3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server known to build on the following kernels (others may also work):
6 2.6.32-696.18.7.el6 (RHEL6.9)
7 3.10.0-862.14.4.el7 (RHEL7.5)
8 3.10.0-957.21.3.el7 (RHEL7.6)
9 4.4.120-92.70 (SLES12 SP2)
10 4.4.180-94.97 (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 3.10.0-957.21.3.el7 (RHEL7.6)
20 3.10.0-1062.el7 (RHEL7.7)
21 4.14.0-49.13.1.el7a (RHEL7.5)
22 4.14.0-115.2.2.el7a (RHEL7.6)
23 4.18.0-80.4.2.el8 (RHEL8.0)
24 4.4.120-92.70 (SLES12 SP2)
25 4.4.180-94.97 (SLES12 SP3)
26 4.12.14-95.19 (SLES12 SP4)
27 4.4.0-142 (Ubuntu 16.04)
28 4.15.0-45 (Ubuntu 18.04)
29 vanilla linux 4.15.0 (ZFS only)
30 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
31 * Recommended ZFS / SPL version: 0.7.13
32 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
33 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
34 stack overflow. For more information, please refer to bugzilla 17630.
35 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
36 the re-exporting nodes to be mounted with "32bitapi" mount option
37 * Note for Ubuntu 14.04.5:
38 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
39 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
43 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
44 for currently supported client and server kernel versions.
45 * Server known to build on the following kernels (others may also work):
46 2.6.32-696.18.7.el6 (RHEL6.9)
47 3.10.0-862.14.4.el7 (RHEL7.5)
48 3.10.0-957.10.1.el7 (RHEL7.6)
49 4.4.120-92.70 (SLES12 SP2)
50 4.4.162-94.69 (SLES12 SP3)
51 4.4.0-142 (Ubuntu 16.04)
52 4.15.0-45 (Ubuntu 18.04)
53 vanilla linux 4.6.7 (ZFS only)
54 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
55 * Client known to build on unpatched kernels (others may also work):
56 2.6.32-696.18.7.el6 (RHEL6.9)
57 2.6.32-754.6.3.el6 (RHEL6.10)
58 3.10.0-862.14.4.el7 (RHEL7.5)
59 3.10.0-957.10.1.el7 (RHEL7.6)
60 4.14.0-49.13.1.el7a (RHEL7.5)
61 4.14.0-115.2.2.el7a (RHEL7.6)
62 4.4.120-92.70 (SLES12 SP2)
63 4.4.162-94.69 (SLES12 SP3)
64 4.12.14-95.13 (SLES12 SP4)
65 4.4.0-142 (Ubuntu 16.04)
66 4.15.0-45 (Ubuntu 18.04)
67 vanilla linux 4.15.0 (ZFS only)
68 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
69 * Recommended ZFS / SPL version: 0.7.13
70 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
71 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
72 stack overflow. For more information, please refer to bugzilla 17630.
73 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
74 the re-exporting nodes to be mounted with "32bitapi" mount option
75 * Note for Ubuntu 14.04.5:
76 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
77 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
82 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
83 for currently supported client and server kernel versions.
