustre is now built against this package. That means you need to
515 install the ofed-kmp package from Novell for the patchless client.
516 Note that the ofed-kmp that Novell ships may not exactly match the
517 kernel version but should still be compatible.
519 Severity : enhancement
521 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
522 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
524 --------------------------------------------------------------------------------
526 2010-07-15 Oracle, Inc.
528 * Server support for kernels:
529 2.6.18-164.11.1.el5 (RHEL 5)
530 2.6.18-164.11.1.0.1.el5 (OEL 5)
531 * Client support for unpatched kernels:
532 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
533 2.6.18-164.11.1.el5 (RHEL 5),
534 2.6.18-164.11.1.0.1.el5 (OEL 5)
535 2.6.16.60-0.42.8 (SLES 10),
537 2.6.29.4-167.fc11 (FC11)
538 * Recommended e2fsprogs version: 1.41.10-sun2
540 Severity : enhancement
542 Description: Distribute lfs_migrate script with lustre
543 Details : The lfs_migrate mentioned in the manual is now included in the
544 lustre tree in lustre/scripts and will distributed in lustre RPMs.
546 Severity : enhancement
548 Description: Failover nid list added to import proc file.
550 Severity : enhancement
552 Description: "som_preview" mount option required to use SOM
553 Details : The clients will need to use som_preview mount option
557 Description: Lustre does not do 1MB IOs to HW RAID
558 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
560 Severity : enhancement
562 Description: lfs quota output is not very convenient for awk/sed-parsing
563 Details : Some positions in lfs quota output table could be empty or
564 non-empty which made it hard to parse it with scripts, now
565 a dash is put instead of space where there is not supposed
570 Description: Update to OFED 1.5.1
571 Details : OFED 1.5.1 will be used in environments that don't use
572 vendor supplied OFED.
576 Description: Remove lnet. prefix from lctl params display
577 Details : Also renamed lnet.memused to lnet_memused to
582 Description: Use conf_param -d to remove permanant settings.
586 Description: Add list_param -R
587 Details : lctl list_param now supports recursive mode.
590 Bugzilla : 21829, 22312
591 Description: Fix llobdstat.
592 Details : 1. Now updating the write_bytes counter
593 2. Port script fixes from 1.8.
597 Description: decrease the usage of memory on clients.
598 Details : 1. On clients, recycle dentries and inodes unused.
599 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
604 Description: llobdstat fix and enhancement
605 Details : add a counter to set a limit to how many samples will be returned
606 fix a wildcard in the path to limit to obdfilter stats only
610 Description: lfs check only allowed for root
611 Details : Code cleanups and added a regresion test of non-root use of
616 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
617 Details : Fix hr->hr_index overload.
621 Description: Fix FIEMAP functionality.
625 Description: FC11 patchless client support added
629 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
630 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
634 Description: New acc-sm script and various test-framework enhancments.
638 Description: Previously landed patch was reverted.
642 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
646 Description: Add ldiskfs maxdirsize mount option.
650 Description: Fix mkfs.lustre support for 16TB LUNs.
654 Description: Print mmp_check_interval
655 Details : print mmp_check_interval and make it possible to abort
656 mount operation in case it takes too long.
660 Description: Add cascading_rw to lustre mpi tests.
664 Description: Store SOM data in EA.
665 Details : Also several SOM related cleanups.
669 Description: Add refcount to md_open_data
673 Description: Use refcount on lov_request_set to prevent memory corruption.
677 Description: Statahead: use parent dir's i_mutex for synchronization.
681 Description: loadgen improvements.
685 Description: Enable peer health detection in ksocklnd and ko2iblnd.
689 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
693 Description: Fix race on obd_devcie:md_stats
697 Description: Fix SLES 11 client builds.
698 Details : Resolved uninitalized pointer gcc warnings.
702 Description: Allow non-root access for "lfs check".
703 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
707 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
709 Severity : enhancement
711 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
712 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
716 Description: Sanity test_27p fails.
717 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
718 Sanity tests 27 were cleaned up and improved.
722 Description: Unify req format on client/servers
723 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
724 buffers and swabbers approach..
729 Description: drop join file stuff.
731 Severity : enhancement
733 Description: Add 2.6.27 and SLES11 (patchless client) support
737 Description: Fix in ptlrpc_expire_one_request() to print the signed time
740 Severity : enhancement
742 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
745 Severity : enhancement
747 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
749 Severity : enhancement
751 Description: correctly handle big reply message.
752 Details : send LNet event if reply is bigger than buffer and adjust this buffer
757 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
759 Severity : enhancement
761 Description: Add support for OEL5.
765 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
767 Severity : enhancement
769 Description: "weak-modules" support
770 Details : Implement "weak-modules" support which enables kernel modules
771 to be used with any kernel that implements the same kABI. In
772 order to achieve this modules are now installed in
773 /lib/modules/$(uname -r)/updates/kernel on all distributions.
777 Description: client could not reconnect to OST because of an active request.
778 Details : The client now retries to reconnect to the same server, if a
779 connect request failed with EBUSY or -EAGAIN. Test case 35b
780 conf-sanity.sh has been created.
784 Description: Don't allow make backward step on assignin osc next id.
785 Details : race between allocation next id and ll_sync thread can be cause
786 of set wrong osc next id and can be kill valid ost objects.
790 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
794 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
798 Description: Error handling in osc_statfs_interpret() has been improved.
799 Details : Check in osc_statfs_interpret() for EBADR.
803 Description: Do not send statfs() requests to OSTs disabled by administrator.
804 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
806 Severity : enhancement
808 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
810 Severity : enhancement
812 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
814 Severity : enhancement
816 Description: Enhancement for lfs(1) command to use numeric uid/gid.
820 Description: Improved cleanup procedure for conf-sanity test_46a.
824 Description: Do not start bulk transfer, if client tried to reconnect.
825 Details : Do not start bulk transfer, if client tried to reconnect,
826 but failed due to active rpcs. Check for exp_abort_active_req
827 has been implemented.
831 Description: Prevent inconsistences between linux and lustre mount structures.
832 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
833 Make the sleep interruptible.
835 Severity : enhancement
837 Description: provide server to client comms path
838 Details : server to client communications path via new LDLM_SET_INFO rpc
842 Description: truncate starts GFP_FS allocation under transaction causing deadlock
843 Details : ldiskfs_truncate calls grab_cache_page which may start page
844 allocation under an open transaction. This may lead to
845 calling prune_icache with consequent lustre reentrance.
849 Description: send correctly lsm on open replay
850 Details : MDS is trust to LSM size on replay open, but client can set wrong size
855 Description: enable client interface failover
856 Details : When a child reconnects from another NID, properly update export
857 nid hash position and ldlm reverse import.
859 Severity : enhancement
861 Description: Communicate OST degraded/readonly state via statfs to MDS
862 Details : Flags in the statfs returned from OSTs indicate whether the
863 OST is in a degraded RAID state, or if the filesystem has
864 turned read-only after a filesystem error is detected.
869 Description: don't panic if EPROTO was hit when reading symlink
870 Details : correctly handling request reference in error cases.
872 Severity : enhancement
874 Description: Add LustreNetLink, a kernel-userspace communcation path.
876 Severity : enhancement
878 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
883 Description: access to llog context before init.
884 Details : move handling CATALOGS file at osc layer and forbid access to llog
888 Frequency : rare, only when client is writing to OST on local node
890 Description: Avoid deadlock for local client writes
891 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
892 memory freeing context. This allows OST threads to set the
893 PF_MEMALLOC flag on task structures in order to allocate memory
894 from reserved pools and complete IO.
895 Use GFP_HIGHUSER for OST allocations for non-local client writes,
896 so that the OST threads generate memory pressure and allow
897 inactive pages to be reclaimed.
899 Severity : enhancement
901 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
906 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
907 Details : move ldlm namespace creation in setup phase to avoid grab
908 _lprocfs_lock with cli_sem held.
912 Description: Temporarily disable grant shrink.
913 Details : Disable the feature for debugging.
916 Frequency : only when formatting test filesystems
918 Description: Unable to run several mkfs.lustre on loop devices at the same time
919 Details : mkfs.lustre returns error 256 on the concurrent loop devices
920 formatting. The solution is to proper handle the error.
922 Severity : enhancement
924 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
926 Severity : enhancement
928 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
933 Description: resolve race between obd_disconnect and class_disconnect_exports
934 Details : if obd_disconnect will be called to already disconnected export he
935 forget release one reference and osc module can't unloaded.
937 Severity : enhancement
939 Description: Allow tuning service thread via /proc
940 Details : For each service a new
941 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
942 created that can be used to set min/max thread counts, and get the
943 current number of running threads.
948 Descriptoin: don't return error if have partially created objects for file.
949 Details : lov_update_create_set uses set->set_success as index for created
950 objects, so if some requests will be failed, they will have hole
951 at end of array and qos_shrink_lsm for allocate correct lsm.
953 Severity : enhancement
955 Description: Update OFED support to 1.4.1
960 Description: Reduce small size read RPC
961 Details : Set read-ahead limit for every file and only do read-ahead when
962 available read-ahead pages are bigger than 1M to avoid small size
965 Severity : enhancement
967 Description: move AT tunable parameters for more consistent usage
968 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
970 Severity : enhancement
972 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
973 Details : allow skip disconnected ost for send statfs request and hide error
979 Description: fix racy locking of mballoc block bitmaps causing BUG
980 Details : The locking of the mballoc buddy bitmap and the in-memory
981 block bitmap was using two different spin locks in some
982 cases. This made it possible to incorrectly access the
983 mballoc bitmap while another process was modifying it,
984 causing a sanity assertion to fail. While no on-disk corruption
985 was reported, there was some risk of this happening.
988 Frequency : rare, on llog test 6
990 Descriptoin: don't allow connect to already connected import
991 Details : allowing connect to already connected import is hide connecting problem.
996 Descriptoin: allow kill process which wait statahead result
997 Details : for some reasons 'ls' can stick in waiting result from statahead,
998 in this case need way for kill this process.
1000 Severity : enhancement
1002 Description: Add state history info file, enhance import info file
1003 Details : Track import connection state changes in a new osc/mdc proc file;
1004 add overview-type data to the osc/mdc import proc file.
1006 Severity : enhancement
1008 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1010 Severity : enhancement
1012 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1017 Descriptoin: OST grant too much space to client even there are not enough space.
1018 Details : Client will shrink its grant cache to OST if there are no write
1019 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1020 this grant cache if there are already not enough avaible space
1021 (left_space < total_clients * 32M).
1024 Frequency : start MDS on uncleanly shutdowned MDS device
1026 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1027 Details : stay in waiting mds<>ost recovery finished produce random bugs
1028 due race between two ll_sync thread for one lov target. send
1029 ACTIVATE event only if connect realy finished and import have
1033 Frequency : rare, connect and disconnect target at same time
1035 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1036 Details : don't call obd_disconnect under lov_lock. this long time
1037 operation and can block ptlrpcd which answer to connect request.
1042 Descriptoin: don't lose wakeup for imp_recovery_waitq
1043 Details : recover_import_no_retry or invalidate_import and import_close can
1044 both sleep on imp_recovery_waitq, but we was send only one wakeup
1048 Frequency : always with long access acl
1050 Descriptoin: mds can't pack reply with long acl.
1051 Details : mds don't control size of acl but they limited by reint/getattr
1054 Severity : enhancement
1056 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1058 Severity : enhancement
1060 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1063 Frequency : start MDS on uncleanly shutdowned MDS device
1065 Descriptoin: aborting recovery hang on MDS
1066 Details : don't throttle destroy RPCs for the MDT.
1069 Frequency : on remount
1071 Description: external journal device not working after the remount
1072 Details : clear dev_rdonly flag for external journal devices in
1078 Description: shutdown vs evict race
1079 Details : client_disconnect_export vs connect request race.
1080 if client will evicted at this time - we start invalidate
1081 thread without referece to import and import can be freed
1085 Frequency : rare, need acl's on inode.
1087 Description: client can't handle ost additional correctly
1088 Details : if ost was added after client connected to mds client can have
1089 hit lnet_try_match_md ... to big messages to wide striped files.
1090 in this case need teach client to handle config events about add
1091 lov target and update client max ea size at that event.
1093 Severity : enhancement
1095 Description: Changelogs
1096 Details : Changelogs are a lightweight mechanism to track filesystem
1097 metadata and namespace changes. The changelog is recorded
1098 permanently on the MDTs, and is periodically "consumed" / purged
1099 when records are no longer needed.
1101 Severity : enhancement
1103 Description: compact fld format with extents
1104 Details : Store range of seq rather than every seq in FLD. Seq
1105 controller update FLD rather than clients. In Case of CMD, mdt0
1106 has FLD, all other metadata server act as non persistent proxy
1107 for FLD queries and cache fld entries in fld cache.
1112 Description: don't skip ost target if they assigned to file
1113 Details : Drop slow OSCs if we can, but not for requested start idx.
1114 This means "if OSC is slow and it is not the requested
1115 start OST, then it can be skipped, otherwise skip it only
1116 if it is inactive/recovering/out-of-space.
1120 Description: more cleanup in mds_lov
1121 Details : not send LOV EA under replay, we can't know about they size at this
1122 time. Don't allow client connect to mds before any ost connected,
1123 for avoid problems with LOV EA size and returning EIO to client.
1125 Severity : enhancement
1127 Description: Interoperability at server side (Disk interoperability)
1129 Severity : enhancement
1131 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1133 Severity : enhancement
1135 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1137 Severity : enhancement
1139 Description: New client IO stack (CLIO).
1141 Severity : enhancement
1143 Description: Commit on sharing. Eliminate inter-client dependencies between
1144 uncommitted transactions by doing transaction commits.
1145 Thereby clients may recovery independently.
1148 Frequency : Create a symlink file with a very long name
1150 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1151 Details : If there is no extra space in the request for early cancels,
1152 ldlm_req_handles_avail() returns 0 instead of a negative value.
1154 Severity : enhancement
1156 Description: Add /proc entry for import status
1157 Details : The mdc, osc, and mgc import directories now have
1158 an import directory that contains useful import data for debugging
1159 connection problems.
1161 Severity : enhancement
1163 Description: Re-disable certain /proc logging
1164 Details : Enable and disable client's offset_stats, extents_stats and
1165 extents_stats_per_process stats logging on the fly.
1168 Frequency : Only on FC kernels 2.6.22+
1170 Description: oops in statahead
1171 Details : Do not drop reference count for the dentry from VFS when lookup,
1172 VFS will do that by itself.
1174 Severity : enhancement
1176 Description: Generic /proc file permissions
1177 Details : Set /Proc file permissions in a more generic way to enable non-
1178 root users operate on some /proc files.
1182 Description: Hitting mdc_commit_close() ASSERTION
1183 Details : Properly handle request reference release in
1184 ll_release_openhandle().
1188 Description: quota recovery deadlock during mds failover
1189 Details : This patch includes att18982, att18236, att18237 in bz14840.
1191 1. fix osts hang when mds does failover with quotaon
1192 2. prevent watchdog storm when osts threads wait for the
1197 Frequency : only patchless client
1198 Description: add workaround for race between add/remove dentry from hash
1200 Severity : enhancement
1202 Description: Allow OST glimpses to return PW locks
1206 Description: LBUG when llog conf file is full
1207 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1211 Description: Prevent import from entering FULL state when server in recovery
1215 Description: service mount cannot take device name with ":"
1216 Details : Only when device name contains ":/" will mount treat it as
1222 Description: replace ptlrpcd with the statahead thread to interpret the async
1223 statahead RPC callback
1227 Frequency : on recovery
1228 Description: I/O failures after umount during fail back
1229 Details : if client reconnected to restarted server we need join to recovery
1230 instead of find server handler is changed and process self
1231 eviction with cancel all locks.
1233 Severity : enhancement
1235 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1237 Severity : enhancement
1239 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1241 Severity : enhancement
1243 Description: Upcall on Lustre log has been dumped
1244 Details : Allow for a user mode script to be called once a Lustre log has
1245 been dumped. It passes the filename of the dumped log to the
1246 script, the location of the script can be specified via
1247 /proc/sys/lnet/debug_log_upcall.
1252 Description: avoid idr_remove called for id which is not allocated.
1253 Details : Move assigment s_dev for clustered nfs to end of initialization,
1254 to avoid problem with error handling.
1259 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1260 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1261 into CONN_USED_HASH and this prodice warning when put connection
1262 again in unused hash.
1265 Severity : enhancement
1267 Description: File striping can now be set to use an arbitrary pool of OSTs.
1269 Severity : enhancement
1271 Description: Export bytes_read/bytes_write count on OSC/OST.
1275 Description: Early reply size mismatch, MGC loses connection
1276 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1277 the connect flags are properly negotiated.
1282 Description: quotas are not honored with O_DIRECT
1283 Details : all writes with the flag O_DIRECT will use grants which leads to
1284 this problem. Now using OBD_BRW_SYNC to guard this.
1288 Description: add quota statistics
1289 Details : 1. sort out quota proc entries and proc code.
1290 2. add quota statistics
1292 Severity : enhancement
1294 Description: enable quota support for HEAD.
1298 Description: Properly propagate oinfo flags from lov to osc for statfs
1299 Details : restore missing copy oi_flags to lov requests.
1301 Severity : enhancement
1303 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1304 : l_getgroups(8), lst(8), routerstat(8)
1305 Details : included man pages for llobdstat(8), llstat(8),
1306 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1308 Severity : enhancement
1310 Description: configure's --enable-quota should check the
1311 : kernel .config for CONFIG_QUOTA
1312 Details : configure is terminated if --enable-quota is passed but
1313 : no quota support is in kernel
1317 Description: Remove portals compatibility
1318 Details : Remove portals compatibility, not interoperable with releases
1323 Description: Resolve device initialization race
1324 Details : Prevent proc handler from accessing devices added to the
1325 obd_devs array but yet be intialized.
1327 Severity : enhancement
1329 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1331 Severity : enhancement
1333 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1338 Description: Using wrong pointer in osc_brw_prep_request
1339 Details : Access to array[-1] can produce panic if kernel compiled with
1340 CONFIG_PAGE_ALLOC enabled
1344 Description: Client runs out of low memory
1345 Details : Consider only lowmem when counting initial number of llap pages
1349 Description: *optional* service tags registration
1350 Details : if the "service tags" package is installed on a Lustre node
1351 When the filesystem is mounted, a local-node service tag will
1352 be created. See http://inventory.sun.com/ for more information
1353 about the Service Tags asset management system.
1357 Description: Kernel BUG tries to release flock
1358 Details : Lustre does not destroy flock lock before last reference goes
1359 away. So always drop flock locks when client is evicted and
1360 perform unlock regardless of successfulness of speaking to MDS.
1364 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1368 Description: sanity test 65a fails if stripecount of -1 is set
1369 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1374 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1375 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1378 Severity : enhancement
1380 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1384 Description: do not process already freed flock
1385 Details : flock can possibly be freed by another thread before it reaches
1386 to ldlm_flock_completion_ast.
1390 Description: LBUG during stress test
1391 Details : Need properly lock accesses the flock deadlock detection list.
1395 Description: oops in page fault handler
1396 Details : kernel page fault handler can return two special 'pages' in error
1397 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1401 Description: timeout with invalidate import.
1402 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1403 should be handled by ptlrpcd. This produce long age waiting and
1404 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1406 Severity : enhancement
1408 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1412 Description: enable MGS and MDT services start separately
1413 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1414 a co-located MGS without starting the MGS, which is a complement
1415 to 'nosvc' mount option.
1419 Frequency : after recovery
1420 Description: precreate to many object's after del orphan.
1421 Details : del orphan st in oscc last_id == next_id and this triger growing
1422 count of precreated objects. Set flag LOW to skip increase count
1423 of precreated objects.
1427 Frequency : rare, on clear nid stats
1428 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1429 Details : when clean nid stats sometimes try destroy live entry,
1430 and this produce panic in free.
1434 Description: Stack overflow during MDS log replay
1435 ease stack pressure by using a thread dealing llog_process.
1439 Description: wait until IO finished before start new when do lock cancel.
1440 Details : VM protocol want old IO finished before start new, in this case
1441 need wait until PG_writeback is cleared until check dirty flag and
1442 call writepages in lock cancel callback.
1444 Severity : enhancement
1446 Description: using special macro for print time and cleanup in includes.
1450 Description: mds_mfd_close() ASSERTION(rc == 0)
1451 Details : In mds_mfd_close(), we need protect inode's writecount change
1452 within its orphan write semaphore to prevent possible races.
1456 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1460 Frequency : rare, on shutdown ost
1461 Description: don't hit live lock with umount ost.
1462 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1463 use shrink_dcache_sb instead.
1467 Description: don't panic with use echo client
1468 Details : echo client pass NULL as client nid pointer and this produce null
1469 pointer dereference.
1473 Description: fix build on ppc32
1474 Details : compile code with -m64 flag produce wrong object file for ppc32.
1478 Description: add message levels for liblustreapi
1482 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1483 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1484 will result in an -EAGAIN and in the exit call path we call
1485 alloc_rr() which will with increasing aggressiveness attempt to
1486 aquire precreated objects on the minimum number of required OSCs.
1490 Description: Use old size assignment to avoid deadlock
1491 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1492 the scheduling while holding a spinlock. We do not need locking
1493 for size in ll_update_inode() because size is only updated from
1494 the MDS for directories or files without objects, so there is no
1495 other place to do the update, and concurrent access to such inodes
1496 are protected by the inode lock.
1500 Description: resolve "_IOWR redefined" build error on SLES10
1504 Description: dump the memory debugging after all modules are unloaded to
1505 suppress false negative in conf_sanity test 39
1507 Severity : enhancement
1509 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1514 Description: fix SLES kernel versioning
1515 Details : the kernel version for our SLES 10 kernel did not include a "-"
1516 before the "smp" at the end. while this was not a problem in
1517 general, it did mean that software trying to use the kernel
1518 version to try to detect a vendor specific kernel would fail.
1519 this was most evident by the OFED build scripts.
1523 Description: Don't update lov_desc members until making sure they are valid
1524 Details : When updating lov_desc members via proc fs, need fix their
1525 validities before doing the real update.
1529 Description: don't put request into delay list while invalidate in flight.
1530 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1531 import in flight. this produce timeout for invalidate and sometimes
1532 can cause stale data.
1536 Frequency : on ppc only
1537 Description: not convert ost objects for directory because it's not exist.
1538 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1542 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1544 Severity : enhancement
1546 Description: Update to support 2.6.22.14 vanilla kernel.
1550 Frequency : rare, on recovery
1551 Description: read procfs can produce deadlock in some situation
1552 Details : Holding lprocfs lock which send rpc can produce block for destroy
1553 obd objects and this also block reconnect with -EALREADY. This
1554 isn't fix all lprocfs bugs - but make it rare.
1556 Severity : enhancement
1558 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1561 Frequency : frequent on X2 node
1563 Description: mdc_set_open_replay_data LBUG
1564 Details : Set replay data for requests that are eligible for replay.
1568 Description: lustre_mgs: operation 101 on unconnected MGS
1569 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1570 MGC, and late on MGC cannot successfully connect to MGS and a lot
1571 of the error messages complaining that MGS is not connected.
1575 Frequency : on network error
1576 Description: panic with double free request if network error
1577 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1578 but it's true only for synchronus enqueue, for async enqueue
1579 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1582 Severity : enhancement
1584 Description: client-side metadata stat-ahead during readdir(directory readahead)
1585 Details : perform client-side metadata stat-ahead when the client detects
1586 readdir and sequential stat of dir entries therein
1589 Frequency : on start mds
1591 Description: Implement get_info(last_id) in obdfilter.
1594 Frequency : occasional
1596 Description: Correctly check stale fid, not start epoch if ost not support SOM
1597 Details : open with flag O_CREATE need set old fid in op_fid3 because
1598 op_fid2 was overwritten with new generated fid, but mds can answer
1599 with one of these two fids and both is not stale. Setattr
1600 incorrectly started an epoch and assume will be called
1601 done_writing, but without SOM done_writing ever being called.
1604 Frequency : rare, depends on device drivers and load
1606 Description: MDS or OSS nodes crash due to stack overflow
1607 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1608 In some cases, in conjunction with device drivers that use a lot
1609 of stack the MDS (or possibly OSS) service threads could overflow
1610 the stack. One change which was identified to consume additional
1611 stack has been reworked to avoid the extra stack usage.
1614 Frequency : occasional
1616 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1617 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1618 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1619 oscc_has_objects and see OSC has no objests and call
1620 oscc_internal_create to resend create request.
1622 Severity : enhancement
1624 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1626 Severity : enhancement
1628 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1631 Frequency : very rare
1633 Description: Fix replay if there is an un-replied request and open
1634 Details : In some cases, older replay request will revert the
1635 mcd->mcd_last_xid on MDS which is used to record the client's
1636 latest sent request.
1638 Severity : enhancement
1640 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1642 Severity : enhancement
1644 Description: Add rhel5 support to HEAD.
1646 Serverity : enhancement
1648 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1653 Description: Don't allow skipping OSTs if index has been specified.
1654 Details : Don't allow skipping OSTs if index has been specified, make
1655 locking in internal create lots better.
1659 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1660 Details : access to bitfield in structure is always rounded to long
1661 and this produce problem with not atomic change any bit.
1665 Description: Lustre make rpms failed.
1666 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1667 builds Lustre rpms from tarball.
1670 Frequency : rare on shutdown ost
1672 Description: If llog cancel was not send before clean_exports phase, this can
1673 produce deadlock in llog code.
1674 Details : If llog thread has last reference to obd and call class_import_put
1675 this produce deadlock because llog_cleanup_commit_master wait when
1676 last llog_commit_thread exited, but this never success because was
1677 called from llog_commit_thread.
1681 Description: allow userland application know is lost one of stripes.
1682 Details : fill lvb_blocks with error code on ost and return it to
1683 application if error flag found.
1687 Description: NULL lov_tgts causing MDS oops
1688 Details : more safe checks for NULL lov_tgts for avoid oops.
1690 Severity : enhancement
1692 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1696 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1697 Details : If an OST is down the MDS will hang indefinitely in
1698 obd_statfs_rqset() waiting for the statfs data. While for
1699 MDS QOS usage of statfs, it should not stuck in waiting.
1701 Severity : enhancement
1703 Description: remote_acl support
1704 Details : Support ACL-based permission check for remote user.
1705 Support setfacl/getfacl for remote user with the utils
1706 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1707 the system "{s,g}etfacl" utils.
1709 Severity : enhancement
1711 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1713 Severity : enhancement
1715 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1719 Description: make mgs_setparam() handle fsname containing dash
1720 Details : fsname containing a dash does not work with lctl conf_param
1722 Severity : enhancement
1724 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1726 Severity : enhancement
1728 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1732 Frequency : rare, at shutdown
1733 Description: access already free / zero obd_namespace.
1734 Details : if client_disconnect_export was called without force flag set,
1735 and exist connect request in flight, this can produce access to
1736 NULL pointer (or already free pointer) when connect_interpret
1737 store ocd flags in obd_namespace.
1741 Frequency : only at startup
1742 Description: not alloc memory with spinlock held.
1743 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1744 if any spinlock held.
1746 Severity : enhancement
1748 Description: make lustre randomly fail allocating memory
1749 Details : Make lustre randomly failed allocating memory for testing purpose.
1751 Severity : enhancement
1753 Description: lost problems with lov objid file
1754 Details : Fixes some scability and access to not inited memory problems
1755 in work with lov objdid file.
1760 Description: lfs find does not continue on file error
1761 Details : Continue other files processing when a file/dir is absent.
1765 Description: Inconsistent usage of lustre_pack_reply()
1766 Details : Standardize the usage of lustre_pack_reply() such that it
1767 always generate a CERROR on failure.
1770 Frequency : occasional
1772 Description: MDS hang or stay in waiting lock
1773 Details : If client receive lock with CBPENDING flag ldlm need send lock
1774 cancel as separate rpc, to avoid situation when cancel request
1775 can't processed due all i/o threads stay in wait lock.
1779 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1783 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1787 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1791 Description: Update to RHEL4 latest kernel.
1793 Severity : enhancement
1795 Description: Build SLES10 patchless client fails
1796 Details : The configure was broken by run ./configure with
1797 --with-linux-obj=.... argument for patchless client. When the
1798 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1799 can't search header adequately. Use absolute path such as
1800 -I($LINUX)/include instead.
1804 Description: interrupt oig_wait produce painc on resend.
1805 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1806 private set, and this produce situation when rq_ptlrpcd_data not
1807 copyed to new allocated request and triggered LBUG on assert
1808 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1809 ptlrpcd set if request is changed and can be safety changed to use
1814 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1815 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1816 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1817 these macros and replace them with their difinitions instead.
1821 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1822 Details : Update which_patch & target file for SLES10 latest kernel.
1825 Description: Remove client patches from SLES 10 kernel.
1826 Details : This causes SLES 10 clients to behave as patchless clients
1827 even on a Lustre-patched (server) kernel.
1829 Severity : enhancement
1831 Description: self-adjustable client's lru lists
1832 Details : use adaptive algorithm for managing client cached locks lru
1833 lists according to current server load, other client's work
1834 pattern, memory activities, etc. Both, server and client
1835 side namespaces provide number of proc tunables for controlling
1838 Severity : enhancement
1840 Description: light-weight GSS support
1841 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1842 and reduce performance overhead.
1844 Severity : enhancement
1846 Description: Linux keyring support
1847 Details : Support using service of Linux keyring for Lustre GSS internal
1848 context refresh/cache mechanism.
1852 Description: Fix errors in lfs documentation
1853 Details : Fixes man pages
1857 Description: don't use GFP_* in generic Lustre code.
1858 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1862 Description: obdclass is limited by single OBD_ALLOC(idarray)
1863 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1867 Description: warnings with build patchless client with vanila 2.6.19 and up
1868 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1869 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1873 Frequency : Only for SLES
1874 Description: sanity_quota fail test_1
1875 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1876 file, which makes TSTID[2] inunique.
1880 Frequency : Only for SLES9
1881 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1882 Details : Update target/ChangeLog/which_patch .
1886 Frequency : Only for test_10 in sanity-quota.sh
1887 Description: a bug in quota test code
1888 Details : checking a test flag in a wrong place
1892 Frequency : Only for test_13 in sanity-quota.sh
1893 Description: a bug in quota test script
1894 Details : shouldn't check overfull allocated quota
1898 Description: jbd statistics
1899 Details : Port older jbd statistics patch for sles10
1901 Severity : enhancement
1903 Description: Add printing inode info into message about error in writepage.
1908 Description: reply_lock_interpret crash due to race with it and lock cancel.
1909 Details : Do not replay locks that are being cancelled. Do not reference
1910 locks by their address during replay, just by their handle.
1914 Frequency : When flocks are used.
1915 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1916 Details : Flock locks might destroy just granted lock if it could be merged
1917 with another existing flock, this is done in completion handler,
1918 so teach ldlm_cli_enquque_fini that this is a valid case for
1924 Description: Oops in read and write path when failing to allocate lock.
1925 Details : Check if lock allocation failed and return error back.
1929 Description: lstripe command fails for valid OST index
1930 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1931 instead of lov->desc.ld_active_tgt_count.
1935 Description: sanity.sh failed test 103
1936 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1941 Description: df doesn't work properly if diskfs blocksize != 4K
1942 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1946 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1947 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1949 Severity : enhancement
1951 Description: Adaptive timeouts
1952 Details : RPC timeouts adapt to changing server load and network
1953 conditions to reduce resend attempts and improve recovery time.
1957 Description: llapi_file_create() does not allow some changes
1958 Details : add llapi_file_open() that allows specifying the mode and
1959 open flags, and also returns an open file handle.
1963 Description: merge and cleanup kernel patches.
1964 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1968 Description: Add journal checksum support.(Kernel part)
1969 Details : The journal checksum feature adds two new flags i.e
1970 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1971 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1972 indicates that the commit block contains the checksum for
1973 the blocks described by the descriptor blocks. Now commit
1974 record can be sent to disk without waiting for descriptor
1975 blocks to be written to disk. This behavior is controlled
1976 using JBD2_FEATURE_ASYNC_COMMIT flag.
1980 Description: OSS needs mutliple precreate threads
1981 Details : Add ability to start more than one create thread per OSS.
1985 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1986 Details : execute lfs setstripe on client
1990 Description: mds_obd_create error creating tmp object
1991 Details : When the user sets quota on root, llog will be affected and can't
1992 create files and write files.
1995 Frequency : Always on ia64 patchless client, and possibly others.
1997 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1998 Details : This allows the patchless client to be loaded on architectures
1999 without this export.
2003 Description: RedHat Update kernel for RHEL5
2004 Details : Add the kernel config file for RHEL5.
2008 Description: RedHat Update kernel for RHEL5
2009 Details : Modify the kernel config file more closer RHEL5.
2013 Description: Build failure against Centos5 (RHEL5)
2014 Details : Define PAGE_SIZE when it isn't present.
2018 Description: test_7 on sanity-quota.sh can't be used on separate machines
2019 Details : add facet to handle it
2023 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2024 Details : with NFS, the anon dentry's parent was set to itself in
2025 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2026 corresponding dentry other than use rec->ur_name.
2028 Severity : enhancement
2030 Description: lfs setstripe enhancement
2031 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2035 Description: enable data checksumming by default
2036 Details : enable checksum by default, allow --disable-checksum
2037 configure option and "-o nochecksum" mount option. Checksums
2038 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2042 Description: lustre support for RHEL5
2043 Details : Add support for RHEL5.
2047 Description: Client eviction due to failover config
2048 Details : after a connection loss, the lustre client should attempt to
2049 reconnect to the last active server first before trying the
2050 other potential connections.
2054 Description: when mds and osts use different quota unit(32bit and 64bit),
2055 quota will be released repeatly.
2056 Details : void sending multiple quota reqs to mds, which will keep the
2057 status between the reqs.
2061 Description: osts not allocated evenly to files
2062 Details : change the condition to increase offset_idx
2065 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2067 Description: Data corruption for OSTs that are formatted larger than 2TB
2069 Details : When generating the bio request for lustre file writes the
2070 sector number would overflow a temporary variable before being
2071 used for the IO. The data reads correctly from Lustre (which
2072 will overflow in a similar manner) but other file data or
2073 filesystem metadata may be corrupted in some cases.
2077 Description: Tune the kernel for good SCSI performance.
2078 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2079 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2084 Description: rewrite ext2-derived code in obdclass/uuid.c
2085 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2086 from scratch preserving functionality.
2090 Description: "lfs find -obd UUID" prints directories
2091 Details : "lfs find -obd UUID" will return all directory names instead
2092 of just file names. It is incorrect because the directories
2093 do not reside on the OSTs.
2097 Description: use i_size_read and i_size_write in 2.6 port
2098 Details : replace inode->i_size access with i_size_read/write()
2101 Frequency : when removing large files
2103 Description: scheduling issue during removal of large Lustre files
2104 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2105 It causes scheduling issues when removing large files (17TB in the
2109 Frequency : only with liblustre clients on XT3
2111 Description: evictions taking too long
2112 Details : allow llrd to evict clients directly on OSTs
2115 Frequency : only on ppc
2117 Description: /proc/fs/lustre/devices broken on ppc
2118 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2119 We should make sure the type of 'index' is loff_t and then cast
2120 explicitly as needed below. Do not assign an explicitly cast
2124 Frequency : only for rhel5
2126 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2127 Details : Modify the target file & which_kernel.
2129 Severity : enhancement
2131 Description: omit set fsid for export NFS
2132 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2133 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2137 Frequency : Always, for kernels after 2.6.16
2138 Description: Fix warning idr_remove called for id=.. which is not allocated.
2139 Details : Last kernels save old s_dev before kill super and not allow
2140 to restore from callback - restore it before call kill_anon_super.
2144 Description: buffer overruns could theoretically occur
2145 Details : llapi_semantic_traverse() modifies the "path" argument by
2146 appending values to the end of the origin string, and a
2147 overrun may occur. Adding buffer overrun check in liblustreapi.
2151 Description: Fix error on 'ls .' at the top of the Lustre mount.
2152 Details : Don't revalidate dentry if it is a root dentry.
2156 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2157 Details : Modify vm-tunables-rhel4.patch.
2161 Description: Kernel config for 2.6.18-vanilla.
2162 Details : Modify targets/2.6-vanilla.target.in.
2163 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2164 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2165 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2166 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2170 Description: improve handling recoverable errors
2171 Details : if request processig with error which can be recoverable on server
2172 request should be resend, otherwise page released from cache and
2177 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2178 Details : Modify target file & which_patch.
2179 A flaw was found in the IA32 system call emulation provided
2180 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2181 value could be stored in the %RAX register, which could trigger an
2182 out-of-bounds system call table access. An untrusted local user
2183 could exploit this flaw to run code in the kernel
2184 (ie a root privilege escalation). (CVE-2007-4573).
2188 Description: O_DIRECT bypasses client statistics.
2189 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2190 read_ahead_stats were updated. copied the stats section from
2191 osc_send_oap_rpc() into async_internal().
2195 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2199 Description: change order of libsysio includes
2200 Details : '#include sysio.h' should always come before '#include xtio.h'
2204 Description: handle "serious error: objid * already exists" more gracefully
2205 Details : If LAST_ID value on disk is smaller than the objects existing in
2206 the O/0/d* directories, it indicates disk corruption and causes an
2207 LBUG(). If the object is 0-length, then we should use the existing
2208 object. This will help to avoid a full fsck in most cases.
2210 Severity : enhancement
2212 Description: adapt the lustre_config script to support the upgrade case
2213 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2214 server targets to 1.6 in parallel.
2218 Description: To avoid grant space > avaible space when the disk is almost
2219 full. Without this patch you might see the error "grant XXXX >
2220 available" or some LBUG about grant, when the disk is almost
2222 Details : In filter_check_grant, for non_grant cache write, we should
2223 check the left space by if (*left > ungranted + bytes), instead
2224 of (*left > ungranted), because only we are sure the left space
2225 is enough for another "bytes", then the ungrant space should be
2226 increase. In client, we should update cl_avail_grant only there
2227 is OBD_MD_FLGRANT in the reply.
2231 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2232 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2233 described in CVE-2007-4573. This problem could allow untrusted
2234 local users to gain root access.
2237 Frequency : when using O_DIRECT and quotas
2239 Description: Incorrect file ownership on O_DIRECT output files
2240 Details : block usage reported by 'lfs quota' does not take into account
2241 files that have been written with O_DIRECT.
2246 Description: touch file failed when fs is not full
2247 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2248 otherwise we can get ENOSP while fs is not full.
2252 Description: parallel lock callbacks
2253 Details : Instead of sending blocking and completion callbacks as separated
2254 requests, adding them to a set and sending in parallel.
2257 Frequency : only for Cray XT3
2258 Bugzilla : 12829/13455
2259 Description: Changing primary group doesn't change the group lustre assigns to
2261 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2262 overridden with the primary group provided by the group upcall,
2263 whereas the supplied fsgid can be trusted if it is in the list of
2264 supplementary groups returned by the group upcall.
2266 Severity : enhancement
2268 Description: Allow masking D_WARNING, D_ERROR messages from console
2269 Details : Console messages can now be disabled via lnet.printk.
2273 Description: User code with malformed file open parameter crashes client node
2274 Details : Before packing join_file req, all the related reference should be
2275 checked carefully in case some malformed flags cause fake
2276 join_file req on client.
2280 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2281 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2282 mode is equal to the granted mode. However, it should also check
2283 that LDLM_FL_CP_REQD is not set.
2287 Description: Heavy nfs access might result in deadlocks
2288 Details : After ELC code landed, it is now improper to enqueue any mds
2289 locks under och_sem, because enqueue might want to decide to
2290 cancel open locks for same inode we are holding och_sem for.
2294 Description: Client eviction while running blogbench
2295 Details : A lot of unlink operations with concurrent I/O can lead to a
2296 deadlock causing evictions. To address the problem, the number of
2297 oustanding OST_DESTROY requests is now throttled to
2298 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2299 callbacks are processed in priority.
2303 Description: enable ACLs on MDS by default
2304 Details : ACLs must be enabled on MDS by default.
2307 Frequency : PPC/PPC64 only
2309 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2310 Details : fix duplicated definitions between asm-ppc64/types.h and
2311 lustre_types.h on PPC.
2314 Frequency : PPC/PPC64 only
2316 Description: asm-ppc/segment.h does not exist
2317 Details : fix compile issue on PPC.
2321 Description: better handle error messages in extents code
2324 Frequency : RHEL4 only
2326 Description: mkfs is very slow on IA64/RHEL4
2327 Details : A performance regression has been discovered in the MPT Fusion
2328 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2329 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2330 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2332 Severity : enhancement
2334 Description: SNMP support enhancement
2335 Details : Adding total number of sampled request for an MDS node in snmp
2338 Severity : enhancement
2340 Description: Optimize ldlm waiting list processing for PR extent locks
2341 Details : When processing waiting list for read extent lock and meeting read
2342 lock that is same or wider to it that is not contended, skip
2343 processing rest of the list and immediatelly return current
2344 status of conflictness, since we are guaranteed there are no
2345 conflicting locks in the rest of the list.
2349 Description: Time out and refuse to reconnect
2350 Details : When the failover node is the primary node, it is possible
2351 to have two identical connections in imp_conn_list. We must
2352 compare not conn's pointers but NIDs, otherwise we can defeat
2353 connection throttling.
2357 Description: port llog fixes from b1_6 into HEAD
2358 Details : Port llog reference couting and some llog cleanups from b1_6
2359 (bug 10800) into HEAD, for protect from panic and access to already
2360 free llog structures.
2364 Description: Detect stride IO mode in read-ahead
2365 Details : When a client does stride read, read-ahead should detect that and
2366 read-ahead pages according to the detected stride pattern.
2370 Description: data checksumming impacts single node performance
2371 Details : add support for several checksum algorithm. Currently, only CRC32
2372 and Adler-32 are supported. The checksum type can be changed on
2373 the fly via /proc/fs/lustre/osc/*/checksum_type.
2377 Description: use adler32 for page checksums
2378 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2383 Description: build for x2 fails
2384 Details : fix compile issue on Cray systems.
2388 Description: Properly match for duplicate locks
2389 Details : Due to different lock order from skiplists code, we need to
2390 traverse entire list for now
2393 Frequency : only on PPC/SLES10
2395 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2396 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2397 be defined. Add a hack in mkfs_lustre.c to work around this compile
2402 Description: LASSERT on MDS when client holding flock lock dies
2403 Details : ldlm pool logic depends on number of granted locks equal to
2404 number of released locks which is not true for flock locks, so
2405 just exclude such locks from consideration.
2409 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2410 Details : Use fsfilt_sync() for both the whole filesystem sync and
2411 individual file sync to eliminate dangerous inode locking
2412 with I_LOCK that can lead to a deadlock.
2416 Description: performance in 1.6.3
2417 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2418 sure that 1MB requests are not fragmented by the block layer.
2420 Severity : enhancement
2422 Description: organize the server-side client stats on per-nid basis
2423 Details : Change the structure of stats under obdfilter and mds to
2431 The "uuid"s file would list the uuids of _active_ exports.
2432 And the clear entry is to clear all stats and stale nids.
2434 Severity : enhancement
2436 Description: eliminate client locks in face of contention
2437 Details : file contention detection and lockless i/o implementation
2438 for contended files.
2442 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2445 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2447 Description: MDS may lose file striping (and hence file data) in some cases
2448 Details : If there are additional extended attributes stored on the MDS,
2449 in particular ACLs, SELinux, or user attributes (if user_xattr
2450 is specified for the client mount options) then there is a risk
2451 of attribute loss. Additionally, the Lustre file striping
2452 needs to be larger than default (e.g. striped over all OSTs),
2453 and an additional attribute must be stored initially in the
2454 inode and then increase in size enough to be moved to the
2455 external attribute block (e.g. ACL growing in size) for file
2460 Description: skiplist implementation simplification
2461 Details : skiplists are used to group compatible locks on granted list
2462 that was implemented as tracking first and last lock of each
2463 lock group the patch changes that to using doubly linked lists
2467 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2468 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2469 because LASSERT on both the data supplied by a client, and the
2470 data on disk is dangerous and incorrect.
2472 Severity : enhancement
2474 Description: Slow truncate/writes to huge files at high offsets.
2475 Details : Directly associate cached pages to lock that protect those pages,
2476 this allows us to quickly find what pages to write and remove
2477 once lock callback is received.
2481 Description: more ldlm soft lockups
2482 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2483 starve other threads from the resource lock for a long time in
2484 case of long waiting queue, so change the debug level from
2485 D_OTHER to the less frequently used D_INFO.
2487 Severity : enhancement
2489 Description: add -gid, -group, -uid, -user options to lfs find
2493 Description: Hung threads in invalidate_inode_pages2_range
2494 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2495 once one is completed. As a result, some RPCs are stuck in the
2496 queue and are never sent.