84 * Server known to build on the following kernels (others may also work):
85 2.6.32-696.18.7.el6 (RHEL6.9)
86 3.10.0-862.14.4.el7 (RHEL7.5)
87 3.10.0-957.10.1.el7 (RHEL7.6)
88 4.4.120-92.70 (SLES12 SP2)
89 4.4.162-94.69 (SLES12 SP3)
90 4.4.0-142 (Ubuntu 16.04)
91 4.15.0-45 (Ubuntu 18.04)
92 vanilla linux 4.6.7 (ZFS only)
93 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
94 * Client known to build on unpatched kernels (others may also work):
95 2.6.32-696.18.7.el6 (RHEL6.9)
96 2.6.32-754.6.3.el6 (RHEL6.10)
97 3.10.0-862.14.4.el7 (RHEL7.5)
98 3.10.0-957.10.1.el7 (RHEL7.6)
99 4.14.0-49.13.1.el7a (RHEL7.5)
100 4.14.0-115.2.2.el7a (RHEL7.6)
101 4.4.120-92.70 (SLES12 SP2)
102 4.4.162-94.69 (SLES12 SP3)
103 4.12.14-95.13 (SLES12 SP4)
104 4.4.0-142 (Ubuntu 16.04)
105 4.15.0-45 (Ubuntu 18.04)
106 vanilla linux 4.15.0 (ZFS only)
107 * Recommended e2fsprogs version: 1.44.5.wc1 or newer
108 * Recommended ZFS / SPL version: 0.7.13
109 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
110 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
111 stack overflow. For more information, please refer to bugzilla 17630.
112 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
113 the re-exporting nodes to be mounted with "32bitapi" mount option
114 * Note for Ubuntu 14.04.5:
115 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
116 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
120 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
121 for currently supported client and server kernel versions.
122 * Server known to build on the following kernels (others may also work):
123 2.6.32-696.18.7.el6 (RHEL6.9)
124 3.10.0-862.14.4.el7 (RHEL7.5)
125 3.10.0-957.el7 (RHEL7.6)
126 4.4.120-92.70 (SLES12 SP2)
127 4.4.162-94.69 (SLES12 SP3)
128 4.4.0-131 (Ubuntu 16.04)
129 4.15.0-32 (Ubuntu 18.04)
130 vanilla linux 4.6.7 (ZFS only)
131 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
132 * Client known to build on unpatched kernels (others may also work):
133 2.6.32-696.18.7.el6 (RHEL6.9)
134 2.6.32-754.6.3.el6 (RHEL6.10)
135 3.10.0-862.14.4.el7 (RHEL7.5)
136 3.10.0-957.el7 (RHEL7.6)
137 4.14.0-49.13.1.el7a (RHEL7.5)
138 4.14.0-115.2.2.el7a (RHEL7.6)
139 4.4.120-92.70 (SLES12 SP2)
140 4.4.162-94.69 (SLES12 SP3)
141 4.4.0-131 (Ubuntu 16.04)
142 4.15.0-32 (Ubuntu 18.04)
143 vanilla linux 4.15.0 (ZFS only)
144 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
145 * Recommended ZFS / SPL version: 0.7.9
146 * Tested with ZFS / SPL version: 0.7.9
147 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
148 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
149 stack overflow. For more information, please refer to bugzilla 17630.
150 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
151 the re-exporting nodes to be mounted with "32bitapi" mount option
152 * Note for Ubuntu 14.04.5:
153 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
154 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
156 --------------------------------------------------------------------------------
159 2018-04-03 Intel Corporation
161 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
162 for currently supported client and server kernel versions.
163 * Server known to build on patched kernels:
164 2.6.32-431.29.2.el6 (RHEL6.5)
165 2.6.32-504.30.3.el6 (RHEL6.6)
166 2.6.32-573.26.1.el6 (RHEL6.7)
167 2.6.32-642.15.1.el6 (RHEL6.8)
168 2.6.32-696.18.7.el6 (RHEL6.9)
169 3.10.0-693.21.1.el7 (RHEL7.4)
170 3.0.101-0.47.71 (SLES11 SP3)
171 3.0.101-107 (SLES11 SP4)
172 3.12.74-60.64.40 (SLES12 SP1)
173 4.4.114-92.64 (SLES12 SP2)
174 4.4.114-94.11 (SLES12 SP3)
175 3.13.0-101 (Ubuntu 14.04, ZFS only)
176 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
177 4.4.0-116 (Ubuntu 16.04)
178 vanilla linux 4.6.7 (ZFS only)
179 * Client known to build on unpatched kernels:
180 2.6.32-431.29.2.el6 (RHEL6.5)
181 2.6.32-504.30.3.el6 (RHEL6.6)
182 2.6.32-573.26.1.el6 (RHEL6.7)
183 2.6.32-642.15.1.el6 (RHEL6.8)
184 2.6.32-696.18.7.el6 (RHEL6.9)
185 3.10.0-693.21.1.el7 (RHEL7.4)
186 3.0.101-0.47.71 (SLES11 SP3)
187 3.0.101-107 (SLES11 SP4)
188 3.12.74-60.64.40 (SLES12 SP1)
189 4.4.114-92.64 (SLES12 SP2)
190 4.4.114-94.11 (SLES12 SP3)
191 3.13.0-101 (Ubuntu 14.04)
192 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
193 4.4.0-116 (Ubuntu 16.04)
195 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
196 * Recommended ZFS / SPL version: 0.7.6
197 * Tested with ZFS / SPL version: 0.7.6
198 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
199 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
200 stack overflow. For more information, please refer to bugzilla 17630.