2500 Description: filter threads hungs on waiting journal commit
2501 Details : Cleanup filter group llog code, then only filter group llog will
2502 be only created in the MDS/OST syncing process.
2506 Description: Procfs and llog threads access destoryed import sometimes.
2507 Details : Sync the import destoryed process with procfs and llog threads by
2508 the import refcount and semaphore.
2510 Severity : enhancement
2512 Description: openlock cache of b1_6 port to HEAD
2517 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2518 Details : If insertion of an extent fails, then discard the inode
2519 preallocation and free data blocks else it can lead to duplicate
2524 Description: don't always update ctime in ext3_xattr_set_handle()
2525 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2526 In some cases the ctime should not be updated, for example for
2527 2.0->1.8 compatibility it is necessary to delete an xattr and it
2528 should not update the ctime.
2532 Bugzilla : 15713/16362
2533 Description: Assertion in iopen_connect_dentry in 1.6.3
2534 Details : looking up an inode via iopen with the wrong generation number can
2535 populate the dcache with a disconneced dentry while the inode
2536 number is in the process of being reallocated. This causes an
2537 assertion failure in iopen since the inode's dentry list contains
2538 both a connected and disconnected dentry.
2542 Description: assertion failure in ldlm_handle2lock()
2543 Details : fix a race between class_handle_unhash() and class_handle2object()
2544 introduced in lustre 1.6.5 by bug 13622.
2549 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2550 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2555 Description: oops in sd_iostats_seq_show()
2556 Details : unloading/reloading the scsi low level driver triggers a kernel
2557 bug when trying to access the sd iostat file.
2562 Description: Kernel panics during QLogic driver reload
2563 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2564 patch, causing memory corruption.
2569 Description: journal_dev option does not work in b1_6
2570 Details : pass mount option during pre-mount.
2572 Severity : enhancement
2574 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2575 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2576 extent information of a file. It can be used to map logical blocks
2577 in a file to physical blocks in the block device.
2581 Description: LDLM soft lockups - improvement
2582 Details : It is be possible to send the lock handle along with each read
2583 or write request because the client is already doing a lock match
2584 itself so there isn't any reason the OST should have to re-do that
2588 Frequency : only on Cray X2
2590 Description: X2 build failures
2591 Details : fix build failures on Cray X2.
2595 Description: xid & resent requests
2596 Details : Initialize RPC XID from clock at startup (randomly if clock is
2599 Severity : enhancement
2601 Description: Add lustre_start utility to start or stop multiple Lustre servers
2606 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2607 Details : In case of memory pressure, list_del() can be called twice on
2608 req->rq_history_list, causing a kernel oops.
2612 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2613 Details : fix stack overflow in the distributed lock manager by defering
2614 export eviction after a failed AST to the elt thread instead of
2615 handling it in the dlm interpret routine.
2619 Description: Convert some comments to new format.
2620 Details : Update documenting comments to match doxygen conventions.
2624 Description: Grammar fixes.
2625 Details : A couple of trivial spelling fixes.
2629 Description: OSD_COUNTERS-mandatory
2630 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2631 to simplify the code.
2635 Description: simplify cmm_device freeing logic.
2636 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2640 Description: Add lockdep support to dt_object_operations locking interface.
2641 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2642 parameter indicating lock ordering. Update mdd code to use new
2647 Description: Introduce failloc constants for lockless IO tests.
2648 Details : Add two new failloc constants to test lockless IO. Only one of
2649 them in implemented---another is checked in yet to be landed
2654 Description: Add lockdep support for inode mutex.
2655 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2656 in the situations where Lustre has to lock more than one inode
2661 Description: Add optional invariants checking support.
2662 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2663 configure switch. This macro is to be used for consistency and
2664 sanity checks that are too expensive to be left in `production'
2669 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2670 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2675 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2676 Details : Use spin_lock_nested() in (the only) situation where more than
2677 one ldlm_lock is locked simultaneously. Also, fix possible
2678 dead-lock in ldlm_lock_change_resource() by enforcing particular
2683 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2684 Details : Instead of specifying each ldlm_lock call-back through separate
2685 parameter, wrap them into struct ldlm_callback_suite.
2689 Description: Kill join_lru obd method and its callers.
2690 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2691 in memory---a requirement implemented through ->o_join_lru() obd
2692 method in HEAD. Remove this method and its users.
2696 Description: Add asynchronous ldlm ENQUEUE completion handler.
2697 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2698 case that stock ldlm_completion_ast() cannot handle as it waits
2699 until lock is granted. Introduce new ldlm_completion_ast_async()
2700 for this. Also comment ldlm_completion_ast().
2704 Description: ldlm_error <-> errno conversion.
2705 Details : Add functions to map (rather arbitrary) between LDLM error codes
2706 and standard errno values. CLIO needs this to prevent LDLM specific
2707 constants from escaping ldlm and osc.
2711 Description: Kill unused ldlm_handle2lock_ns() function.
2712 Details : Kill unused ldlm_handle2lock_ns() function.
2716 Description: Add lu_ref support to ldlm_lock
2717 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2718 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2719 and ldlm_resource. LDLM interface has to be changed, because code that
2720 releases a reference on a lock, has to "know" what reference this is.
2721 In the most frequent case
2723 lock = ldlm_handle2lock(handle);
2725 LDLM_LOCK_PUT(lock);
2727 no changes are required. When any other reference (received _not_ from
2728 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2729 instead of LDLM_LOCK_PUT().
2731 Arguably, changes are pervasive, and interface requires some discipline
2732 for proper use. On the other hand, it was very instrumental in finding
2733 a few leaked lock references.
2737 Description: Add ldlm_lock_addref_try().
2738 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2739 attempts to addref a lock that might be being canceled
2744 Description: Add ldlm_weigh_callback().
2745 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2746 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2747 hard-coded `number of pages' logic.
2751 Description: Add lockdep annotations to llog code.
2752 Details : Use appropriately tagged _nested() locking calls in the places
2753 where llog takes more than one ->lgh_lock lock.
2757 Description: Add loi_kms_set().
2758 Details : Wrap kms updates into a helper function.
2762 Description: Constify instances of struct lsm_operations.
2763 Details : Constify instances of struct lsm_operations.
2767 Description: lu_conf support.
2768 Details : On a server, a file system object is uniquely identified
2769 by a fid, which is sufficient to locate and load all object
2770 state (inode). On a client, on the other hand, more data are
2771 necessary instantiate an object. Change lu_object_find() and
2772 friends to take additional `lu_conf' argument describing object.
2773 Typically this includes layout information.
2777 Description: lu_context fixes.
2778 Details : Introduce new lu_context functions that are needed on the client
2779 side, where some system threads (ptlrpcd) are shared by multiple
2780 modules, and so cannot be stopped during module shutdown.
2784 Description: Add start and stop methods to lu_device_type_operations.
2785 Details : Introduce two new methods in lu_device_type_operations, that are
2786 invoked when first instance of a given type is created and last one
2787 is destroyed respectively. This is need by CLIO.
2791 Description: Add lu_ref support to struct lu_device.
2792 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2793 track leaked references.
2797 Description: Introduce lu_kmem_descr.
2798 Details : lu_kmem_descr and its companion interface allow to create
2799 and destroy a number of kmem caches at once.
2803 Description: Fix lu_object finalization race.
2804 Details : Fix a race between lu_object_find() finding an object and its
2805 concurrent finalization. This race is (most likely) not possible
2806 on the server, but might happen on the client.
2810 Description: Introduce lu_ref interface.
2811 Details : lu_ref is a debugging module allowing to track references to
2812 a given object. It is quite cpu expensive, and has to be
2813 explicitly enabled with --enable-lu_ref. See usage description
2818 Description: Factor lu_site procfs stats into a separate function.
2819 Details : Separate lu_site stats printing code into a separate function
2820 that can be reused on a client.
2824 Description: Constify instances of struct {lu,dt,md}_device_operations.
2825 Details : Constify instances of struct {lu,dt,md}_device_operations.
2829 Description: Introduce struct md_site and move meta-data specific parts of
2830 struct lu_site here.
2831 Details : Move md-specific fields out of struct lu_site into special struct
2832 md_site, so that lu_site can be used on a client.
2836 Description: Kill mdd_lov_destroy().
2837 Details : Remove unused mdd code.
2841 Description: Add st_block checking to multistat.c.
2842 Details : Add st_block checking to multistat.c.
2846 Description: Add lu_ref support to struct obd_device.
2847 Details : Add lu_ref tracking to obd_device.
2851 Description: Kill obd_set_fail_loc().
2852 Details : Remove unused code.
2856 Description: Add special type for ptlrpc_request interpret functions.
2857 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2858 there. Actual usage will be in CLIO.
2862 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2863 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2864 doesn't work with lockdep.
2868 Description: Add rwv.c test program.
2869 Details : New testing program exercising readv(2) and writev(2) (Qian).
2873 Description: Add sendfile.c test program.
2874 Details : New testing program exercising sendfile(2) (Jay).
2878 Description: Ratelimit a message that can be very frequent.
2879 Details : Ratelimit a memory allocation failure message that can
2884 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2885 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2889 Description: Liblustre build fixes.
2890 Details : Liblustre build fixes.
2894 Description: libcfs: add cfs_{need,cond}_resched() interface.
2895 Details : libcfs: add cfs_{need,cond}_resched() definition and
2896 implementations for Linux, NT, and liblustre.
2898 Severity : enhancement
2900 Description: More exported tunables for mballoc
2901 Details : Add support for tunable preallocation window and new tunables for
2902 large/small requests
2906 Description: Detect corruption of block bitmap and checking for preallocations
2907 Details : Checks validity of on-disk block bitmap. Also it does better
2908 checking of number of applied preallocations. When corruption is
2909 found, it turns filesystem readonly to prevent further corruptions.
2913 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2914 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2915 window to 0. If it does hit cache miss, and read-pattern is still
2916 stride-io mode, does not reset the stride window, but also does not
2917 increase the stride window length in this case.
2921 Frequency : only for big-endian servers
2922 Description: Check if system is big-endian while mounting fs with extents feature
2923 Details : Mounting a filesystem with extents feature will fail on big-endian
2924 systems since ext3-based ldiskfs is not supported on big-endian
2925 systems. This can be over-riden with "bigendian_extents" mount option.
2927 Severity : enhancement
2929 Description: The root squash functionality
2930 Details : A security feature, which is to prevent users from being able
2931 to mount lustre on their desktop, run as root, and delete
2932 all of the files in the filesystem. The goal is accomplished by
2933 remapping user id (UID) and group id (GID) of the root user to
2934 a UID and GID specified by the system administartor via Lustre
2935 configuration management server (MGS). The functionality also
2936 allows to specify sets of clients for which the remapping does
2941 Description: Excessive recovery window
2942 Details : With AT enabled, the recovery window can be excessively long (6000+
2943 seconds). To address this problem, we no longer use
2944 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2945 timeout no longer depends on the service time, it is set to
2946 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2947 time via pb_service_time.
2951 Description: Watchdog triggered on MDS failover
2952 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2953 the OSTs know that the MDS "UUID" can be reused for the same export
2954 from a different NID, so we do not need to wait for the export to be
2958 Frequency : rare, only if using MMP with Linux RAID
2960 Description: MMP doesn't work with Linux RAID
2961 Details : While using HA for Lustre servers with Linux RAID, it is possible
2962 that MMP will not detect multiple mounts. To make this work we
2963 need to unplug the device queue in RAID when the MMP block is being
2964 written. Also while reading the MMP block, we should read it from
2965 disk and not the cached one.
2967 Severity : enhancement
2969 Description: open file using fid
2970 Details : A file can be opened using just its fid, like
2971 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2974 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2977 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2978 Details : add proper locking for creation of jbd_4k slab cache
2982 Description: MMP check in ext3_remount() fails without displaying any error
2983 Details : When multiple mount protection fails during remount, proper error
2986 Severity : enhancement
2988 Description: Allow stripe size to be up to 4G-64k
2989 Details : Fix math logic to allow large stripe sizes.
2993 Description: add check for >8TB ldiskfs filesystems
2994 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2995 Don't allow >8TB ldiskfs filesystems to be mounted without
2996 force_over_8tb mount option
3001 Descriptoin: OSS complains "too many blocks" and crashes
3002 Details : In filter_commitrw_write():
3003 1. lquota_chkquota();
3004 2. obdo_from_inode();
3005 3. lquota_pending_commit();
3006 "2." has changed uid and gid in this case, then caused
3007 lquota_chkquota() and lquota_pending_commit() to use
3008 different uid and gid. That is the root of the bug.
3012 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3013 Details : Detect and print out the 'stripe count = -1' case properly.
3017 Description: Out or order replies might be lost on replay
3018 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3019 request with tid smaller than one currently being added, add it
3020 to the start, not end of the list.
3022 --------------------------------------------------------------------------------
3024 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3026 * Support for kernels:
3027 2.6.5-7.283 (SLES 9),
3028 2.6.9-55.EL (RHEL 4),
3029 2.6.16.46-0.14 (SLES 10),
3030 2.6.18.8 vanilla (kernel.org)
3031 * Client support for unpatched kernels:
3032 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3033 2.6.16 - 2.6.22 vanilla (kernel.org)
3034 * Due to recently discovered recovery problems, we do not recommend
3035 using patchless RHEL 4 clients with this or any earlier release.
3036 * Recommended e2fsprogs version: 1.39.cfs8
3037 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3038 * Starting with this release, the ldiskfs backing filesystem required
3039 by Lustre is now in its own package, lustre-ldiskfs. This package
3040 should be installed. It is versioned separately from Lustre and
3041 may be released separately in future.
3045 Description: block reactivating mgc import until all deactivates complete
3046 Details : Fix race when failing back MDT/MGS to itself (testing)
3049 Frequency : at statup only
3051 Description: mds_lov_synchronize race leads to various problems
3052 Details : simultaneous MDT->OST connections at startup can cause the
3053 sync to abort, leaving the OSC in a bad state.
3055 Severity : enhancement
3057 Description: add optional extra BUILD_VERSION info
3058 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3059 to override the lustre version.
3062 Frequency : 2.6.18 servers only
3064 Description: ll_kern_mount() doesn't release the module reference
3065 Details : The ldiskfs module reference count never drops down to 0
3066 because ll_kern_mount() doesn't release the module reference.
3071 Description: server LBUG when using old ost_num_threads parameter
3072 Details : Accept the old ost_num_threads parameter but warn that it
3073 is deprecated, and fix an off-by-one error that caused an LBUG.
3078 Description: Transient SCSI error results in persistent IO issue
3079 Details : iobuf->dr_error is not reinitialized to 0 between two
3083 Frequency : sometimes when underlying device returns I/O errors
3085 Description: OSTs not going read-only during write failures
3086 Details : OSTs are not remounted read-only when the journal commit threads
3087 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3088 instead of the ext3 wrappers.
3092 Description: poor connect scaling with increasing client count
3093 Details : Don't run filter_grant_sanity_check for more than 100 exports
3094 to improve scaling for large numbers of clients.
3097 Frequency : SLES10 only
3099 Description: sanity-quota.sh quotacheck failed: rc = -22
3100 Details : Quotas cannot be enabled on SLES10.
3103 Frequency : liblustre clients only
3105 Description: getdirentries does not give error when run on compute nodes
3106 Details : getdirentries does not fail when the size specified as an argument
3107 is too small to contain at least one entry
3109 Severity : enhancement
3111 Description: Add LNET router traceability for debug purposes
3112 Details : If a checksum failure occurs with a router as part of the
3113 IO path, the NID of the last router that forwarded the bulk data
3114 is printed so it can be identified.
3119 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3120 Details : Due to a race condition, liblustre clients were occasionally
3121 evicted incorrectly.
3123 Severity : enhancement
3125 Description: lfs setstripe use optional parameters instead of postional
3128 Severity : enhancement
3130 Description: Nanosecond timestamp support for ldiskfs
3131 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3132 resolution timestamps. There is not yet support for this at
3133 the Lustre filesystem level.
3136 Frequency : during server recovery
3138 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3139 Details : request with rq_no_resend flag not awake l_wait_event if they get
3143 Frequency : nfs export on patchless client
3145 Description: connectathon hang when test nfs export over patchless client
3146 Details : Disconnected dentry cannot be found with lookup, so we do not need
3147 to unhash it or make it invalid
3150 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3151 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3152 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3157 Description: open req refcounting wrong on reconnect
3158 Details : If reconnect happened between getting open reply from server and
3159 call to mdc_set_replay_data in ll_file_open, we will schedule
3160 replay for unreferenced request that we are about to free.
3161 Subsequent close will crash in variety of ways.
3162 Check that request is still eligible for replay in
3163 mdc_set_replay_data().
3168 Description: disable writes to filesystem when reading health_check file
3169 Details : the default for reading the health_check proc file has changed
3170 to NOT do a journal transaction and write to disk, because this
3171 can cause reads of the /proc file to hang and block HA state
3172 checking on a healthy but otherwise heavily loaded system. It
3173 is possible to return to the previous behaviour during configure
3174 with --enable-health-write.
3176 Severity : enhancement
3178 Description: 64-bit inode version
3179 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3180 to the inode. This will be required for version-based recovery.
3185 Description: MDS fails to start if a duplicate client export is detected
3186 Details : in some rare cases it was possible for a client to connect to
3187 an MDS multiple times. Upon recovery the MDS would detect this
3188 and fail during startup. Handle this more gracefully.
3190 Severity : enhancement
3192 Description: Add -o localflock option to simulate old noflock
3194 Details : This will achieve local-only flock/fcntl locks
3200 Description: log_commit_thread vs filter_destroy race leads to crash
3201 Details : Take import reference before releasing llog record semaphore
3206 Description: Wrong request locking in request set processing
3207 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3208 imp_delayed_list, in this place should be used imp_lock.
3211 Frequency : when reconnection
3213 Description: Grant Leak when osc reconnect to OST
3214 Details : When osc reconnect ost, OST(filter) should check whether it
3215 should grant more space to client by comparing fed_grant and
3216 cl_avail_grant, and return the granted space to client instead
3217 of "new granted" space, because client will call osc_init_grant
3218 to update the client grant space info.
3221 Frequency : when client reconnect to OST
3223 Description: Grant Leak when osc do resend and replay bulk write
3224 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3225 bulk write request, because the grant info while be sync between
3226 OSC and OST when reconnect, and we should ignore the grant info
3227 these of resend/replay write req.
3232 Description: Grant space more than available left space sometimes.
3233 Details : When then OST is about to be full, if two bulk writing from
3234 different clients came to OST. Accord the avaliable space of the
3235 OST, the first req should be permitted, and the second one
3236 should be denied by ENOSPC. But if the seconde arrived before
3237 the first one is committed. The OST might wrongly permit second
3238 writing, which will cause grant space > available space.
3241 Frequency : when client is evicted
3243 Description: Grant might be wrongly erased when osc is evicted by OST
3244 Details : when the import is evicted by server, it will fork another
3245 thread ptlrpc_invalidate_import_thread to invalidate the
3246 import, where the grant will be set to 0. While the original
3247 thread will update the grant it got when connecting. So if
3248 the former happened latter, the grant will be wrongly errased
3249 because of this race.
3254 Description: Checking Stale with correct fid
3255 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3256 to check whether it is stale, because sometimes, we want the
3257 enqueue happened anyway, and op_data.fid2 will not be initialized.
3259 Severity : enhancement
3261 Description: update patchless client
3262 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3265 Frequency : only with 2.4 kernel
3267 Description: random memory corruption
3268 Details : size of struct ll_inode_info is to big for union inode.u and this
3269 can be cause of random memory corruption.
3274 Description: Memory leak in recovery
3275 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3276 It should also check obd_fail before fsfilt_start, otherwise if
3277 fsfilt_start return -EROFS,(failover mds during mds recovery).
3278 then the req will return with repmsg->transno = 0 and rc = EROFS.
3279 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3280 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3281 be freed no matter whether fsfilt_commit success or not.
3284 Frequency : only with huge count clients
3286 Description: Prevents from taking the superblock lock in llap_from_page for
3288 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3289 for llap_from_page call in ll_removepage prevents from taking the
3290 superblock lock for a soon died page.
3295 Description: Not check open intent error before release open handle
3296 Details : in some rare cases, the open intent error is not checked before
3297 release open handle, which may cause
3298 ASSERTION(open_req->rq_transno != 0), because it tries to release
3299 the failed open handle.
3304 Description: Set cat log bitmap only after create log success.
3305 Details : in some rare cases, the cat log bitmap is set too early. and it
3306 should be set only after create log success.
3310 Description: Accessing a block bevice can re-enable I/O when Lustre is
3311 tearing down a device.
3312 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3317 Description: service threads may hog cpus when there are a lot of requests
3318 Details : Insert cond_resched to give other threads a chance to use some CPU
3323 Description: the cat log was not initialized in recovery
3324 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3325 unlink log on mds will not be initialized until mds post
3326 recovery. And also in mds post recovery, the unlink log will
3327 initialization will be done asynchronausly, so there will be race
3328 between add unlink log and unlink log initialization.
3332 Description: brw_stats were being printed incorrectly
3333 Details : brw_stats were being printed as log2 but all of them were not
3334 recorded as log2. Also remove some code duplication arising from
3335 filter_tally_{read,write}.
3339 Frequency : rare, only in recovery.
3340 Description: ASSERTION(req->rq_type != LI_POISON) failed
3341 Details : imp_lock should be held while iterating over imp_sending_list for
3342 prevent destroy request after get timeout in ptlrpc_queue_wait.
3346 Description: replay-single.sh test 52 fails
3347 Details : A lock's skiplist need to be cleanup when it being unlinked
3348 from its resource list.
3352 Description: Short directio read returns full requested size rather than
3354 Details : Direct I/O operations should return actual amount of bytes
3355 transferred rather than requested size.
3357 Severity : enhancement
3359 Description: metadata RPC reduction (e.g. for rm performance)
3360 Details : decrease the amount of synchronous RPC between clients and servers
3361 by canceling conflicing lock before the operation on the client
3362 and packing thier handles into the main operation RPC to server.
3364 Severity : enhancement
3366 Description: Async OSC create to avoid the blocking unnecessarily.
3367 Details : If an OST has no remaining object, system will block on the
3368 creation when it needs to create a new object on this OST. Now,
3369 ways use pre-created objects when available, instead of blocking on
3370 an empty osc while others are not empty. If we must block, we
3371 block for the shortest possible period of time.
3375 Description: improve handling recoverable errors
3376 Details : if request processig with error which can be recoverable on server
3377 request should be resend, otherwise page released from cache and
3380 Severity : enhancement
3382 Description: refine locking for avoid write wrong info into lov_objid file
3383 Details : fix possible races with add new target and write/update data in
3386 --------------------------------------------------------------------------------
3388 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3393 Frequency : on some architectures
3395 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3396 Details : Uninitialized flags sometimes cause configuration commands to
3400 Frequency : patchless clients only
3402 Description: missing __iget() symbol export
3403 Details : The __iget() symbol export is missing. To avoid the need for
3404 this on patchless clients the deathrow inode reaper is turned
3405 off, and we depend on the VM to clean up old inodes. This
3406 dependency was during via the fix for bug 12181.
3408 --------------------------------------------------------------------------------
3410 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3412 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3413 INTEROPERATE with older versions automatically. In many cases a
3414 special upgrade step is needed. Please read the
3415 user documentation before upgrading any part of a live system.
3416 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3417 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3418 * WARNING: Lustre configuration and startup changes are required with
3419 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3423 Severity : enhancement
3425 Description: MountConf
3426 Details : Lustre configuration is now managed via mkfs and mount
3427 commands instead of lmc and lconf. New obd types (MGS, MGC)
3428 are added for dynamic configuration management. See
3429 https://mail.clusterfs.com/wikis/lustre/MountConf for
3432 Severity : enhancement
3434 Description: dynamic OST addition
3435 Details : OSTs can now be added to a live filesystem
3437 Severity : enhancement
3439 Description: startup order invariance
3440 Details : MDTs and OSTs can be started in any order. Clients only
3441 require the MDT to complete startup.
3443 Severity : enhancement
3445 Description: parallel, asynchronous orphan cleanup
3446 Details : orphan cleanup is now performed in separate threads for each
3447 OST, allowing parallel non-blocking operation.
3449 Severity : enhancement
3451 Description: optimized stripe assignment
3452 Details : stripe assignments are now made based on ost space available,
3453 ost previous usage, and OSS previous usage, in order to try
3454 to optimize storage space and networking resources.
3456 Severity : enhancement
3458 Description: Permanently set tunables
3459 Details : All writable /proc/fs/lustre tunables can now be permanently
3460 set on a per-server basis, at mkfs time or on a live system.
3462 Severity : enhancement
3464 Description: Lustre message v2
3465 Details : Add lustre message format v2.
3467 Severity : enhancement
3469 Description: client OST exclusion list
3470 Details : Clients can be started with a list of OSTs that should be
3471 declared "inactive" for known non-responsive OSTs.
3475 Description: SPEC SFS validation failure on NFS v2 over lustre.
3476 Details : Changes the blocksize for regular files to be 2x RPC size,
3477 and not depend on stripe size.
3479 Severity : enhancement
3481 Description: Multiple MD RPCs in flight.
3482 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3483 To avoid overly-overloading MDS, introduce a limit on number of
3484 MDS RPCs in flight for a single client and add /proc controls
3485 to adjust this limit.
3487 Severity : enhancement
3489 Description: client read/write statistics
3490 Details : Add client read/write call usage stats for performance
3491 analysis of user processes.
3492 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3493 file access. extents_stats shows chunk size distribution.
3494 extents_stats_per_process show chunk size distribution per
3497 Severity : enhancement
3499 Description: mds statistics
3500 Details : Add detailed mds operations statistics in
3501 /proc/fs/lustre/mds/*/stats.
3505 Description: Failure of copying files with lustre special EAs.
3506 Details : Client side always return success for setxattr call for lustre
3507 special xattr (currently only "trusted.lov").
3512 Description: Refcount LNET uuids
3513 Details : The global LNET uuid list grew linearly with every startup;
3514 refcount repeated list entries instead of always adding to
3517 Severity : enhancement
3519 Description: Dynamic service threads
3520 Details : Within a small range, start extra service threads
3521 automatically when the request queue builds up.
3524 Frequency : mixed-endian client/server environments
3526 Description: mixed-endian crashes
3527 Details : The new msg_v2 system had some failures in mixed-endian
3530 Severity : enhancement
3532 Description: Easy OST removal
3533 Details : OSTs can be permanently deactivated with e.g. 'lctl
3534 conf_param lustre-OST0001.osc.active=0'
3536 Severity : enhancement
3538 Description: MGS proc entries
3539 Details : Added basic proc entries for the MGS showing what filesystems
3542 Severity : enhancement
3544 Description: provide MGS failover
3545 Details : Added config lock reacquisition after MGS server failover.
3547 Severity : enhancement
3549 Description: add Linux 2.4 support
3550 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3554 Description: a large application tries to do I/O to the same resource and dies
3555 in the middle of it.
3556 Details : Check the req->rq_arrival time after the call to
3557 ost_brw_lock_get(), but before we do anything about
3558 processing it & sending the BULK transfer request. This
3559 should help move old stale pending locks off the queue as
3560 quickly as obd_timeout.
3563 Frequency : when an incorrect nid is specified during startup
3565 Description: ptlrpc connect to non-existant node causes kernel crash
3566 Details : LNET can't be re-entered from an event callback, which
3567 happened when we expire a message after the export has been
3568 cleaned up. Instead, hand the zombie cleanup off to another
3571 Severity : enhancement
3573 Description: plain/inodebits lock performance improvement
3574 Details : Grouping plain/inodebits in granted list by their request modes
3575 and bits policy, thus improving the performance of search through
3579 Frequency : only if OST filesystem is corrupted
3581 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3582 Details : for a short time RPCs with bulk IO are in the replay list,
3583 but replay of bulk IOs is unimplemented. If the OST filesystem
3584 is corrupted due to disk cache incoherency and then replay is
3585 started it is possible to trip an assertion. Avoid putting
3586 committed RPCs into the replay list at all to avoid this issue.
3589 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3591 Description: System hang on startup
3592 Details : This bug allowed the liblustre (e.g. catamount) client to
3593 return to the app before handling all startup RPCs. This
3594 could leave the node unresponsive to lustre network traffic
3595 and manifested as a server ptllnd timeout.
3597 Severity : enhancement
3599 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3600 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3601 boolean to control whether to print peer debug info when a
3602 client's RPC times out.
3605 Frequency : only for kernels with patches from Lustre below 1.4.3
3607 Description: Remove old rdonly API
3608 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3611 Frequency : only for devices with external journals
3613 Description: Set external device read-only also
3614 Details : During a commanded failover stop, we set the disk device
3615 read-only while the server shuts down. We now also set any
3616 external journal device read-only at the same time.
3619 Frequency : when upgrading from 1.4 while trying to change parameters
3621 Description: The wrong (new) MDC name was used when setting parameters for
3622 upgraded MDT's. Also allows changing of OSC (and MDC)
3623 parameters if --writeconf is specified at tunefs upgrade time.
3626 Frequency : when setting specific ost indicies
3628 Description: QOS code breaks on skipped indicies
3629 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3630 created with --index need not be sequential.
3632 Severity : enhancement
3634 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3635 Details : The uninit_groups feature works in conjunction with the kernel
3636 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3637 up the pass1 processing of e2fsck. This is a read-only feature
3638 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3639 mount filesystems that have had this feature enabled.
3641 Severity : enhancement
3643 Description: Improve multi-block allocation algorithm to avoid fragmentation
3644 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3645 allocation locality and avoid filesystem fragmentation.
3647 ------------------------------------------------------------------------------
3649 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3651 * Support for kernels:
3652 2.4.21-47.0.1.EL (RHEL 3)
3653 2.6.5-7.283 (SLES 9)
3654 2.6.9-42.0.10.EL (RHEL 4)
3655 2.6.12.6 vanilla (kernel.org)
3656 2.6.16.27-0.9 (SLES 10)
3657 * Recommended e2fsprogs version: 1.39.cfs5
3659 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3663 Frequency : occasional, depends on client load and configuration
3664 Bugzilla : 12181, 12203
3665 Description: data loss for recently-modified files
3667 Details : In some cases it is possible that recently written or created
3668 files may not be written to disk in a timely manner (this should
3669 normally be within 30s unless client IO load is very high).
3670 The problem appears as zero-length files or files that are a
3671 multiple of 1MB in size after a client crash or client eviction
3672 that are missing data at the end of the file.
3674 This problem is more likely to be hit on clients where files are
3675 repeatedly created and unlinked in the same directory, clients
3676 have a large amount of RAM, have many CPUs, the filesystem has
3677 many OSTs, the clients are rebooted frequently, and/or the files
3678 are not accessed by other nodes after being written.
3680 The presence of the problem can be detected by looking at
3681 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3682 smaller than the second (nr_unused) then dirty files will not
3683 be flushed automatically to disk. "sync; sleep 10" should be
3684 run several times on the node before unmounting it to update
3685 Lustre (this is also safe to run on nodes without this problem).
3687 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3688 that can cause this same problem, so customers using that kernel
3689 also need to update the kernel in addition to Lustre. In order
3690 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3692 It is normal that files written just before a client crash (less
3693 than 30s) may not yet have been flushed to disk, even for local
3697 Frequency : frequent on thin XT3 nodes
3699 Description: UUID collision on thin XT3 Linux nodes
3700 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3701 randomly, since we relied on an insufficiently-seeded PRNG.
3706 Description: OSS hangs after "All ost request buffers busy"
3707 Details : A deadlock between quota and journal operations caused OSS
3708 hangs after printing "All ost request buffers busy."
3711 Frequency : always on liblustre builds
3713 Description: Cleanup compiler warnings on liblustre
3716 Frequency : always on liblustre builds on XT3
3718 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3722 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3723 > 32000 subdirectories
3724 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3725 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3726 subdirectory count crosses 32000. This will aid e2fsck to
3727 correctly handle more than 32000 subdirectories.
3730 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3732 Description: System hang on startup
3733 Details : This bug allowed the liblustre (e.g. catamount) client to
3734 return to the app before handling all startup RPCs. This
3735 could leave the node unresponsive to lustre network traffic
3736 and manifested as a server ptllnd timeout.
3738 Severity : enhancement
3740 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3741 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3742 boolean to control whether to print peer debug info when a
3743 client's RPC times out.
3748 Description: make O_SYNC working on 2.6 kernels
3749 Details : 2.6 kernels use different method for mark pages for write,
3750 so need add a code to lustre for O_SYNC work.
3755 Description: Failure to close file and release space on NFS
3756 Details : Put inode details into lock acquired in ll_intent_file_open.
3757 Use mdc_intent_lock in ll_intent_open to properly
3758 detect all kind of errors unhandled by mdc_enqueue.
3763 Description: proc file read during shutdown sometimes raced obd removal,
3765 Details : Add lock to prevent obd access after proc file removal.
3768 Frequency : Only for files larger than 4GB on 32-bit clients.
3770 Description: improperly doing page alignment of locks
3771 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3772 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3777 Description: RPCs being resent when they shouldn't be
3778 Details : Some RPCs that should not be resent are being resent. This
3779 can cause inconsistencies in the RPC state machine. Do not
3780 resend such requests.
3783 Frequency : rare, only with NFS export
3785 Description: Crash on NFS re-export node
3786 Details : under very unusual load conditions an assertion is hit in
3787 ll_intent_file_open()
3790 Frequency : only if OST filesystem is corrupted
3792 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3793 Details : for a short time RPCs with bulk IO are in the replay list,
3794 but replay of bulk IOs is unimplemented. If the OST filesystem
3795 is corrupted due to disk cache incoherency and then replay is
3796 started it is possible to trip an assertion. Avoid putting
3797 committed RPCs into the replay list at all to avoid this issue.
3802 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3803 Details : Large single O_DIRECT read and write calls can fail to allocate
3804 a sufficiently large buffer to process the request. In case of
3805 allocation failure the allocation is retried with a smaller
3806 buffer and broken into smaller requests.
3808 Severity : enhancement
3810 Description: Add -o localflock option to simulate old noflock behaviour.
3811 Details : This will achieve local-only flock/fcntl locks coherentness.
3816 Description: versioning check is incomplete
3817 Details : Checking the version difference of client vs. server, report
3818 error if the gap is too big.
3823 Description: add support PG_writeback bit
3824 Details : add support for PG_writeback bit for Lustre, for more carefull
3825 work with page cache in 2.6 kernel. This also fix some deadlocks
3826 and remove hack for work O_SYNC with 2.6 kernel.
3828 Severity : enhancement
3830 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3831 Details : The uninit_groups feature works in conjunction with the kernel
3832 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3833 up the pass1 processing of e2fsck. This is a read-only feature
3834 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3835 mount filesystems that have had this feature enabled.
3837 Severity : enhancement
3839 Description: Improve multi-block allocation algorithm to avoid fragmentation
3840 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3841 allocation locality and avoid filesystem fragmentation.
3843 ------------------------------------------------------------------------------
3845 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3847 * Support for kernels:
3848 2.6.9-42.0.3.EL (RHEL 4)
3849 2.6.5-7.276 (SLES 9)
3850 2.4.21-47.0.1.EL (RHEL 3)
3851 2.6.12.6 vanilla (kernel.org)
3852 2.6.16.21-0.8 (SLES10)
3853 * Recommended e2fsprogs version: 1.39.cfs2-0
3855 * The backwards-compatible /proc/sys/portals symlink has been removed
3856 in this release. Before upgrading, please ensure that you change
3857 any configuration scripts or /etc/sysctl.conf files that access
3858 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3859 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3860 in advance of the upgrade on any system running Lustre 1.4.6 or
3861 newer, since /proc/sys/lnet was added in that version.
3862 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3866 Frequency : only when quota is used
3868 Description: avoid scanning export list for quota master
3869 Details : Change the algorithms to avoid scanning export list in order
3870 to improve the efficiency.
3873 Frequency : MDS failover only, very rarely
3875 Description: "went back in time" messages on mds failover
3876 Details : The greatest transno may be lost when the current operation
3877 finishes with an error (transno==0) and the client's last_rcvd
3878 record is over-written. Save the greatest transno in the
3879 mds_last_transno for this case.
3882 Frequency : always for specific kernels and striping counts
3884 Description: client may get "Matching packet too big" without ACL support
3885 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3886 when trying to access files in certain configurations. The
3887 clients should in fact be denied when mounting because they do
3888 not understand ACLs.
3891 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3893 Description: many clients connecting with IO in progress causes connect timeouts
3894 Details : Avoid synchronous journal commits to avoid delays caused by many
3895 clients connecting/disconnecting when bulk IO is in progress.
3896 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3897 OST_IO_PORTAL to avoid delays behind potentially many pending
3901 Frequency : occasionally with multiple writers to a single file
3903 Description: shared writes to file may result in wrong size reported by stat()
3904 Details : Allow growing of kms when extent lock is cancelled
3907 Frequency : always with random mmap IO to multi-striped file
3909 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3910 Details : Only if the hole is at the end of OST object so that kms is too
3911 small. Fix is to increase kms accordingly in ll_nopage.
3914 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3916 Description: writes to a missing object would leak memory on the OST
3917 Details : If there is an inconsistency between the MDS and OST filesystems,
3918 such that the MDS references an object that doesn't exist, writes
3919 to that object will leak memory due to incorrect cleanup in the
3920 error handling path, eventually running out of memory on the OST.
3925 Description: Creating too long symlink causes lustre errors
3926 Details : Check symlink and name lengths before sending requests to MDS.
3929 Frequency : only if flock is enabled (not on by default)
3931 Description: posix locks not released on fd closure on 2.6.9+
3932 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3933 kernels, this caused such locks not to be released on fd close and
3934 then assertions on fs unmount about still used locks.
3937 Frequency : MDS failover only, very rarely
3939 Description: clients may get ASSERTION(granted_lock != NULL)
3940 Details : When request was taking a long time, and a client was resending
3941 a getattr by name lock request. The were multiple lock requests
3942 with the same client lock handle and
3943 mds_getattr_name->fixup_handle_for_resent_request found one of the
3944 lock handles but later failed with ASSERTION(granted_lock != NULL).
3949 Description: handle->h_buffer_credits > 0, assertion failure
3950 Details : h_buffer_credits is zero after truncate, causing assertion
3951 failure. This patch extends the transaction or creates a new
3955 Frequency : NFS re-export or patchless client
3956 Bugzilla : 11179, 10796
3957 Description: Crash on NFS re-export node (__d_move)
3958 Details : We do not want to hash the dentry if we don't have a lock.
3959 But if this dentry is later used in d_move, we'd hit uninitialised
3960 list head d_hash, so we just do this to init d_hash field but
3961 leave dentry unhashed.
3964 Frequency : NFS re-export or patchless client
3966 Description: NFS exports has problem with symbolic link
3967 Details : lustre client didn't properly install dentry when re-exported
3968 to NFS or running patchless client.
3971 Frequency : NFS re-export or patchless client
3973 Description: Various nfs/patchless fixes.
3974 Details : fixes reuse disconected alias for lookup process - this fixes
3975 warning "find_exported_dentry: npd != pd",
3976 fix permission error with open files at nfs.
3977 fix apply umask when do revalidate.
3980 Frequency : occasional
3982 Description: Crash on NFS re-export node
3983 Details : calling clear_page() on the wrong pointer triggered oops in
3984 generic_mapping_read().
3987 Frequency : rarely, using O_DIRECT IO
3989 Description: unaligned directio crashes client with LASSERT
3990 Details : check for unaligned buffers before trying any requests.
3993 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3995 Description: stale data returned from RAID cache
3996 Details : If only a small amount of IO is done to the RAID device before
3997 reading it again it is possible to get stale data from the RAID
3998 cache instead of reading it from disk.
4001 Frequency : always for sles10 kernel
4003 Description: sles10 support
4004 Details : ll_follow_link: compile fixes and using of nd_set_link
4005 under newer kernels.
4008 Frequency : depends on arch, kernel and compiler version, always on sles10
4011 Description: recursive or deep enough symlinks cause stack overflow
4012 Details : getting rid of large stack-allocated variable in
4016 Frequency : depends on hardware
4018 Description: lustre write performance loss in the SLES10 kernel
4019 Details : the performance loss is caused by using of write barriers in the
4020 ext3 code. The SLES10 kernel turns barrier support on by
4021 default. The fix is to undo that change for ldiskfs.
4023 ------------------------------------------------------------------------------
4025 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4027 * Support for kernels:
4028 2.6.9-42.0.3EL (RHEL 4)
4029 2.6.5-7.276 (SLES 9)
4030 2.4.21-47.0.1.EL (RHEL 3)
4031 2.6.12.6 vanilla (kernel.org)
4035 Frequency : quota enabled and large files being deleted
4037 Description: releasing more than 4GB of quota at once hangs OST
4038 Details : If a user deletes more than 4GB of files on a single OST it
4039 will cause the OST to spin in an infinite loop. Release
4040 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4045 Description: statfs data retrieved from /proc may be stale or zero
4046 Details : When reading per-device statfs data from /proc, in the
4047 {kbytes,files}_{total,free,avail} files, it may appear
4048 as zero or be out of date.
4051 Frequency : systems with MD RAID1 external journal devices
4053 Description: lconf's call to blkid is confused by RAID1 journal devices
4054 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4055 of returning all block devices that match the journal UUID.
4058 Frequency : always, for aggregate stripe size over 4GB
4060 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4061 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4062 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4065 Frequency : always if "lfs find" used on a local file/directory
4067 Description: "lfs find" segfaults if used on a local file/directory
4068 Details : The case where a directory component was not specified wasn't
4069 handled correctly. Handle this properly.
4072 Frequency : always on ppc64
4074 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4075 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4076 perform endianness conversion.
4079 Frequency : rarely (truncate to non-zero file size after write under load)
4080 Bugzilla : 10730, 10687
4081 Description: Files padded with zeros to next 4K multiple
4082 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4083 it is possible that files that are truncated to a non-zero size
4084 immediately after being written are filled with zero bytes beyond
4085 the truncated size. No file data is lost.
4087 Severity : enhancement
4089 Description: Allow recovery/failover for liblustre clients.
4090 Details : liblustre clients were unaware of failover configurations until
4093 Severity : enhancement
4095 Description: user file locks should fail when not mounting with flock option
4096 Details : Set up an error-returning stub in ll_file_operations.lock field
4097 to prevent incorrect behaviour when client is mounted without
4098 flock option. Also, set up properly f_op->flock field for
4102 Frequency : always on ia64
4104 Description: "lfs df" loops on printing out MDS statfs information
4105 Details : The obd_ioctl_data was not initialized and in some systems
4106 this caused a failure during the ioctl that did not return
4107 an error. Initialize the struct and return an error on failure.
4110 Frequency : SLES 9 only
4112 Description: Error of copying files with lustre special EAs as root
4113 Details : Client side always return success for setxattr call for lustre
4114 special xattr (currently only "trusted.lov").
4117 Frequency : rarely on clusters with both ia64+i386 clients
4119 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4120 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4121 clients are doing concurrent writes to the same file it is
4122 possible that the ia64 clients may LASSERT because the OST
4123 extent locks are not PAGE_SIZE aligned. Ensure that grown
4124 locks are always aligned on the request boundary.
4127 Frequency : specific use, occasional
4129 Description: Overwriting in use executable truncates on-disk binary image
4130 Details : If one node attempts to overwrite an executable in use by
4131 another node, we now correctly return ETXTBSY instead of
4132 truncating the file.
4134 Severity : enhancement
4136 Description: Async OSC create to avoid the blocking unnecessarily.
4137 Details : If an OST has no remaining object, system will block on the creating
4138 when need to create a new object on this OST. Now, ways use
4139 pre-created objects when available, instead of blocking on an
4140 empty osc while others are not empty. If we must block, we block
4141 for the shortest possible period of time.
4146 Description: chmod on Lustre root is propagated to other clients
4147 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4148 invalid by the follow_mount time.
4151 Frequency : liblustre clients only
4153 Description: Race in 'instant cancel' lock handling could lead to such locks
4154 never to be granted in case of SMP MDS
4155 Details : Do not destroy not yet granted but cbpending locks in
4159 Frequency : replay/resend of open
4161 Description: non null lock assetion failure in mds_intent_policy
4162 Details : Trying to replay/resend lockless open requests resulted in
4163 mds_open() returning 0 with no lock. Now it sets a flag if
4164 a lock is going to be returned.