201 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
202 the re-exporting nodes to be mounted with "32bitapi" mount option
203 * Note for Ubuntu 14.04.5:
204 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
205 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
207 --------------------------------------------------------------------------------
209 2017-07-05 Intel Corporation
211 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
212 for currently supported client and server kernel versions.
213 * Server known to build on patched kernels:
214 2.6.32-431.29.2.el6 (RHEL6.5)
215 2.6.32-504.30.3.el6 (RHEL6.6)
216 2.6.32-573.26.1.el6 (RHEL6.7)
217 2.6.32-642.15.1.el6 (RHEL6.8)
218 2.6.32-696.3.1.el6 (RHEL6.9)
219 3.10.0-514.21.1.el7 (RHEL7.3)
220 3.0.101-0.47.71 (SLES11 SP3)
221 3.0.101-100 (SLES11 SP4)
222 3.12.74-60.64.40 (SLES12 SP1)
223 4.4.59-92.17 (SLES12 SP2)
224 vanilla linux 4.6.7 (ZFS only)
225 * Client known to build on unpatched kernels:
226 2.6.32-431.29.2.el6 (RHEL6.5)
227 2.6.32-504.30.3.el6 (RHEL6.6)
228 2.6.32-573.26.1.el6 (RHEL6.7)
229 2.6.32-642.15.1.el6 (RHEL6.8)
230 2.6.32-696.3.1.el6 (RHEL6.9)
231 3.10.0-514.21.1.el7 (RHEL7.3)
232 3.0.101-0.47.71 (SLES11 SP3)
233 3.0.101-100 (SLES11 SP4)
234 3.12.74-60.64.40 (SLES12 SP1)
235 4.4.59-92.17 (SLES12 SP2)
237 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
238 * Recommended ZFS / SPL version: 0.6.5.9
239 * Tested with ZFS / SPL version: 0.6.5.9
240 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
241 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
242 stack overflow. For more information, please refer to bugzilla 17630.
243 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
244 the re-exporting nodes to be mounted with "32bitapi" mount option
246 --------------------------------------------------------------------------------
248 12-07-2016 Intel Corporation
250 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
251 for currently supported client and server kernel versions.
252 * Server known to build on patched kernels:
253 2.6.32-431.29.2.el6 (RHEL6.5)
254 2.6.32-504.30.3.el6 (RHEL6.6)
255 2.6.32-573.26.1.el6 (RHEL6.7)
256 2.6.32-642.11.1.el6 (RHEL6.8)
257 3.10.0-514.el7 (RHEL7.3)
258 3.0.101-0.47.71 (SLES11 SP3)
259 3.0.101-80 (SLES11 SP4)
260 3.12.62-60.64.8 (SLES12 SP1)
261 vanilla linux 4.5.7 (ZFS only)
262 * Client known to build on unpatched kernels:
263 2.6.32-431.29.2.el6 (RHEL6.5)
264 2.6.32-504.30.3.el6 (RHEL6.6)
265 2.6.32-573.26.1.el6 (RHEL6.7)
266 2.6.32-642.11.1.el6 (RHEL6.8)
267 3.10.0-514.el7 (RHEL7.3)
268 3.0.101-0.47.71 (SLES11 SP3)
269 3.0.101-80 (SLES11 SP4)
270 3.12.62-60.64.8 (SLES12 SP1)
272 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
273 * Recommended ZFS / SPL version: 0.6.5.8
274 * Tested with ZFS / SPL version: 0.6.5.8
275 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
276 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
277 stack overflow. For more information, please refer to bugzilla 17630.