4166 Severity : enhancement
4168 Description: Checksum enhancements
4169 Details : New checksum enhancements allow for resending RPCs that failed
4172 Severity : enhancement
4174 Description: Tunables on number of dirty pages in cacche
4175 Details : Allow to set limit on number of dirty pages cached.
4180 Description: client crash on unmount - lock still has references
4181 Details : In some error handling cases it was possible to leak a lock
4182 reference on a client while accessing a file. This was not
4183 harmful to the client during operation, but would cause the
4184 client to crash when the filesystem is unmounted.
4187 Frequency : specific case, rare
4189 Description: ETXTBSY on mds though file not in use by client
4190 Details : ETXTBSY is no longer incorrectly returned when attempting to
4191 chmod or chown a directory that the user previously tried to
4192 execute or a currently-executing binary.
4195 Frequency : extremely rare except on liblustre-based clients
4197 Description: Lustre space not freed when files are deleted
4198 Details : Clean up open-unlinked files after client eviction. Previously
4199 the unlink was skipped and the files remained as orphans.
4204 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4205 Details : In some cases it was possible to send a blocking callback to
4206 a client doing a glimpse, even though that client didn't get
4207 a lock granted. When the glimpse lock is cancelled on the OST
4208 the freed lock is left on the waiting list and corrupted the list.
4211 Frequency : all core dumps
4213 Description: Broke core dumps to lustre
4214 Details : Negative dentry may be unhashed if parent does not have UPDATE
4215 lock, but some callers, e.g. do_coredump, expect dentry to be
4216 hashed after successful create, hash it in ll_create_it.
4218 ------------------------------------------------------------------------------
4220 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4222 * Support for kernels:
4223 2.6.9-42.0.2.EL (RHEL 4)
4224 2.6.5-7.276 (SLES 9)
4225 2.4.21-40.EL (RHEL 3)
4226 2.6.12.6 vanilla (kernel.org)
4230 Frequency : always on RHEL 3
4232 Description: Number of open files grows over time
4233 Details : The number of open files grows over time, whether or not
4234 Lustre is started. This was due to a filp leak introduced
4235 by one of our kernel patches.
4237 ------------------------------------------------------------------------------
4239 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4241 * Support for kernels:
4242 2.6.9-42.EL (RHEL 4)
4243 2.6.5-7.276 (SLES 9)
4244 2.4.21-40.EL (RHEL 3)
4245 2.6.12.6 vanilla (kernel.org)
4250 Bugzilla : 5719, 9635, 9792, 9684
4251 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4252 Details : If a server is under heavy load and cannot reply to new
4253 connection requests before the client resends the (re)connect,
4254 the connection handling code can behave badly if two service
4255 threads are concurrently handing separate (re)connections from
4256 the same client. Add better locking to the connection handling
4257 code, and ensure that only a single connection will be processed
4258 for a given client UUID, even if the lock is dropped.
4260 Severity : enhancement
4262 Description: add TCP zero-copy support to kernel
4263 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4264 sends if the hardware supports scatter-gather and checksumming.
4265 This allows socklnd to do client-write and server-read more
4266 efficiently and reduce CPU utilization from skbuf copying.
4269 Frequency : only if NFS exporting from client
4271 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4272 Details : If setting attributes on a file created under NFS that had
4273 never been opened it would be possible to oops the client
4274 if the file had no objects.
4277 Frequency : always for liblustre
4279 Description: liblustre client does MDS+OSTs setattr RPC for each write
4280 Details : When doing a write from a liblustre client, the client
4281 incorrectly issued an RPC to the MDS and each OST the file was
4282 striped over in order to update the timestamps. When writing
4283 with small chunks and many clients this could overwhelm the MDS
4284 with RPCs. In all cases it would slow down the write because
4285 these RPCs are unnecessary.
4287 Severity : enhancement
4289 Description: allow number of MDS service threads to be changed at module load
4290 Details : It is now possible to change the number of MDS service threads
4291 running. Adding "options mds mds_num_threads={N}" to the MDS's
4292 /etc/modprobe.conf will set the number of threads for the next
4293 time Lustre is restarted (assuming the "mds" module is also
4294 reloaded at that time). The default number of threads will
4295 stay the same, 32 for most systems.
4300 Description: OST crash if filesystem is unformatted or corrupt
4301 Details : If an OST is started on a device that has never been formatted
4302 or if the filesystem is corrupt and cannot even mount then the
4303 error handling cleanup routines would dereference a NULL pointer.
4308 Description: NULL pointer deref in llap_from_page.
4309 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4310 incorrect) mapping if the page was truncated/reclaimed while it was
4311 searched for. Check for this condition and skip such pages when
4312 doing readahead. Introduce extra check to llap_from_page() to
4313 verify page->mapping->host is non-NULL (so page is not anonymous).
4316 Frequency : Sometimes when using sys_sendfile
4318 Description: "page not covered by a lock" warnings from ll_readpage
4319 Details : sendfile called ll_readpage without right page locks present.
4320 Now we introduced ll_file_sendfile that does necessary locking
4321 around call to generic_file_sendfile() much like we do in
4325 Frequency : with certain MDS communication failures at client mount time
4327 Description: NULL pointer deref after failed client mount
4328 Details : a client connection request may delayed by the network layer
4329 and not be sent until after the PTLRPC layer has timed out the
4330 request. If the client fails the mount immediately it will try
4331 to clean up before the network times out the request. Add a
4332 reference from the request import to the obd device and delay
4333 the cleanup until the network drops the request.
4336 Frequency : occasionally during client (re)connect
4338 Description: assertion failure during client (re)connect
4339 Details : processing a client connection request may be delayed by the
4340 client or server longer than the client connect timeout. This
4341 causes the client to resend the connection request. If the
4342 original connection request is replied in this interval, the
4343 client may trip an assertion failure in ptlrpc_connect_interpret()
4344 which thought it would be the only running connect process.
4347 Frequency : only with obd_echo servers and clients that are rebooted
4349 Description: kernel BUG accessing uninitialized data structure
4350 Details : When running an obd_echo server it did not start the ping_evictor
4351 thread, and when a client was evicted an uninitialized data
4352 structure was accessed. Start the ping_evictor in the RPC
4353 service startup instead of the OBD startup.
4355 Severity : enhancement
4356 Bugzilla : 10193 (patchless)
4357 Description: Remove dependency on various unexported kernel interfaces.
4358 Details : No longer need reparent_to_init, exit_mm, exit_files,
4359 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4362 Frequency : rare (only users of deprecated and unsupported LDAP config)
4364 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4365 Details : LDAP apparently contains 'lustreName' attributes instead of
4366 'name'. A simple remapping of the name is sufficient.
4369 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4371 Description: waiting_locks_callback trips kernel BUG if client is evicted
4372 Details : Running with the dump_on_timeout debug flag turned on makes
4373 it possible that the waiting_locks_callback() can try to dump
4374 the Lustre kernel debug logs from an interrupt handler. Defer
4375 this log dumping to the expired_lock_main() thread.
4377 Severity : enhancement
4379 Description: Support NFS exporting on 2.6 kernels.
4380 Details : Implement non-rawops metadata methods for NFS server to use without
4381 changing NFS server code.
4384 Frequency : very rare (synthetic metadata workload only)
4386 Description: two racing renames might cause an MDS thread to deadlock
4387 Details : Running the "racer" program may cause one MDS thread to rename
4388 a file from being the source of a rename to being the target of
4389 a rename at exactly the same time that another thread is doing
4390 so, and the second thread has already enqueued these locks after
4391 doing a lookup of the target and is trying to relock them in
4392 order. Ensure that we don't try to re-lock the same resource.
4395 Frequency : only very large systems with liblustre clients
4397 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4398 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4399 all OSTs in parallel. This allows the eviction of stale liblustre
4400 clients to proceed much faster than if they were done in series,
4401 and also offers similar improvements for other set_info RPCs.
4406 Description: excessive CPU usage during initial read phase on client
4407 Details : During the initial read phase on a client, it would agressively
4408 retry readahead on the file, consuming too much CPU and impacting
4409 performance (since 1.4.5.8). Improve the readahead algorithm
4410 to avoid this, and also improve some other common cases (read
4411 of small files in particular, where "small" is files smaller than
4412 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4417 Description: MDS crash when receiving packet with unknown intent.
4418 Details : Do not LBUG in unknown intent case, just return -EFAULT
4420 Severity : enhancement
4421 Bugzilla : 9293, 9385
4422 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4423 Details : Do not serialize getattr (non-intent version) and statfs.
4426 Frequency : occasional, when OST network is overloaded/intermittent
4428 Description: client evicted by OST after bulk IO timeout
4429 Details : If a client sends a bulk IO request (read or write) the OST
4430 may evict the client if it is unresposive to its data GET/PUT
4431 request. This is incorrect if the network is overloaded (takes
4432 too long to transfer the RPC data) or dropped the OST GET/PUT
4433 request. There is no need to evict the client at all, since
4434 the pinger and/or lock callbacks will handle this, and the
4435 client can restart the bulk request.
4438 Frequency : Always when mmapping file with no objects
4440 Description: client crashes when mmapping file with no objects
4441 Details : Check that we actually have objects in a file before doing any
4442 operations on objects in ll_vm_open, ll_vm_close and
4448 Description: Request leak when working with deleted CWD
4449 Details : Introduce advanced request refcount tracking for requests
4450 referenced from lustre intent.
4452 Severity : Enhancement
4454 Description: Cache open file handles on client.
4455 Details : MDS now will return special lock along with openhandle, if
4456 requested and client is allowed to hold openhandle, even if unused,
4457 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4458 closing files for every read/write openration.
4460 Severity : Enhancement
4462 Description: Cache open negative dentries on client when possible.
4463 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4464 negative dentries on lock revocation.
4469 Description: Remounting a client read-only wasn't possible with a zconf mount
4470 Details : It wasn't possible to remount a client read-only with llmount.
4472 Severity : enhancement
4473 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4474 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4475 MPICH is included to add Lustre-specific ADIO interfaces.
4476 This is based closely on the UFS ADIO layer and only differs
4477 in file creation, in order to allow the OST striping to be set.
4478 This is user-contributed code and not supported by CFS.
4483 Description: extended inode attributes (immutable, append-only) work improperly
4484 when 2.4 and 2.6 kernels are used on client/server or vice versa
4485 Details : Introduce kernel-independent values for these flags.
4487 Severity : enhancement
4490 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4491 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4492 granularity. Now, Fractional MB granularity is be supported,
4493 this is very useful for low memory system.
4495 Severity : enhancement
4497 Description: Getattr by fid
4498 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4499 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4500 have locks and avoids getattr rpc requests for it on every stat.
4503 Frequency : Always, for filesystems larger than 2TB
4505 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4506 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4507 filesystems larger than 2TB to be mounted with mballoc enabled.
4510 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4512 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4513 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4514 filesystems larger than 2TB is mounted then the kernel will
4515 silently corrupt the start of the filesystem. CONFIG_LBD is
4516 enabled for all CFS-supported kernels, but the possibility of
4517 this happening with a modified kernel config exists.
4519 Severity : enhancement
4521 Description: add client O_DIRECT support for 2.6 kernels
4522 Details : It is now possible to do O_DIRECT reads and writes to files
4523 in the Lustre client mountpoint on 2.6 kernel clients.
4525 Severity : enhancement
4527 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4528 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4529 parallel, not waiting for response from every OST before sending
4530 a rpc to the next OST.
4535 Description: setattr vs write race when updating file timestamps
4536 Details : Client processes that update a file timestamp into the past
4537 right after writing to the file (e.g. tar) it is possible that
4538 the updated file modification time can be reset to the current
4539 time due to a race between processing the setattr and write RPC.
4541 Severity : enhancement
4543 Description: Bring 'lfs find' closer in line with regular Linux find.
4544 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4545 -print0 options and obtains all the needed info through the lustre
4548 Severity : enhancement
4550 Description: support up to 1024 configured devices on one node
4551 Details : change obd_dev array from statically allocated to dynamically
4552 allocated structs as they are first used to reduce memory usage
4557 Description: Flush dirty partially truncated pages during truncate
4558 Details : Immediatelly flush partially truncated pages in filter_setattr,
4559 this way we completely avoid having any pages in page cache on OST
4560 and can retire ugly workarounds during writes to flush such pages.
4565 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4566 Details : protect inode from truncation within vfs_unlink() context
4567 just take a reference before calling vfs_unlink() and release it
4568 when parent's i_sem is free.
4573 Description: last_id value checked outside lock on OST caused LASSERT failure
4574 Details : If there were multiple MDS->OST object precreate requests in
4575 flight, it was possible that the OST's last object id was checked
4576 outside a lock and incorrectly tripped an assertion. Move checks
4577 inside locks, and discard old precreate requests.
4580 Frequency : always, if extents are used on OSTs
4582 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4583 Details : OSTs using the extents format would not zero the high 16 bits of
4584 the index physical block number. This is not a problem for any
4585 OST filesystems smaller than 16TB, and no kernels support ext3
4586 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4587 new/modified files) and can be fixed for existing filesystems
4588 with e2fsprogs-1.39-cfs1.
4593 Description: import connection selection may be incorrect if timer wraps
4594 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4595 import connections to be ignored if the 32-bit jiffies counter
4596 wraps. Use a 64-bit jiffies counter.
4599 Frequency : very large clusters immediately after boot
4601 Description: LNET request buffers exhausted under heavy short-term load
4602 Details : If a large number of client requests are generated on a service
4603 that has previously never seen so many requests it is possible
4604 that the request buffer growth cannot keep up with the spike in
4605 demand. Instead of dropping incoming requests, they are held in
4606 the LND until the RPC service can accept more requests.
4609 Frequency : Sometimes during replay
4611 Description: Assertion failure in ll_local_open after replay.
4612 Details : If replay happened on an open request reply before we were able
4613 to set replay handler, reply will become not swabbed tripping the
4614 assertion in ll_local_open. Now we set the handler right after
4615 recognising of open request
4618 Frequency : very rare
4620 Description: kernel reports "badness in vsnprintf"
4621 Details : Reading from the "recovery_status" /proc file in small chunks
4622 may cause a negative length in lprocfs_obd_rd_recovery_status()
4623 call to vsnprintf() (which is otherwise harmless). Exit early
4624 if there is no more space in the output buffer.
4626 Severity : enhancement
4628 Description: clear OBD RPC statistics by writing to them
4629 Details : It is now possible to clear the OBD RPC statistics by writing
4630 to the "stats" file.
4635 Description: Client mtime is not the same on different clients after utimes
4636 Details : In some cases, the client was using the utimes() syscall on
4637 a file cached on another node. The clients now validate the
4638 ctime from the MDS + OSTs to determine which one is right.
4643 Description: Inability to activate failout mode
4644 Details : lconf script incorrectly assumed that in pythong string's numeric
4645 value is used in comparisons.
4648 Frequency : always with multiple stripes per file
4650 Description: Inefficient object allocation for mutli-stripe files
4651 Details : When selecting which OSTs to stripe files over, for files with
4652 a stripe count that divides evenly into the number of OSTs,
4653 the MDS is always picking the same starting OST for each file.
4654 Return the OST selection heuristic to the original design.
4659 Description: mount failures may take full timeout to return an error
4660 Details : Under some heavy load conditions it is possible that a
4661 failed mount can wait for the full obd_timeout interval,
4662 possibly several minutes, before reporting an error.
4663 Instead return an error as soon as the status is known.
4665 Frequency : quota enabled and large files being deleted
4667 Description: releasing more than 4GB of quota at once hangs OST
4668 Details : If a user deletes more than 4GB of files on a single OST it
4669 will cause the OST to spin in an infinite loop. Release
4670 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4675 Description: statfs data retrieved from /proc may be stale or zero
4676 Details : When reading per-device statfs data from /proc, in the
4677 {kbytes,files}_{total,free,avail} files, it may appear
4678 as zero or be out of date.
4681 Frequency : systems with MD RAID1 external journal devices
4683 Description: lconf's call to blkid is confused by RAID1 journal devices
4684 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4685 of returning all block devices that match the journal UUID.
4688 Frequency : always, for aggregate stripe size over 4GB
4690 Description: assertion fails when trying to use 4GB stripe size
4691 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4692 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4695 Frequency : always on ppc64
4697 Description: the first write on an ext3 filesystem with mballoc got stuck
4698 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4699 perform endianness conversion.
4701 ------------------------------------------------------------------------------
4703 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4705 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4706 INTEROPERATE with older versions automatically. Please read the
4707 user documentation before upgrading any part of a live system.
4708 * WARNING: Lustre networking configuration changes are required with
4709 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4712 * Support for kernels:
4713 2.6.9-22.0.2.EL (RHEL 4)
4714 2.6.5-7.244 (SLES 9)
4715 2.6.12.6 vanilla (kernel.org)
4718 Severity : enhancement
4719 Bugzilla : 7981/8208
4720 Description: Introduced Lustre Networking (LNET)
4721 Details : LNET is new networking infrastructure for Lustre, it includes
4722 a reorganized network configuration mode (see the user
4723 documentation for full details) as well as support for routing
4724 between different network fabrics. Lustre Networking Devices
4725 (LNDs) for the supported network fabrics have also been
4726 created for this new infrastructure.
4728 Severity : enhancement
4729 Description: Introduced Access control lists
4730 Details : clients can set ACLs on files and directories in order to have
4731 more fine-grained permissions than the standard Unix UGO+RWX.
4732 The MDS must be started with the "-o acl" mount option.
4734 Severity : enhancement
4735 Description: Introduced filesystem quotas
4736 Details : Administrators may now establish per-user quotas on the
4739 Severity : enhancement
4741 Description: Configuration change for the XT3
4742 The PTLLND is now used to run Lustre over Portals on the XT3
4743 The configure option(s) --with-cray-portals are no longer used.
4744 Rather --with-portals=<path-to-portals-includes> is used to
4745 enable building on the XT3. In addition to enable XT3 specific
4746 features the option --enable-cray-xt3 must be used.
4751 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4752 Details : A race between allocating a new llcd and re-getting the llcd_lock
4753 allowed another thread to grab newly-allocated llcd.
4755 Severity : enhancement
4757 Description: 2.6 OST async journal commit and locking fix to improve performance
4758 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4759 2.6 kernels are now async as they already were in 2.4 kernels so
4760 that they can commit concurrently with the network bulk transfer.
4761 For block-allocated files the filter allocation semaphore is held
4762 to avoid filesystem fragmentation during allocation. BKL lock
4763 removed for 2.6 xattr operations where it is no longer needed.
4768 Description: lconf incorrectly determined whether two IP networks could talk
4769 Details : In some more complicated routing and multiple-network
4770 configurations, lconf will avoid trying to make a network
4771 connection to a disjoint part of the IP space. It was doing the
4772 math incorrectly for one set of cases.
4777 Description: Fix for potential infinite loop processing records in an llog.
4778 Details : If an llog record is corrupted/zeroed, it is possible to loop
4779 forever in llog_process(). Validate the llog record length
4780 and skip the remainder of the block on error.
4783 Frequency : occasional (liblustre only)
4785 Description: liblustre could not open files whose last component is a symlink
4786 Details : sysio_path_walk() would incorrectly pass the open intent to
4787 intermediate path components.
4790 Frequency : rare (liblustre only with non-standard tuning)
4791 Bugzilla : 7201 (7350)
4792 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4793 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4794 after it has been released, only do this for referenced locks
4796 Severity : enhancement
4798 Description: specifying an (invalid) directory default stripe_size of -1
4799 would reset the directory default striping
4800 Details : stripe_size -1 was used internally to signal directory stripe
4801 removal, now use "all default" to signal dir stripe removal
4802 as a directory striping of "all default" is not useful
4805 Frequency : common for large clusters running liblustre clients
4807 Description: doing an ls when liblustre clients are running is slow
4808 Details : sending a glimpse AST to a liblustre client waits for every AST
4809 to time out, as liblustre clients will not respond. Since they
4810 cannot cache data we refresh the OST lock LVB from disk instead.
4812 Severity : enhancement
4814 Description: doing an ls at the same time as file IO can be slow
4815 Details : enqueue and other "small" requests can be blocked behind many
4816 large IO requests. Create a new OST IO portal for non-IO
4817 requests so they can be processed faster.
4820 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4822 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4823 created file with mode 000
4824 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4825 allow an ftruncate() as a non-root user to a file with mode 000.
4826 Lustre now respects this flag to disable mode checks when
4827 truncating a file owned by the user
4830 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4833 Description: Revoking locks from clients that went dead or catatonic might take
4835 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4836 cancellation of such locks instant on servers without waiting for
4837 any reply from clients. Clients drops these locks when cancel
4838 notification from server is received without replying.
4841 Frequency : liblustre-only, when liblustre client dies or becomes busy
4843 Description: Doing ls on Linux clients can take a long time with active
4845 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4846 granting such locks to it in the first place if possible. Locks
4847 are taken by proxy on the OST during the read or write and
4848 dropped immediately afterward. Add connect flags handling, do
4849 not grant locks to liblustre clients for glimpse ASTs.
4851 Severity : enhancement
4853 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4854 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4855 sequential reads and behaves badly with applications doing random
4856 reads. Improve it by reading ahead at least read region, and
4857 avoiding excessive large RPC for small reads.
4859 Severity : enhancement
4861 Description: Creating more than 1000 files for a single job may cause a load
4862 imbalance on the OSTs if there are also a large number of OSTs.
4863 Details : qos_prep_create() uses an OST index reseed value that is an
4864 even multiple of the number of available OSTs so that if the
4865 reseed happens in the middle of the object allocation it will
4866 still utilize the OSTs as uniformly as possible.
4871 Description: OST or MDS may oops in ping_evictor_main()
4872 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4873 but doesn't restart at beginning of obd_exports_timed list
4876 Severity : enhancement
4878 Description: improve by-nid export eviction on the MDS and OST
4879 Details : allow multiple exports with the same NID to be evicted at one
4880 time without re-searching the exports list.
4883 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4885 Description: MDS may oops in groups_free()
4886 Details : in rare race conditions a newly allocated group_info struct is
4887 freed again, and this can be NULL. The 2.4 compatibility code
4888 for groups_free() checked for a NULL pointer, but 2.6 did not.
4891 Frequency : common for liblustre clients doing little filesystem IO
4892 Bugzilla : 9352, 7313
4893 Description: server may evict liblustre clients accessing contended locks
4894 Details : if a client is granted a lock or receives a completion AST
4895 with a blocking AST already set it would not reply to the AST
4896 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4899 Frequency : lfs setstripe, only systems with more than 160 OSTs
4901 Description: unable to set striping with a starting offset beyond OST 160
4902 Details : llapi_create_file() incorrectly limited the starting stripe
4903 index to the maximum single-file stripe count.
4906 Frequency : LDAP users only
4908 Description: lconf did not handle in-kernel recovery with LDAP properly
4909 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4911 Severity : enhancement
4913 Description: bind OST threads to NUMA nodes to improve performance
4914 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4915 node and do their allocations there to localize memory access
4917 Severity : enhancement
4919 Description: llmount can determine client NID directly from Myrinet (GM)
4920 Details : the client NID code from gmnalnid was moved directly into
4921 llmount, removing the need to use this or specifying the
4922 client NID explicitly when mounting GM clients with zeroconf
4925 Frequency : if client is started with down MDS
4927 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4928 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4929 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4930 ptlrpc_import_delayed_req() because we didn't check if the
4931 request was interrupted, and we also didn't break out of the
4932 event loop if there was no timeout
4937 Description: data loss during non-page-aligned writes to a single file from
4938 both multiple nodes and multiple threads on one node at same time
4939 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4940 inconsistency led to false short-reads, that were cached and later
4941 used by ->prepare_write() to fill in partially written page,
4942 leading to data loss.
4945 Frequency : always, if lconf --abort_recovery used
4947 Description: lconf --abort_recovery fails with 'Operation not supported'
4948 Details : lconf was attempting to abort recovery on the MDT device and not
4951 Severity : enhancement
4953 Description: remove cleanup logs
4954 Details : replace lconf-generated cleanup logs with lustre internal
4955 cleanup routines. Eliminates the need for client-cleanup and
4958 Severity : enhancement
4960 Description: add support for EAs (user and system) on lustre filesystems
4961 Details : it is now possible to store extended attributes in the Lustre
4962 client filesystem, and with the user_xattr mount option it
4963 is possible to allow users to store EAs on their files also
4965 Severity : enhancement
4967 Description: Add possibility (config option) to show minimal available OST free
4969 Details : When compiled with --enable-mindf configure option, statfs(2)
4970 (and so, df) will return least minimal free space available from
4971 all OSTs as amount of free space on FS, instead of summary of
4972 free spaces of all OSTs.
4974 Severity : enhancement
4976 Description: do not expand extent locks acquired on OST-side
4977 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4978 by server: they are not cached anyway.
4981 Frequency : when mmap is used/binaries executed from Lustre
4983 Description: Unmmap pages before throwing them away from read cache.
4984 Details : llap_shrink cache now attempts to unmap pages before discarding
4985 them (if unmapping failed - do not discard). SLES9 kernel has
4986 extra checks that trigger if this unmapping is not done first.
4991 Description: lconf didn't resolve symlinks before checking to see whether a
4992 given mountpoint was already in use
4995 Frequency : when migrating failover services
4996 Bugzilla : 6395, 9514
4997 Description: When migrating a subset of services from a node (e.g. failback
4998 from a failover service node) the remaining services would
4999 time out and evict clients.
5000 Details : lconf --force (implied by --failover) sets the global obd_timeout
5001 to 5 seconds in order to quickly disconnect, but this caused
5002 other RPCs to time out too quickly. Do not change the global
5003 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5005 Severity : enhancement
5006 Frequency : if MDS is started with down OST
5007 Bugzilla : 9439,5706
5008 Description: Allow startup/shutdown of an MDS without depending on the
5009 availability of the OSTs.
5010 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5011 Add appropriate locking and lov-osc refcounts for safe
5012 cleaning. Add osc abort_inflight calls in case the
5013 synchronize never started.
5016 Frequency : occasional (Cray XT3 only)
5018 Description: root not authorized to access files in CRAY_PORTALS environment
5019 Details : The client process capabilities were not honoured on the MDS in
5020 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5021 been accessed by an authorized user then root was able to access
5022 the file on the local client also. The root user capabilities
5023 are now allowed on the MDS, as this environment has secure UID.
5026 Frequency : occasional
5028 Description: ldiskfs "too long searching" message happens too often
5029 Details : A debugging message (otherwise harmless) prints too often on
5030 the OST console. This has been reduced to only happen when
5031 there are fragmentation problems on the filesystem.
5036 Description: Division by zero in statfs when all OSCs are inactive
5037 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5038 lov_statfs divides by value returned by lov_get_stripecnt().
5042 Bugzilla : 9489, 3273
5043 Description: First write from each client to each OST was only 4kB in size,
5044 to initialize client writeback cache, which caused sub-optimal
5045 RPCs and poor layout on disk for the first written file.
5046 Details : Clients now request an initial cache grant at (re)connect time
5047 and so that they can start streaming writes to the cache right
5048 away and always do full-sized RPCs if there is enough data.
5049 If the OST is rebooted the client also re-establishes its grant
5050 so that client cached writes will be honoured under the grant.
5055 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5056 Details : Now I/O RPCs go to different portal number and (presumably) fast
5057 lock requests (and glimses) and other RPCs get their own service
5058 threads pool that should be able to service those RPCs
5061 Severity : enhancement
5063 Description: Ability to exchange lustre version between client and servers and
5064 issue warnings at client side if client is too old. Also for
5065 liblustre clients there is ability to refuse connection of too old
5067 Details : New 'version' field is added to connect data structure that is
5068 filled with version info. That info is later checked by server and
5072 Frequency : rare, liblustre only.
5073 Bugzilla : 9296, 9581
5074 Description: Two simultaneous writes from liblustre at offset within same page
5075 might proceed at the same time overwriting eachother with stale
5077 Details : I/O lock withing llu_file_prwv was released too early, before data
5078 actually was hitting the wire. Extended lock-holding time until
5079 server acknowledges receiving data.
5082 Frequency : extremely rare. Never observed in practice.
5084 Description: avoid generating lustre_handle cookie of 0.
5085 Details : class_handle_hash() generates handle cookies by incrementing
5086 global counter, and can hit 0 occasionaly (this is unlikely, but
5087 not impossible, because initial value of cookie counter is
5088 selected randonly). Value of 0 is used as a sentinel meaning
5089 "unassigned handle" --- avoid it. Also coalesce two critical
5090 sections in this function into one.
5092 Severity : enhancement
5094 Description: allow liblustre clients to delegate truncate locking to OST
5095 Details : To avoid overhead of locking, liblustre client instructs OST to
5096 take extent lock in ost_punch() on client's behalf. New connection
5097 flag is added to handle backward compatibility.
5099 Severity : enhancement
5100 Bugzilla : 4928, 7341, 9758
5101 Description: allow number of OST service threads to be specified
5102 Details : a module parameter allows the number of OST service threads
5103 to be specified via "options ost ost_num_threads={N}" in the
5104 OSS's /etc/modules.conf or /etc/modprobe.conf.
5108 Bugzilla : 6146, 9635, 9895
5109 Description: servers crash with bad pointer in target_handle_connect()
5110 Details : In rare cases when a client is reconnecting it was possible that
5111 the connection request was the last reference for that export.
5112 We would temporarily drop the export reference and get a new
5113 one, but this may have been the last reference and the export
5114 was just destroyed. Get new reference before dropping old one.
5116 Severity : enhancement
5117 Frequency : if client is started with failover MDS
5119 Description: Allow multiple MDS hostnames in the mount command
5120 Details : Try to read the configuration from all specified MDS
5121 hostnames during a client mount in case the "primary"
5124 Severity : enhancement
5126 Description: Stop sending data to evicted clients as soon as possible.
5127 Details : Check if the client we are about to send or are sending data to
5128 was evicted already. (Check is done every second of waiting,
5129 for which l_wait_event interface was extended to allow checking
5130 of exit condition at specified intervals).
5133 Frequency : rare, normally only when NFS exporting is done from client
5135 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5137 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5140 Frequency : always, if srand() is called before liblustre initialization
5142 Description: Liblustre uses system PRNG disturbing its usage by user application
5143 Details : Introduce internal to lustre fast and high-quality PRNG for
5144 lustre usage and make liblustre and some other places in generic
5145 lustre code to use it.
5147 Severity : enhancement
5148 Bugzilla : 9477, 9557, 9870
5149 Description: Verify that the MDS configuration logs are updated when xml is
5150 Details : Check if the .xml configuration logs are newer than the config
5151 logs stored on the MDS and report an error if this is the case.
5152 Request --write-conf, or allow starting with --old_conf.
5154 Severity : enhancement
5156 Description: Handle symlinks in the path when checking if Lustre is mounted.
5157 Details : Resolve intermediate symlinks when checking if a client has
5158 mounted a filesystem to avoid duplicate client mounts.
5163 Description: lconf can hit an error exception but still return success.
5164 Details : The lconf command catches the Command error exception at the top
5165 level script context and will exit with the associated exit
5166 status, but doesn't ensure that this exit status is non-zero.
5171 Description: failure of ptlrpc thread startup can cause oops
5172 Details : Starting a ptlrpc service thread can fail if there are a large
5173 number of threads or the server memory is very fragmented.
5174 Handle this without oopsing.
5177 Frequency : always, only if liblustre and non-default acceptor port was used
5179 Description: liblustre cannot connect to servers with non-default acceptor port
5180 Details : tcpnal_set_default_params() was not called and was therefore
5181 ignoring the environment varaible TCPNAL_PORT, as well as other
5182 TCPNAL_ environment variables
5187 Description: two objects could be created on the same OST for a single file
5188 Details : If an OST is down, in some cases it was possible to create two
5189 objects on a single OST for a single file. No problems other
5190 than potential performance impact and spurious error messages.
5194 Bugzilla : 5681, 9562
5195 Description: Client may oops in ll_unhash_aliases
5196 Details : Client dcache may become inconsistent in race condition.
5197 In some cases "getcwd" can fail if the current directory is
5203 Description: Inode refcounting problems in NFS export code
5204 Details : link_raw functions used to call d_instantiate without obtaining
5205 extra inode reference first.
5209 Bugzilla : 9942, 9903
5210 Description: Referencing freed requests leading to crash, memleaks with NFS.
5211 Details : We used to require that call to ll_revalidate_it was always
5212 followed by ll_lookup_it. Also with revalidate_special() it is
5213 possible to call ll_revalidate_it() twice for the same dentry
5214 even if first occurence returned success. This fix changes semantic
5215 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5216 reference on a request referred from the intent.
5217 ll_intent_release() then releases such a request.
5220 Frequency : rare, normally benchmark loads only
5222 Description: unlinked inodes were kept in memory on the client
5223 Details : If a client is repeatedly creating and unlinking files it
5224 can accumulate a lot of stale inodes in the inode slab cache.
5225 If there is no other client load running this can cause the
5226 client node to run out of memory. Instead flush old inodes
5227 from client cache that have the same inode number as a new inode.
5230 Frequency : SLES9 2.6.5 kernel and long filenames only
5231 Bugzilla : 9969, 10379
5232 Description: utime reports stale NFS file handle
5233 Details : SLES9 uses out-of-dentry names in some cases, which confused
5234 the lustre dentry revalidation. Change it to always use the
5238 Frequency : rare, unless heavy write-truncate concurrency is continuous
5239 Bugzilla : 4180, 6984, 7171, 9963, 9331
5240 Description: OST becomes very slow and/or deadlocked during object unlink
5241 Details : filter_destroy() was holding onto the parent directory lock
5242 while truncating+unlinking objects. For very large objects this
5243 may block other threads for a long time and slow overall OST
5244 responsiveness. It may also be possible to get a lock ordering
5245 deadlock in this case, or run out of journal credits because of
5246 the combined truncate+unlink. Solution is to do object truncate
5247 first in one transaction without parent lock, and then do the
5248 final unlink in a new transaction with the parent lock. This
5249 reduces the lock hold time dramatically.
5252 Frequency : rare, 2.4 kernels only
5254 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5255 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5256 before unmounting their filesystem in order to avoid deadlock.
5257 The kernel_locked() function in 2.4 kernels only checks whether
5258 the kernel lock is held, not whether it is this process that is
5259 holding it as 2.6 kernels do.
5264 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5265 Details : The client ptlrpc code may be trying to reconnect to a down
5266 server before a previous connection attempt has timed out.
5267 Increase the reconnect interval to be longer than the connection
5268 timeout interval to avoid sending duplicate connections to
5272 Frequency : echo_client brw_test command
5274 Description: fix echo_client to work with OST preallocated code
5275 Details : OST preallocation code (5137) didn't take echo_client IO path
5276 into account: echo_client calls filter methods outside of any
5277 OST thread and, hence, there is no per-thread preallocated
5278 pages and buffers to use. Solution: hijack pga pages for IO. As
5279 a byproduct, this avoids unnecessary data copying.
5283 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5284 Description: Client can oops in mdc_commit_close() after open replay
5285 Details : It was possible for the MDS to return an open request with no
5286 transaction number in mds_finish_transno() if the client was
5287 evicted, but without actually returning an error. Clients
5288 would later try to replay that open and may trip an assertion
5289 Simplify the client close codepath, and always return an error
5290 from the MDS in case the open is not successful.
5293 Frequency : rare, 2.6 OSTs only
5295 Description: OST may deadlock under high load on fragmented files
5296 Details : If there was a heavy load and highly-fragmented OST filesystems
5297 it was possible to have all the OST threads deadlock waiting on
5298 allocation of biovecs, because the biovecs were not released
5299 until the entire RPC IO was completed. Instead, release biovecs
5300 as soon as they are complete to ensure forward IO progress.
5302 Severity : enhancement
5304 Description: Support for specifying external journal device at mount
5305 Details : If an OST or MDS device is formatted with an external journal
5306 device, this device major/minor is stored in the ext3 superblock
5307 and may not be valid for failover. Allow detecting and
5308 specifying the external journal at mount time.
5313 Description: Mounting an MDS with pending unlinked files may cause oops
5314 Details : target_finish_recovery() calls mds_postrecov() which returned
5315 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5316 considers this an error and immediately begins cleaning up the
5317 lov, just after starting the mds_lov process
5319 Severity : enhancement
5321 Description: Implement 'lfs df' to report actual free space on per-OST basis
5322 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5323 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5324 usage of objects; '-h' to report in human readable format.
5326 ------------------------------------------------------------------------------
5328 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5335 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5336 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5337 at startup during orphan unlinking. Instead, skip buddy update
5338 before setup, as it will be regenerated after recovery is complete.
5343 Description: If an OST is inactive, its locks might reference stale inodes.
5344 Details : lov_change_cbdata() must iterate over all namespaces, even if
5345 they are inactive to clear inode references from the lock.
5347 Severity : enhancement
5348 Frequency : occasional, if non-standard max_dirty_mb used
5350 Description: Client will block write RPCs if not enough grant
5351 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5352 then the client will block writes while waiting for another RPC
5353 to complete instead of consuming its dirty limit. With change
5354 we get improved performance when max_dirty_mb is small.
5356 Severity : enhancement
5357 Bugzilla : 3389, 6253
5358 Description: Add support for supplementary groups on the MDS.
5359 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5360 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5361 lookups for user supplementary groups into a cache.
5365 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5366 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5367 for liblustre O_EXCL is enforced by MDS.
5371 Description: readdir never returns NULL in liblustre.
5372 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5373 dentry in struct dirent.
5377 Frequency : liblustre only, depends on application IO pattern
5378 Description: liblustre clients evicted if not contacting servers
5379 Details : Don't put liblustre clients into the ping_evictor list, so
5380 they will not be evicted by the pinger ever.
5382 Severity : enhancement
5384 Description: Add ability to evict clients by NID from MDS.
5385 Details : By echoing "nid:$NID" string into
5386 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5387 $NID would be instantly evicted from this MDS and from all active
5388 OSTs connected to it.
5392 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5393 Details : lookup_it_finish() used to query file size from OSTs that was not
5398 Description: service threads change working directory to that of init
5399 Details : Starting lustre service threads may pin the working directory
5400 of the parent thread, making that filesystem busy. Threads
5401 now change to the working directory of init to avoid this.
5405 Frequency : during shutdown only
5406 Description: shutdown with a failed MDS or OST can cause unmount to hang
5407 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5412 Frequency : 2.6 only
5413 Description: chmod/chown may include an extra supplementary group
5414 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5415 supplementary group and if none is specified this is used.
5418 Bugzilla : 5479 (6816)
5420 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5421 Details : If lookup is in progress on a file that is unlinked we might try
5422 to revalidate the inode and fail in revalidate after lookup is
5423 complete and ll_file_open() enqueues the open again but
5424 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5427 Frequency : always, if lconf --abort_recovery used
5429 Description: lconf --abort_recovery fails with 'Operation not supported'
5430 Details : lconf was attempting to abort recovery on the MDT device and not
5433 ------------------------------------------------------------------------------
5435 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5440 Frequency : rare (only unsupported configurations with a node running as an
5442 Bugzilla : 6514, 5137
5443 Description: Mounting a Lustre file system on a node running as an OST could
5445 Details : OSTs now preallocates memory needed to write out data at
5446 startup, instead of when needed, to avoid having to
5447 allocate memory in possibly low memory situations.
5448 Specifically, if the file system is mounted on on OST,
5449 memory pressure could force it to try to write out data,
5450 which it needed to allocate memory to do. Due to the low
5451 memory, it would be unable to do so and the node would
5452 become unresponsive.
5454 Severity : enhancement
5456 Description: Addition of lconf --service command line option
5457 Details : lconf now accepts a '--service <arg>' option, which is
5458 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5460 Severity : enhancement
5462 Description: Failover mode is now the default for OSTs.
5463 Details : By default, OSTs will now run in failover mode. To return to
5464 the old behaviour, add '--failout' to the lmc line for OSTs.
5466 Severity : enhancement
5468 Description: Health checks are now provided for MDS and OSTs
5469 Details : Additional detailed health check information on MSD and OSTs
5470 is now provided through the procfs health_check value.
5473 Frequency : occasional, depends on IO load
5475 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5476 after numerous create/delete cycles
5477 Details : The ext3 inode allocation policy would not allocate new inodes
5478 very well on the OSTs because there are no new directories
5479 being created. Instead we look for groups with free space if
5480 the parent directories are nearly full.
5485 Description: Network or server problems during mount may cause partially
5486 mounted clients instead of returning an error.
5487 Details : The config llog parsing code may overwrite the error return
5488 code during mount error handling, returning success instead
5494 Description: MDS can fail to allocate large reply buffers
5495 Details : After long uptimes the MDS can fail to allocate large reply
5496 buffers (e.g. zconf client mount config records) due to memory
5497 fragmentation or consumption by the buffer cache. Preallocate
5498 some large reply buffers so that these replies can be sent even
5499 under memory pressure.
5503 Frequency : rare (liblustre)
5504 Description: fsx running with liblustre complained that using truncate() to
5505 extend the file doesn't work. This patch corrects that issue.
5506 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5507 ATTR_SIZE and lsm use in llu_setattr_raw.
5511 Frequency : rare, only 2.6 kernels
5512 Description: Unusual file access patterns on the MDS may result in inode
5513 data being lost in very rare circumstances.
5514 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5515 code in the 2.6 kernel caused the inode and/or EA data not to
5516 be read from disk, causing single-file corruption.
5520 Frequency : rare, only 2.6 filesystems using extents
5521 Description: Heavy concurrent write and delete load may cause data corruption.
5522 Details : It was possible under high-load situations to have an extent
5523 metadata block in the block device cache from a just-unlinked
5524 file overwrite a newly-allocated data block. We now unmap any
5525 metadata buffers that alias just-allocated data blocks.
5529 Frequency : filesystems with default stripe_count larger than 77
5530 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5531 Details : lconf specifies an inode size of 4096 bytes when the default
5532 stripe_count is larger than 77. This conflicts with the default
5533 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5534 this case to avoid pinning too much memory for large EAs.
5536 ------------------------------------------------------------------------------
5538 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5543 Frequency : rare (extremely heavy IO load with hundreds of clients)
5545 Description: Client is evicted, gets IO error writing to file
5546 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5547 caused clients to be evicted for AST timeouts. The fixes in
5548 bug 5192 mean we no longer need to have such short AST timeouts
5549 so ldlm_timeout has been increased.
5552 Frequency : occasional during --force or --failover shutdown under load
5553 Bugzilla : 5949, 4834
5554 Description: Server oops/LBUG if stopped with --force or --failover under load
5555 Details : a collection of import/export refcount and cleanup ordering
5556 issues fixed for safer force cleanup
5559 Frequency : only filesystems larger than 120 OSTs
5560 Bugzilla : 5990, 6223
5561 Description: lfs getstripe would oops on a very large filesystem
5562 Details : lov_getconfig used kfree on vmalloc'd memory
5565 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5566 Bugzilla : 6242, 6243
5567 Description: reading from files that had been truncated to a non-zero size
5568 but never opened returned no data
5569 Details : ll_file_read() reads zeros from no-object files to EOF
5574 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5575 Details : There's little harm in aborting MDS/OSS recovery and letting it
5576 try again, so I removed the LASSERT and return an error instead.
5578 Severity : enhancement
5580 Description: New debugging infrastructure for tracking down data corruption
5581 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5582 (b) cover more of the client-side code path, and (c) try to narrow
5583 down where problems occurred
5587 Bugzilla : 3819, 4364, 4397, 6313
5588 Description: Racing close and eviction MDS could cause assertion in mds_close
5589 Details : It was possible to get multiple mfd references during close and
5590 client eviction, leading to one thread referencing a freed mfd.
5592 Severity: : enhancement
5593 Bugzilla : 3262, 6359
5594 Description: Attempts to reconnect to servers are now more aggressive.
5595 Details : This builds on the enhanced upcall-less recovery that was added
5596 in 1.4.2. When trying to reconnect to servers, clients will
5597 now try each server in the failover group every 10 seconds. By
5598 default, clients would previously try one server every 25 seconds.
5603 Description: After recovery, certain operations trigger a failed
5604 assertion on a client.
5605 Details : Failing over an mds, using lconf -d --failover, while a
5606 client was doing a readdir() call would cause the client to
5607 LBUG after recovery completed and the readdir() was resent.
5609 Severity : enhancement
5611 Description: Default groups are now added by lconf
5612 Details : You can now run lconf --group <servicename> without having to
5613 manually add groups with lmc.
5616 Frequency : occasional
5618 Description: Nodes with an elan id of 0 trigger a failed assertion
5621 Frequency : always when accessing e.g. tty/console device nodes
5623 Description: tty and some other devices nodes cannot be used on lustre
5624 Details : file's private_data field is used by device data and lustre
5625 values in there got lost. New field was added to struct file to
5626 store fs-specific private data.