278 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
279 the re-exporting nodes to be mounted with "32bitapi" mount option
281 --------------------------------------------------------------------------------
283 02-29-2016 Intel Corporation
285 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
286 for currently supported client and server kernel versions.
287 * Server known to build on patched kernels:
288 2.6.32-431.29.2.el6 (RHEL6.5)
289 2.6.32-504.30.3.el6 (RHEL6.6)
290 2.6.32-573.12.1.el6 (RHEL6.7)
291 3.10.0-327.3.1.el7 (RHEL7.2)
292 3.0.101-0.47.71 (SLES11 SP3)
293 3.0.101-68 (SLES11 SP4)
294 vanilla linux 4.2.1 (ZFS only)
295 * Client known to build on unpatched kernels:
296 2.6.32-431.29.2.el6 (RHEL6.5)
297 2.6.32-504.30.3.el6 (RHEL6.6)
298 2.6.32-573.12.1.el6 (RHEL6.7)
299 3.10.0-327.3.1.el7 (RHEL7.2)
300 3.0.101-0.47.71 (SLES11 SP3)
301 3.0.101-68 (SLES11 SP4)
304 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
305 * Recommended ZFS / SPL version: 0.6.4.2
306 * Tested with ZFS / SPL version: 0.6.4.2
307 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
308 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
309 stack overflow. For more information, please refer to bugzilla 17630.
310 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
311 the re-exporting nodes to be mounted with "32bitapi" mount option
313 --------------------------------------------------------------------------------
315 03-10-2015 Intel Corporation
317 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
318 for currently supported client and server kernel versions.
319 * Server known to build on patched kernels:
320 2.6.32-431.29.2.el6 (RHEL6.5)
321 2.6.32-504.8.1.el6 (RHEL6.6)
322 3.0.101-0.46 (SLES11 SP3)
323 * Client known to build on unpatched kernels:
324 2.6.32-431.29.2.el6 (RHEL6.5)
325 2.6.32-504.8.1.el6 (RHEL6.6)
326 3.10.0-123.20.1.el7 (RHEL7)
327 3.0.101-0.46 (SLES11 SP3)
328 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
329 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
330 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
331 stack overflow. For more information, please refer to bugzilla 17630.
332 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
333 the re-exporting nodes to be mounted with "32bitapi" mount option
335 Severity : enhancement
337 Description: control OST-index in IDIF via ROCOMPAT flag.
338 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
339 last_rcvd file. For new formatted OST device, it will be auto set;
340 for the case of upgrading from old OST device, you can enable it
341 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
342 enabled, for new created OST-object, its IDIF-in-LMA will contain
343 the OST-index; for the existing OST-object, the OSD will convert
344 old format IDIF as new format IDIF with OST-index stored in the
345 LMA EA when accessing such OST-object or via OI scrub. Once such
346 flag is enabled, it cannot be reverted back, so the system cannot
347 be downgraded to the orignal incompatible version.
349 --------------------------------------------------------------------------------
351 07-30-2014 Intel Corporation
353 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
354 for currently supported client and server kernel versions.
355 * Server known to build on patched kernels:
356 2.6.32-431.20.3.el6 (RHEL6.5)
357 3.0.101-0.31 (SLES11 SP3)
358 * Client known to build on unpatched kernels:
359 2.6.32-431.20.3.el6 (RHEL6.5)
360 3.0.101-0.31 (SLES11 SP3)
363 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
364 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
365 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
366 stack overflow. For more information, please refer to bugzilla 17630.