5629 Frequency : when exporting Lustre via NFS
5631 Description: NFSD failed occasionally when looking up a path component
5632 Details : NFSD is looking up ".." which was broken in ext3 directories
5633 that had grown large enough to become hashed.
5636 Frequency : Clusters with multiple interfaces not on the same subnet
5638 Description: Nodes will repeatedly try to reconnect to an interface which it
5639 cannot reach and report an error to the log.
5640 Details : Extra peer list entries will be created by lconf with some peers
5641 unreachable. lconf now validates the peer before adding it.
5644 Frequency : Only if a default stripe is set on the filesystem root.
5646 Description: Setting a default stripe on the filesystem root prevented the
5647 filesystem from being remounted.
5648 Details : The client was sending extra request flags in the root getattr
5649 request and did not allocate a reply buffer for the dir EA.
5652 Frequency : occasional, higher if lots of files are accessed by one client
5653 Bugzilla : 6159, 6097
5654 Description: Client trips assertion regarding lsm mismatch/magic
5655 Details : While revalidating inodes the VFS looks up inodes with ifind()
5656 and in rare cases can find an inode that is being freed.
5657 The ll_test_inode() code will free the lsm during ifind()
5658 when it finds an existing inode and then the VFS later attaches
5659 this free lsm to a new inode.
5663 Bugzilla : 6422, 7030
5664 Description: MDS deadlock between mkdir and client eviction
5665 Details : Creating a new file via mkdir or mknod (starting a transaction
5666 and getting the ns lock) can deadlock with client eviction
5667 (gets ns lock and trying to finish a synchronous transaction).
5670 Frequency : occasional
5671 Description: While starting a server, the fsfilt_ext3 module could not be
5673 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5674 kernels. Previously, lconf would still use the ext3 name
5675 when trying to load modules. Now, it will correctly use
5676 ext3 on 2.4 and ldiskfs on 2.6.
5678 Severity : enhancement
5679 Description: The default stripe count has been changed to 1
5680 Details : The interpretation of the default stripe count (0, to lfs
5681 or lmc) has been changed to mean striping across a single
5682 OST, rather than all available. For general usage we have
5683 found a stripe count of 1 or 2 works best.
5685 Severity : enhancement
5686 Description: Add support for compiling against Cray portals.
5687 Details : Conditional compiling for some areas that are different
5691 Frequency : occasional
5692 Bugzilla : 6409, 6834
5693 Description: Creating files with an explicit stripe count may lead to
5694 a failed assertion on the MDS
5695 Details : If some OSTs are full or unavailable, creating files may
5696 trigger a failed assertion on the MDS. Now, Lustre will
5697 try to use other servers or return an error to the
5701 Frequency : occasional
5703 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5704 caused lock timeout during readahead (since 1.4.2).
5705 Details : Processes doing readahead might match a lock that hasn't been
5706 granted yet if there are overlapping and conflicting lock
5707 requests. The readahead process waits on ungranted lock
5708 (original lock is CBPENDING), while OST waits for that process
5709 to cancel CBPENDING read lock and eventually evicts client.
5711 Severity : enhancement
5713 Description: Initial enabling of flock support for clients
5714 Details : Implements fcntl advisory locking and file status functions.
5715 This feature is provided as an optional mount flag (default
5716 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5717 locking are implemented yet, and those that are are not guaranteed
5718 to be completely correct in production environments.
5719 mount -t lustre -o [flock|noflock] ...
5722 Frequency : occasional
5724 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5725 assertion in the ext3 JBD (journaling) layer.
5726 Details : The b_committed_data struct is protected by the big kernel lock
5727 in 2.4 kernels, serializing journal_commit_transaction() and
5728 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5729 access to this struct. In 2.6 kernels there is finer grained
5730 locking to improve SMP performance of the JBD layer.
5734 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5739 Description: merge and cleanup kernel patches.
5742 -----------------------------------------------------------------------------
5744 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5746 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5747 versions of Lustre. Please update all servers and clients to
5748 version 1.4.2 or later at the same time. You must also run
5749 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5750 to update the configuration logs.
5752 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5753 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5754 - swab llog records of type '0' so we get proper header size/idx (5861)
5755 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5756 - fix rename of one directory over another leaking an inode (5953)
5757 - avoid SetPageDirty on 2.6 (5981)
5758 - don't re-add just-being-destroyed locks to the waiting list (5653)
5759 - when creating new directories, inherit the parent's custom
5760 striping settings if present parent (3048)
5761 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5762 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5763 - don't hold client locks on temporary worklist from l_lru (5666)
5764 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5765 - automatically evict dead clients (5921)
5766 - Update file size properly in create+truncate+fstat case (6196)
5767 - Do not unhash mountpoint dentries, do not allow removal of
5769 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5770 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5771 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5772 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5773 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5774 - Correctly handle reads of files with no objects (6243)
5775 - lctl recover will also mark a device active if deactivate used (5933)
5777 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5778 - allow --write-conf on an MDS with different nettype than client (5619)
5779 - don't write config llogs to MDS for mounts not from that MDS (5617)
5780 - lconf should create multiple TCP connections from a client (5201)
5781 - init scripts are now turned off by default; run chkconfig --on
5782 lustre and chkconfig --on lustrefs to use them
5783 - upcalls are no longer needed for clients to recover to failover
5785 - add --abort-recovery option to lconf to abort recovery on device
5787 - add support for an arbitrary number of OSTs (3026)
5788 - Quota support protocol changes.
5789 - forward compatibility changes to wire structs (6007)
5790 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5791 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5792 - improved "lustre status" script
5793 - initialize blocksize for non-regular files (6062)
5794 - added --disable-server and --disable-client configure options (5782)
5795 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5796 - Vanilla 2.4.29 support
5797 - increase maximum number of obd devices to 520 (6242)
5798 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5799 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5800 SLES 9 2.6.5 kernels
5801 - limit stripes per file to 160 (the maximum EA size) (6093)
5803 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5806 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5807 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5808 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5809 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5810 - small ext3 extents cleanups and fixes (5733)
5811 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5812 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5813 - avoid needless client->OST connect, fix handle mismatch (5317)
5814 - fix DLM error path that led to out-of-sync client, long delays (5779)
5815 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5816 - fix several locking issues related to i_size (5492,5624,5654,5672)
5817 - don't move pending lock onto export if it is already evicted (5683)
5818 - fix kernel oops when creating .foo in unlinked directory (5548)
5819 - fix deadlock in obdfilter statistics vs. object create (5811)
5820 - use time_{before,after} to avoid timer jiffies wrap (5882)
5821 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5822 - Do not leak request if server was not able to process it (5154)
5823 - If mds_open unable to find parent dir, make that negative lookup(5154)
5824 - don't create new directories with extent-mapping (5909, 5936)
5826 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5827 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5828 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5829 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5830 - support for RHEL4 kernel on i686 (5773)
5831 - provide error messages when incompatible logs are encountered (5898)
5833 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5834 * version 1.4.0.10 (1.4.1 release candidate 1)
5836 - don't keep a lock reference when lock is not granted (4238)
5837 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5838 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5839 - drop import inflight refcount on signal_completed_replay error (5255)
5840 - unlock page after async write error during send (3677)
5841 - handle missing objects in filter_preprw_read properly (5265)
5842 - no transno return for symlink open, don't save no-trasno open (3440)
5843 - don't try to complete elan receive that already failed (4012)
5844 - free RPC server reply state on error (5406)
5845 - clean up thread from ptlrpc_start_thread() on error (5160)
5846 - readahead could read extra page into cache that wasn't ejected (5388)
5847 - prevent races in class_attach/setup/cleanup/detach (5260)
5848 - don't dereference de->d_inode after l_dput of de (5458)
5849 - use "int" for stripe value returned from lock_to_stripe (5544)
5850 - mballoc allocation and error-checking fixes in 2.6 (5504)
5851 - block device patches to fix I/O request sizes in 2.6 (5482)
5852 - look up hostnames for IB nals (5602)
5853 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5854 - don't start multiple acceptors for the same port (5277)
5855 - fix incorrect LASSERT in mds_getattr_name (5635)
5856 - export a proc file for general "ping" checking (5628)
5857 - fix "lfs check" to not block when the MDS is down (5628)
5859 - service request history (4965)
5860 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5861 - create an "evict_client" /proc entry on OSTs, like the MDS has
5862 - fix mount usage message, return errors per mount(8) (5168)
5863 - change grep [] to grep "[]" in tests so they work in more UMLs
5864 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5865 - remove ancient LOV_MAGIC_V0 EA support (5047)
5866 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5867 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5868 - allow readahead allocations to fail when low on memory (5383)
5869 - mmap locking landed again, after considerable improvement (2828)
5870 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5872 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5875 - send OST transaction number in read/write reply to free req (4966)
5876 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5877 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5878 - the watchdog thread now runs as interruptible (5246)
5879 - flock/lockf fixes (but it's still disabled, pending 5135)
5880 - don't use EXT3 constants in llite code (5094)
5881 - memory shortage at startup could cause assertion (5176)
5883 - reorganization of lov code
5884 - single portals codebase
5886 - add extents/mballoc support (5025)
5887 - direct I/O reads in the obdfilter (4048)
5888 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5890 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5893 - send OST transaction number in read/write reply to free req (4966)
5894 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5895 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5896 - the watchdog thread now runs as interruptible (5246)
5897 - handle missing objects in filter_preprw_read properly (5265)
5898 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5899 - drop import inflight refcount on signal_completed_replay error (5255)
5900 - unlock page after async write error during send (3677)
5901 - return original error code on reconstructed replies (3761)
5902 - no transno return for symlink open, don't save no-trasno open (3440)
5904 - add pid to ldlm debugging output (4922)
5905 - bump the watchdog timeouts -- we can't handle 30sec yet
5906 - extra debugging for orphan dentry/inode bug (5259)
5908 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5911 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5912 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5913 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5914 - fix NULL dereference in /proc/sys/portals/routes (4827)
5915 - allow failed mdc_close() operations to be interrupted (4561)
5916 - stop precreate on OST before MDS would time out on it (4778)
5917 - don't send partial-page writes before EOF from client (4410)
5918 - discard client grant for sub-page writes on large-page clients (4520)
5919 - don't free dentries not owned by NFS code, check generation (4806)
5920 - fix lsm leak if mds_create_objects() fails (4801)
5921 - limit debug_daemon file size, always print CERROR messages (4789)
5922 - use transno after validating reply (3892)
5923 - process timed out requests if import state changes (3754)
5924 - update mtime on OST during writes, return in glimpse (4829)
5925 - add mkfsoptions to LDAP (4679)
5926 - use ->max_readahead method instead of zapping global ra (5039)
5927 - don't interrupt __l_wait_event() during strace
5929 - add software watchdogs to catch hung threads quickly (4941)
5930 - make lustrefs init script start after nfs is mounted
5931 - fix CWARN/ERROR duplication (4930)
5932 - return async write errors to application if possible (2248)
5933 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5934 - print NAL number in %x format (4645)
5935 - update barely-supported suse-2.4.21-171 series (4842)
5936 - support for sles 9 %post scripts
5937 - support for building 2.6 kernel-source packages
5938 - support for sles km_* packages
5940 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5943 - ignore -ENOENT errors in osc_destroy (3639)
5944 - notify osc create thread that OSC is being cleaned up (4600)
5945 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5946 - reconstruct ost_handle() like mds_handle() (4657)
5947 - create a new thread to do import eviction to avoid deadlock (3969)
5948 - let lconf resolve symlinked-to devices (4629)
5949 - don't unlink "objects" from directory with default EA (4554)
5950 - hold socknal file ref over connect in case target is down (4394)
5951 - allow more than 32000 subdirectories in a single directory (3244)
5952 - fix blocks count for O_DIRECT writes (3751)
5953 - OST returns ENOSPC from object create when no space left (4539)
5954 - don't send truncate RPC if file size isn't changing (4410)
5955 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5956 - bind to privileged port in socknal and tcpnal (3689)
5958 - rate limit CERROR/CWARN console message to avoid overload (4519)
5959 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5960 - basic mmap support (3918)
5961 - kernel patch series update from b1_4 (4711)
5963 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5966 - avoid crash during MDS cleanup with OST shut down (2775)
5967 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5968 - don't use bad inodes on the MDS (3744)
5969 - dynamic object preallocation to improve recovery speed (4236)
5970 - don't hold spinlock over lock dumping or change debug flags (4401)
5971 - don't zero obd_dev when it is force cleaned (3651)
5972 - print grants to console if they go negative (4431)
5973 - "lctl deactivate" will stop automatic recovery attempts (3406)
5974 - look for existing locks in ldlm_handle_enqueue() (3764)
5975 - don't resolve lock handle twice in recovery avoiding race (4401)
5976 - revalidate should check working dir is a directory (4134)
5978 - don't always mark "slow" obdfilter messages as errors (4418)
5980 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5983 - don't close LustreDB during write_conf until it is done (3860)
5984 - fix typo in lconf for_each_profile (3821)
5985 - allow dumping logs from multiple threads at one time (3820)
5986 - don't allow multiple threads in OSC recovery (3812)
5987 - fix debug_size parameters (3864)
5988 - fix mds_postrecov to initialize import for llog ctxt (3121)
5989 - replace config semaphore with spinlock (3306)
5990 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5991 - don't allow enqueue to complete on a destroyed export (3822)
5992 - down write_lock before checking llog header bitmap (3825)
5993 - recover from lock replay timeout (3764)
5994 - up llog sem before sending rpc (3652)
5995 - reduce ns lock hold times when setting kms (3267)
5996 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5997 - fix NULL deref and obd_dev leak on setup error (3312)
5998 - replace some LBUG about llog ops with error handling (3841)
5999 - don't match INVALID dentries from d_lookup and spin (3784)
6000 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6001 - fix invalid assertion in ptlrpc_set_wait (3880)
6003 - add libwrap support for the TCP acceptor (3996)
6004 - add /proc/sys/portals/routes for non-root route listing (3994)
6005 - allow setting MDS UUID in .xml (2580)
6006 - print the stack of a process that LBUGs (4228)
6008 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6011 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6012 - make sure to unset replay flag from failed open requests (3440)
6013 - if default stripe count is 0, use OST count for inode size (3636)
6014 - update parent mtime/ctime on client for create/unlink (2611)
6015 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6016 - free recovery state on server during a forced cleanup (3571)
6017 - unregister_reply for resent reqs (3063)
6018 - loop back devices mounting and status check on 2.6 (3563)
6019 - fix resource-creation race that can provoke i_size == 0 (3513)
6020 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6021 - more debugging for page-accounting assertion (3746)
6022 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6023 - avoid deadlock after precreation failure (3758)
6024 - fix race and lock order deadlock in orphan handling (3450, 3750)
6025 - add validity checks when grabbing inodes from l_ast_data (3599)
6027 - add /proc/.../recovery_status to obdfilter (3428)
6028 - lightweight CDEBUG infrastructure, debug daemon (3668)
6029 - change default OSC RPC parameters to be better on small clusters
6030 - turn off OST read cache for files smaller than 32MB
6031 - install man pages and include them in rpms (3100)
6032 - add new init script for (un)mounting lustre filesystems (2593)
6033 - run chkconfig in %post for init scripts (3701)
6034 - drop scimac NAL (unmaintained)
6036 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6039 - clean kiobufs before and after use (3485)
6040 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6041 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6042 - change init script to fail more clearly if not run as root (1528)
6043 - allow clients to reconnect during replay (1742)
6044 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6045 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6046 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6047 - fix cond_resched (3554)
6048 - only evict unfinished clients after recovery (3515)
6049 - allow bulk resend, prevent data loss (3570)
6050 - dynamic ptlrpc request buffer allocation (2102)
6051 - don't allow unlinking open directory if it isn't empty (2904)
6052 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6053 - remove extraneous obd dereference causing LASSERT failure (3334)
6054 - don't use get_cycles() when creating temp. files on the mds (3156)
6055 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6056 - handle EEXIST for set-stripe, set proper directory name (3336)
6058 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6059 - fix ksocknal_fmb_callback() error messages (2918)
6061 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6064 - don't copy lvb into (possibly NULL) reply on error (2983)
6065 - don't deref dentry after dput, don't free lvb on error (2922)
6066 - use the kms to determine writeback rpc length (2947)
6067 - increment oti_logcookies when osc is inactive (2948)
6068 - update client's i_blocks count via lvb messages (2543)
6069 - handle intent open/close of special files properly (1557)
6070 - mount MDS with errors=remount-ro, like obdfilter (2009)
6071 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6072 - don't use cancelling-locks' kms values (2947)
6073 - use highest lock extent for kms, not last one (2925)
6074 - don't dereference ERR_PTR() dentry in error handling path (3107)
6075 - fix thread race in portals_debug_dumplog() (3122)
6076 - create lprocfs device entries at setup instead of at attach (1519)
6077 - common AST error handler, don't evict client on completion race (3145)
6078 - zero nameidata in detach_mnt in 2.6 (3118)
6079 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6080 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6081 - initialize RPC timeout timer earlier for 2.6 (3219)
6082 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6083 - print nal/nid for unknown nid (3258)
6084 - additional checks for oscc recovery before doing precreate (3284)
6085 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6086 - don't crash in mdc_close for bad permissions on open (3285)
6087 - zero i_rdev for non-device files (3147)
6088 - clear page->private before handing to FS, better assertion (3119)
6089 - tune the read pipeline (3236)
6090 - fix incorrect decref of invalidated dentry (2350)
6091 - provide read-ahead stats and refine rpc in flight stats (3328)
6092 - don't hold journal transaction open across create RPC (3313)
6093 - update atime on MDS at close time (3265)
6094 - close LDAP connection when recovering to avoid server load (3315)
6095 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6096 - don't leak open file on MDS after open resend (3325)
6097 - serialize filter_precreate and filter_destroy_precreated (3329)
6098 - loop device shouldn't call sync_dev() for nul device (3092)
6099 - clear page cache after eviction (2766)
6100 - resynchronize MDS->OST in background (2824)
6101 - refuse to mount the same filesystem twice on same mountpoint (3394)
6102 - allow llmount to create routes for mounting behind routers (3320)
6103 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6104 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6105 - fix rare problem with rename on htree directories (3417)
6107 - allow default OST striping configuration per directory (1414)
6108 - fix compilation for qswnal for 2.6 kernels (3125)
6109 - increase maximum number of MDS request buffers for large systems
6110 - change liblustreapi to be useful for external progs like lfsck (3098)
6111 - increase local configuration timeout for slow disks (3353)
6112 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6114 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6117 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6118 - don't overwrite extent policy data in reply if lock was blocked (2901)
6119 - drop filter export grants atomically with removal from device (2663)
6120 - del obd_self_export from work_list in class_disconnect_exports (2908)
6121 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6122 - swab reply message in mdc_close, other PPC fixes (2464)
6123 - fix destroying of named logs (2325)
6124 - overwrite old logs when running lconf --write_conf (2264)
6125 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6126 - fix race in target_handle_connect (2898)
6127 - mds_reint_create() should take same inode create lock (2926)
6128 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6129 - don't close files for self_export to avoid uninitialized obd (2936)
6130 - allow MDS with the same name as client node (2939)
6131 - hold dentry reference for closed log files for unlink (2325)
6132 - reserve space for all logs during transactions (2059)
6133 - don't evict page beyond end of stripe extent (2925)
6134 - don't oops on a deleted current working directory (2399)
6135 - handle hard links to targets without a parent properly (2517)
6136 - don't dereference NULL lock when racing during eviction (2867)
6137 - don't grow lock extents when lots of conflicting locks (2919)
6139 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6142 - account for cache space usage on clients to avoid data loss (974)
6143 - lfsck support in lustre kernel code (2349)
6144 - reduce journal credits needed for BRW writes (2370)
6145 - orphan handling to avoid losing space on client/server crashes
6146 - ptlrpcd can be blocked, stopping ALL progress (2477)
6147 - use lock value blocks to assist in proper KMS, faster stat (1021)
6148 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6149 - recovery for initial connections (2355)
6150 - fixes for mds_cleanup_orphans (1934)
6151 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6152 - block all file creations until orphan recovery completes (1901)
6153 - client remove rq_connection from request struct (2423)
6154 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6155 - recovery timer race (2670)
6156 - mdc_close recovey bug (2532)
6157 - ptlrpc cleanup bug (2710)
6158 - mds timeout on local locks (2588)
6159 - namespace lock held during RPCs (2431)
6160 - handle interrupted sync write properly (2503)
6161 - don't try to handle a message that hasn't been replied to (2699)
6162 - client assert failure during cleanup after abort recovery (2701)
6163 - leak mdc device after failed mount (2712)
6164 - ptlrpc_check_set allows timedout requests to complete (2714)
6165 - wait for inflight reqs when ptlrpcd finishes (2710)
6166 - make sure unregistered services are removed from the srv_list
6167 - reset bulk XID's when resending them (caught by 1138 test)
6168 - unregister_bulk after timeout
6169 - fix lconf error (2694)
6170 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6171 - readahead locks pages, leaves pending causing memory pressure (2673)
6172 - increase OST request buffers to 4096 on large machines (2729)
6173 - fix up permission of existing directories in simple_mkdir (2661)
6174 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6175 - don't assign transno to errored transactions (2742)
6176 - don't delete objects on OST if given a bogus objid from MDS (2751)
6177 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6178 - if rq_no_resend, then timeout request after recovery (2432)
6179 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6180 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6181 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6182 - MDS recovery shouldn't delete objects that it has given out (2730)
6183 - if enqueue arrives after completion, don't clobber LVB (2819)
6184 - don't unlock pages twice when trigger_group_io returns error (2814)
6185 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6186 - don't write pages to disk if there was an error (1450)
6187 - don't ping imports that have recovery disabled (2676)
6188 - take buffered bytes into account when balancing socknal conn (2817)
6189 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6190 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6191 - remove little-endian swabbing of llog records (1987)
6192 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6193 - retry reposting request buffers if they fail (1191)
6194 - grow extent at grant time to avoid granting a revoked lock (2809)
6195 - lock revoke doesn't evict page if covered by a second lock (2765)
6196 - disable VM readahead to avoid reading outside lock extents (2805)
6198 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6199 - updated kernel patches for hp-2.4.20 kernel (2681)
6201 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6204 - fix truncated write corruption (2366)
6205 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6207 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6208 - ensure len(uuid) < 37 in lmc (1171)
6209 - fix ia64 OOPS in llog_test (2255)
6210 - zero end of page at obdfilter for partial page writes (2648)
6211 - don't leave stale dentries around after renames (bug 2428)
6212 - fix timeouts when evicting a client with a single lock held (2642)
6213 - set deadline for the initial HELLO message to drain (2634)
6214 - print out dotted-quad IP addresses in the socknal (2302)
6216 - additional debugging for MDS client eviction problem (2443)
6217 - fix mkfsoptions support for osts (2603, 2604)
6219 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6222 - add series for the vanilla 2.6.0 kernel
6223 - add series for the vanilla 2.4.24 kernel
6224 - add series for a cray x86/64 UL kernel drop
6225 - fix xattr patches for the vanilla 2.4.19 series
6227 - generate true UUIDs in lmc (1171)
6228 - have portals stack dumping break in UML (2466)
6229 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6230 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6231 - wake callback waiting threads on client eviction (2460)
6232 - Add --ptldebug and --subsystem to lmc (1719)
6233 - update assertion to allow safe interrupt allocation
6234 - set rq_no_resend for cancel requests (2432)
6235 - recalculate ptlrpcd timeout after resend (2494)
6236 - call vfs_rmdir when removing pending directories (2368)
6237 - fix renaming a file to itself (2429)
6238 - lmc creates a default one-stripe lov (2454)
6239 - expand procfs space to handle large clusters (2326)
6240 - increase UML stack to avoid overflow
6241 - update lconf's list of debug and subsystem masks
6242 - fix lfs find --obd (2510)
6243 - /proc tunable for disabling filter read caching (2591)
6244 - stop rpm packages from altering slapd.conf (2301)
6245 - disable nagle in the socknal under 0conf (2578)
6246 - choose mds inode size based on stripe count (2572)
6247 - fix kernel-source rpm problems (2516)
6249 - add --disable-doc to avoid pdf generation (2421)
6250 - update documentation, tests, type-os, comments
6251 - avoid format warnings on ia64
6252 - remove the TOE NAL
6253 - tiny code cleanups by removing unused fields
6255 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6258 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6259 - remove the most blatant lies from BUILDING file (2371)
6260 - change default debug level to reasonable production setting
6261 - reduce client side cache size to reduce cache flush time
6262 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6263 - make TCP zerocopy and pinger support enabled by default (2476)
6264 - sync writes completed after process exits caused crashes (2319)
6265 - maintain correct mount count on the MDS (2356)
6266 - backout 1557, because 2316 wasn't really fixed
6267 - better file I/O statistics gathering in /proc
6268 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6269 - another kernel patch to fix zero-copy TCP function export
6270 - don't take duplicate lock when processing re-sent getattr (2420)
6271 - lctl uses obd_self_export instead of creating new conn (2353)
6272 - MDS/OST recovery case which requires object creation asserted (2425)
6273 - move lfs from /usr/sbin to /usr/bin in packages
6274 - fix race between mds_client_add and mds_client_free (2417)
6275 - use kmalloc instead of slabs in portals (2430)
6276 - don't create duplicate records when a failover MDS is present (2442)
6277 - remove unnecessary mount age check (2332)
6278 - don't remove directory inodes from locks prematurely (2451)
6279 - don't break if MDS service name is the same as hostname (2103)
6280 - fix races in client write RPC generation when cache full (2482)
6282 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6285 - remove now-unused request->rq_obd (278)
6286 - if an allocation fails, print out how much memory we've used (1933)
6287 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6288 - add the "configurable stack size" patch to most series files (1256)
6289 - ability to write large log records, for 100+ OST configs (2306)
6290 - fix NULL deref when filter_prep fails (2314)
6291 - fix operator precedence error in filter_sync
6292 - dynamic allocation of socknal TX descriptors (2315)
6293 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6294 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6295 - ensure that CWARN messages in Portals always get to the syslog
6296 - __init/__exit are not for prototype decls (ldlm_init/exit)
6297 - x86-64 compile warning fixes
6298 - fix gateway LMC keyword conflict (2318)
6299 - fix MDS lock inversions in getattr/reint paths (1844)
6300 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6301 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6302 - DLM race condition prevented some lock evictions (2328)
6303 - ENOMEM detection and retry on socknal sends (2230)
6304 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6305 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6306 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6307 - better error messages when a client is rejected during recovery (1505)
6308 - avoid cancelling locks which were never granted, after failure (2330)
6309 - fix i_sem/journal inversion in mds_client_add (2333)
6310 - fix truncate/getattr lock cycle deadlock (2334)
6311 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6312 - fix two rare exit paths which could leak an l_lock() ref (2321)
6313 - fencepost error in MDS/OST orphan recovery (2226)
6314 - make log record alignment 8 bytes (1988)
6315 - lstripe now fails when requested offset > ost_count (2237)
6316 - ensure that all kernel series have a complete list.h (1607)
6317 - fix crashes in special-file operations (2316)
6318 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6320 - allow configurable automake binary, for testing new versions
6321 - small update to the lfs documentation
6323 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6325 * fix negative export reference count in fsfilt_sync (2312)
6327 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6328 * release candidate 0.9.1
6330 - orphans are moved into the PENDING directory for possible recovery
6331 - replayed opens now open by fid for orphan/rename safety (1042)
6332 - last close of an orphan inode generates a transno (683)
6333 - chdir() and mount() now pin the directory entry (1020)
6334 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6335 - discard very old requests without processing them (1502)
6336 - remove some common, well-understood CERRORs (1505)
6337 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6338 - clear "grant" flags in OST replies until OST grant code lands (1644)
6339 - fix read performance by not clobbering i_blksize on client (1598)
6340 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6341 - make LRU size a /proc tunable, clears locks when reduced (707)
6342 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6343 - server locks take references on exports now (1558)
6344 - build fixes for 2.4.20-rh trees (1663)
6345 - return an error from lov_create if all OSCs are inactive (1751)
6346 - fix import levels when a reconnect happens without a timeout (1597)
6347 - exit early from mds_open if we get a lookup error (1749)
6348 - partial page read at EOF wouldn't wait for disk before sending (1642)
6349 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6350 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6351 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6352 - write/truncate lock inversion (1639)
6353 - Don't auto-load obdclass, portals modules during cleanup (1495)
6354 - fix timestamps from jumping to "now" (1763)
6355 - extra journal assertions (1648)
6356 - add an extra multiunlink test (1771)
6357 - fix read_record/write_record API (1776)
6358 - fix leak of offset_extent, possible incorrect i_size later (1772)
6359 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6360 - Debugging for the kqswnal_get_idle_tx problems (1820)
6361 - Allow recovery to be attempted multiple times (1536)
6362 - Write out MDS last_rcvd file after it is first created (1600)
6363 - Fix tx_descriptor leak in failed transmit situations (1827)
6364 - ext3 journaling fixes for assertion failure after IO error (1871)
6365 - class_export_put() on freed export after completion AST error (1896)
6366 - Fix revalidate looping in VFS (1322)
6367 - Don't access a freed export during MDS_REINT timeout (1521)
6368 - Add open-unlink recovery support on the MDS (1673,1764)
6369 - Return an error if no MDS data was read from last_rcvd (1946)
6370 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6371 - Don't setup a disk device that doesn't match exported UUID (317)
6372 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6373 - avoid committing NULL handle in force close
6374 - local.sh is now a one-stripe LOV configuration
6375 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6376 - fix ext3 htree duplicate directory entry corruption (1516)
6377 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6378 - update to new LOV EA format (2097)
6379 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6381 2003-06-15 Phil Schwan <phil@clusterfs.com>
6384 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6385 - per-import/export recovery handling (958, 931, 959)
6386 - multiple last-rcvd slots, for serving multiple FSes (949)
6387 - connections are again shared between multiple imp/exports (963, 964)
6388 - "umount -f" would hang if any requests needed to be sent (393, 978)
6389 - avoid pinning large req buffer by copying for queued messages (989)
6390 - add "uuid" to "lctl device" command to help upcalls (991)
6391 - "open" RPCs with transnos would confuse recovery counters (1037)
6392 - do proper endian conversion of all wire messages (288, 340, 891)
6393 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6394 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6395 - fix ptlrpc_connection leak in target_handle_connect (1174)
6396 - fix import refcounting bug in OST and MDS cleanup (1134)
6397 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6398 - fix very unlikely obd_types race condition (501)
6399 - remove osc_open hack for echo_client (1187)
6400 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6401 - a failure in read_inode2 leads to deadlock (1139)
6402 - cancel ack-locks as soon as transaction is committed (1072)
6403 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6404 - make sure to commitrw after any preprw to avoid deadlock (1162)
6405 - failing to execute a file in a lustre FS would lock inode (1203)
6406 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6407 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6408 - obd_uuid2tgt would walk off the end of the list (1255)
6409 - on IA64 the osc would give portals incorrect bulk size (1258)
6410 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6411 - fix lock inversion caused by new llite matching code (1282)
6412 - limit the number of dirty pages on a client to 10MB (1286)
6413 - timed out locks were not being corrected cancelled (1289)
6414 - fix O_DIRECT above 4GB on IA-32 (1292)
6415 * major user-visible changes
6416 - fail out/fail over policy now controlled by the upcall (993)
6418 - add OBD_PING to check server availability and failure (954)
6419 - lustre messages are now sent in sending host order (288, 340, 891)
6420 - add eadatalen to MDS getattr reply (340)
6421 - OST read replies may contain second buffer, with per-page status (593)
6423 2003-03-11 Phil Schwan <phil@clusterfs.com>
6426 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6427 - failed open()s could cause deadlock; fixed (867, 869)
6428 - stop cancelling OST locks when files are closed (481)
6429 - overlapping XID spaces caused network corruption (851, 853)
6430 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6431 - setattr_raw added, to do single-RPC, server-side setattrs
6432 - lmc/lconf syntax change for OST UUIDs
6433 - fix crashy race condition between ptlrpc_free_req and osc_close
6434 - don't use request in mdc_enqueue if we hit a timeout (889)
6435 - don't set the inode i_size for regular files from the MDS (896)
6436 - handle out of order completion AST (842)
6437 - don't LBUG if a lock request times out after receiving AST (913)
6438 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6439 - if a bad lock AST arrives, send an error instead of dropping entirely
6440 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6441 - fix leak in bulk IO when only partially completed (899, 900, 926)
6442 - fix O_DIRECT for ia64 (55)
6443 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6444 - C-z after timeout could hang a process forever; fixed (977)
6446 - client-side I/O cache (678, 924, 929, 941, 970)
6448 - READPAGE and SETATTRs which don't take server-side locks get
6451 2003-02-11 Phil Schwan <phil@clusterfs.com>
6454 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6455 - Fix fsfilt_extN_readpage() to read a full page of directory
6456 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6457 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6458 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6459 - Most intent operations were removed, in favour of a new RPC mode
6460 that does a single RPC to the server and bypasses most of the VFS
6461 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6462 - Aggressively cancel local locks on DLM servers
6463 - mds_reint_unlink sends EA to the client if it's the last nlink.
6464 client uses that EA to unlink OST objects.
6465 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6466 - recursive symlinks were fixed (439)
6467 - fixed NULL deref in DEBUG_REQ
6468 - filter_update_lastobjid no longer calls sync, which annoyed extN
6469 - fixed multi-client small-writes to a single file problem (445)
6470 - fixed mtime updates during file writes (607)
6471 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6472 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6473 - fixed Linux OST size reporting problem (444, 656)
6474 - OST now updates object mtime with writes or setattr (607, 619)
6475 - client verifies file size before zeroing page past EOF (445)
6476 - OST now writes last allocated objid to disk with allocation (108)
6477 - LOV on echo now works (409)
6479 - mds_reint_unlink sends a new buffer, with the EA included. this
6480 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6481 set if a regular file was being unlinked, and it was the last link
6482 - use PtlGet from the target for bulk writes (315)
6483 - OST now updates object mtime with writes or setattr (607, 619)
6484 - LDLM now has a grant-time callback to revalidate locked items, if
6486 - Many MDS operations were reorganized to combat race conditions
6488 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6489 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6491 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6494 - Fully reactivate OST imports after reconnection (512, others)
6495 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6496 - More graceful error handling for truncating on dead OST (515)
6497 - Don't error out unless we're actually accessing dead stripes (474)
6498 - Fix garbage sizes when stripes are missing (410)
6499 - LRU counters were broken, causing constant lock purge (433, 432)
6500 - garbage on read from stripes with failed OSTs (441)
6501 - mark OSCs as active before reconnecting during recovery (438)
6502 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6503 - lfind did not preserve OST order in output (443)
6504 - symlinks cause hung clients, incorrect data (439)
6505 - stop dereferencing request after dropping refcount (457)
6506 - don't LASSERT(spin_is_locked) on non-SMP (455)
6507 - fixes for many rename() bugs
6508 - fstat didn't correctly synchronize attributes (399)
6509 - server must handle lock cancellation during blocking AST prep (487)
6510 - bulk descriptors were free()d too soon (511)
6511 - fix paths in lconf, which would load incorrect modules (451, 507)
6512 - fix confusing lconf 'host not found' error message (386)
6513 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6514 - fix race condition in mdc_blocking_ast() for inode access (526)
6515 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6516 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6517 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6518 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6519 - fix lstripe in several ways (kernel side) (527)
6520 - fix request leak in ldlm_cli_enqueue (262)
6521 - incorrect OSC was marked inactive after OST failure
6522 - call mds_fs_cleanup before unmounting filesystem (524)
6523 - fix races between taking ns_lock and ldlm_lock_change_resource
6524 - fix races updating LOV export open file list
6525 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6526 - fix recovery NULL deref in ldlm_cli_cancel_unused
6527 - fix some DLM races by using new hash table for lock handles (419)
6528 - permit the client to specify desired inodes, at replay
6529 - duplicate requests when we queue them for replay reintegration
6530 - fix last_rcvd offset calculation
6531 - sync after each recovered transaction, so we always make progress
6532 - never, not always, ERESTART requests without transnos
6533 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6535 - skip replay if the MDS says that the client is already connected
6536 - don't check for a recovery-enabled export to match lctl's UUID
6537 - don't INC_USE_COUNT for phantom exports
6538 - don't crash when cleaning up phantom exports (567)
6539 - don't double-finish or set replay data for errored mdc_open requests
6540 - abort requests when they time out, so we don't get old replies
6541 - send/receive replies for AST messages again
6542 - if the client says that it doesn't have the lock, cancel it on the
6544 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6545 mark it as destroyed, it will all work out when decref is called
6546 - fix module use counts (22, 581)
6548 - ASTs now expect a reply (server cancels lock on error reply)
6550 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6553 - fix many simultaneous client startup (392)
6554 - fix dentry->d_it clobbering
6555 - credentials weren't being shipped for readdir/getattr operations
6556 - remove invalid assertions triggered during some concurrent MD
6558 - proper Lustre versions added (336, 389)
6559 - fix memory leak for create error case (398)
6560 - fix LOV locking bug that would get cli/srv out of sync
6561 - fix echo client over LOV (409)
6562 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6563 - fix double-O_EXCL intent crash (424)
6564 - avoid sending multiple lock CANCELs (352)
6566 - MDS can do multi-client recovery (modulo bugs in new code)
6568 - many updates, edits, cleanups
6570 2002-11-18 Phil Schwan <phil@clusterfs.com>
6573 - fix null d_it dereference (346)
6574 - fix full OST/dbench hang (333)
6575 - fix permission problem with file removal (286)
6576 - fix removal of OSCs from LOV when they fail
6577 - fix NULL deref during bulk timeout (214)
6578 - fix problems related to multiple filesystems on one MDS (241)
6579 - fixed serious subtle metadata locking bugs
6580 - free locks on clients when inodes are removed due to memory
6582 - fix inode pointer in lock data (285)
6583 - partial support for multiple MDS on a single host (241)
6584 - data locks weren't cancelled at clear_inode time (290, 311)
6585 - intent locks could lead to unbounded lock growth (205)
6586 - added a maximum lock count, an LRU list, and a flusher
6587 - fix multiple rename (365)
6588 - properly abstracted the echo client
6589 - OSC locked 1 byte too many; fixed
6590 - rewrote brw callback code:
6591 - fixed recovery bugs related to LOVs (306)
6592 - fixed too-many-pages-in-one-write crash (191)
6593 - fixed (again) crash in sync_io_timeout (214)
6594 - probably fixed callback-related race (385)
6596 - Add capability to MDS protocol
6597 - LDLM cancellations and callbacks on different portals
6599 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6602 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6603 - defer open object destruction to close time (601981, 138)
6604 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6605 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6606 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6607 - don't call obd_setattr in ll_file_release for destroyed objects
6608 * protocol change to lustre_msg: move |version| and add |flags|
6609 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6610 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6611 * added replay of create, unlink, link and rename operations during
6612 MDS failover; recovery should be much more robust now
6613 * remove failed OSCs from LOVs (only lov_create uses this so far)
6614 * the lustre-HOWTO was brought (more) up to date (582544)
6616 2002-10-23 Phil Schwan <phil@clusterfs.com>
6619 - in-use dentries weren't being reused properly (617851)
6620 - prevent multiple LDLM setup (599178)
6621 - fix LOV size calculations for truncate (617853)
6622 - fix client handling of MDS intent errors (POSIX)
6623 - fix permission bug in lovstripe.c test (624321)
6624 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6625 - truncate past end of file could corrupt data
6626 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6627 - a race in recovery could return ETIMEDOUT to apps (623947)
6628 - building outside the source directory was fixed
6629 * the lustre-HOWTO was brought (more) up to date (582544)
6630 * major progress was made on recovery functionality
6632 2002-10-10 Phil Schwan <phil@clusterfs.com>
6635 - recovery deadlock fix
6636 - rm -rf causes LBUG fix (617817)
6637 - file open by multiple tasks fix (618962)
6638 - directory permissions bugs (602707 and 620007)
6639 - journal_stop fixed with locking (611313)
6640 - O_APPEND failures resolved (618273, perhaps 614459)
6641 - lconf PATH fix (619770)
6642 - IA64 build fix (621450)
6643 - RPC buffer sizes scale with amount of memory
6645 2002-10-01 Phil Schwan <phil@clusterfs.com>
6648 - locks would be cancelled without throwing away data pages,
6649 resulting in inconsistent data (605627)
6650 - inode attributes were not always being refreshed (605627, 612449)
6651 - lconf now continues to cleanup after lctl reports an error
6652 - MDS now enforces user permissions (602707)
6653 - lprocfs cleanup fixed, but not yet enabled (614157)
6654 - fixed infinite server hang, should a client not respond to an AST
6655 - avoid going into recovery if user calls readlink() with a buffer
6656 that's too small (613941)
6657 - AST RPCs no longer require replies (614867) -- this may be changed
6658 - don't crash server if client sends an IOV that's too big (611336)
6659 - fixed lock conversion deadlock (611892)
6660 - fixed the following of symlinks (614622)
6661 * recovery: the server can remove locks from a client that dies, other
6662 clients can make progress
6663 * more extN patch fixes
6664 * compile-time configurable ptlrpc buffer allocations
6666 - collaborative read cache document
6667 - Lustre Lite Performance CDR document-in-progress
6669 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6672 - fix typo in patch-2.4.18
6674 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6677 - clear ptlrpc request each time in handle_incoming_request()
6678 - unlink of files now destroys the object on the OST
6680 2002-09-19 Peter Braam <braam@clusterfs.com>
6682 * add hard link support
6683 * change obdfile creation method
6684 * kernel patch changed
6686 2002-09-19 Peter Braam <braam@clusterfs.com>
6689 - stack overflow bug in extN fixed
6691 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6693 * documentation updates
6694 - add man pages for config tools
6695 - update tests/README to describe testing with new config tools
6696 - finish metadata API descriptions
6697 * bug fixes and cleanups
6698 - statfs workaround for 16TB limit
6699 - LOV stripe allocation improved, can stripe on subset of OSTs
6700 - LOV file size/IO offset was wrong for files > 4GB in size
6701 - object EA data was being dropped, caused files to be unreadable
6702 - memory overflow with non-LOV OST caused memory corruption
6703 - fixed regression tests to work with new config tools, obdfilter
6704 - fixed bug when directory size became larger than 1 block
6705 - fixed bug (for single client case) when PWD was deleted
6706 - invalidate local directory pages when doing intent-based ops
6707 - avoid LDLM oops when lock callback contained bad data
6709 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6711 * documentation updates
6712 * bug fixes and cleanups
6713 - configuration tools
6716 - 64-bit compile warnings
6717 - 64-bit internal statfs data
6719 * test_brw on persistent OST devices
6721 * lprocfs (disabled)
6723 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6725 * documentation updates
6726 * bug fixes and cleanups
6727 * configuration tools
6729 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6732 * many small fixes to 0_5_4
6733 * io/network handling
6734 * thinkos in MDS operations
6736 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6739 * crucial basic fixes to 0.5.3
6740 * IOR, Iozone work over Elan
6743 2002-08-07 Phil Schwan <phil@clusterfs.com>
6744 * version 0_5_3, our first alpha
6745 * we use the new Portals iovs
6746 * documentation updates
6747 * bug fixes and cleanups
6748 * small changes in the DLM wire protocol
6750 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6751 * version 0_5_1 with some initial stability,
6752 * locking on MD and file I/O.
6753 * documentation updates
6754 * several bug fixes since 0.5.0
6755 * small changes in wire protocol
6757 2002-07-18 Phil Schwan <phil@clusterfs.com>
6759 * delivered as Lustre Light Alpha
6760 * fixed a crash after handling invalid MDS requests
6761 * fixed directory pages for architectures with non-4k pages sizes
6763 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6764 * release version v0_4_4
6765 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6767 * Fixes a number of interruption problems with OST operations.
6768 * Update documentation for portals header changes
6769 * Move all wire protocol structs/defines to lustre_idl.h
6770 * Fixes symlink length bug.
6771 * Add tcpdump to repository.
6773 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6774 * release version v0_4_3
6775 * Fixes statfs for inodes on extN.
6776 * Fixes bug in runtests which would delete /etc/hosts.
6777 * Use 64-bit object IDs wherever possible (not into VFS though)
6778 Remove ost_get_info, which is unused by lustre, and out of date.
6780 2002-07-03 Peter Braam <braam@clusterfs.com>
6781 * release version v0_4_2 Fixes a lookup error (type not passed)
6782 * move forward to head of Portals
6783 * move forward to latest Lustre kernel
6785 2002-06-25 Peter Braam <braam@clusterfs.com>
6786 * release version v0_4_1. Hopefully stable on single node use.