367 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
368 the re-exporting nodes to be mounted with "32bitapi" mount option
370 --------------------------------------------------------------------------------
372 10-31-2013 Intel Corporation
374 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
375 * Server builds for kernels:
376 2.6.32-358.18.1.el6 (RHEL6.4)
377 * Client builds for unpatched kernels:
378 2.6.32-358.18.1.el6 (RHEL6.4)
379 3.0.92-0.8 (SLES11 SP3)
381 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
382 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
383 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
384 stack overflow. For more information, please refer to bugzilla 17630.
385 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
386 the re-exporting nodes to be mounted with "32bitapi" mount option
388 --------------------------------------------------------------------------------
390 05-30-2013 Intel Corporation
392 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
393 * Server builds for kernels:
394 2.6.32-358.6.2.el6 (RHEL6.4)
395 * Client builds for unpatched kernels:
396 2.6.32-358.6.2.el6 (RHEL6.4)
397 3.0.74-0.6.6 (SLES11 SP2)
399 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
400 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
401 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
402 stack overflow. For more information, please refer to bugzilla 17630.
403 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
404 the re-exporting nodes to be mounted with "32bitapi" mount option
406 --------------------------------------------------------------------------------
408 9-30-2012 Whamcloud, Inc.
410 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
411 * Server support for kernels:
412 2.6.32-279.5.1.el6 (RHEL6)
413 * Client support for unpatched kernels:
414 2.6.32-279.5.1.el6 (RHEL6)
416 2.6.18-238.19.1.el5 (RHEL5)
417 2.6.16.60-0.69.1 (SLES10)
418 * Recommended e2fsprogs version: 1.41.90.wc4
419 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
420 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
421 stack overflow. For more information, please refer to bugzilla 17630.
422 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
423 the re-exporting nodes to be mounted with "32bitapi" mount option
425 Severity : enhancement
427 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
429 Severity : enhancement
431 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
433 Severity : enhancement
435 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
437 Severity : enhancement
439 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
441 Severity : enhancement
443 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
445 --------------------------------------------------------------------------------
447 04-30-2012 Whamcloud, Inc.
449 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
450 * Server support for kernels:
451 2.6.32-220.4.2.el6 (RHEL6)
452 * Client support for unpatched kernels:
453 2.6.32-220.4.2.el6 (RHEL6)
455 2.6.18-238.19.1.el5 (RHEL5)
456 2.6.16.60-0.69.1 (SLES10)
457 * Recommended e2fsprogs version: 1.41.90.wc4
458 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
459 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
460 stack overflow. For more information, please refer to bugzilla 17630.
461 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
462 the re-exporting nodes to be mounted with "32bitapi" mount option
464 Severity : enhancement
466 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
468 Severity : enhancement
470 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
472 Severity : enhancement
474 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
476 Severity : enhancement
478 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
480 --------------------------------------------------------------------------------
482 09-30-2011 Whamcloud, Inc.
484 * Server support for kernels:
485 2.6.18-238.19.1.el5 (RHEL5)
486 2.6.32-131.6.1.el6 (RHEL6)
487 * Client support for unpatched kernels:
488 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
489 2.6.18-238.19.1.el5 (RHEL5)
490 2.6.32-131.6.1.el6 (RHEL6)
491 2.6.29.4-167.fc11 (FC11)
492 2.6.16.60-0.69.1 (SLES10)
494 * Recommended e2fsprogs version: 1.41.90.wc3
495 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
496 Lustre filesystem with 4K stack may cause a stack overflow. For more
497 information, please refer to bugzilla 17630.
498 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
499 mounted with "32bitapi" mount option
501 Severity : enhancement
502 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
504 Severity : enhancement
505 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
507 Severity : enhancement
508 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
511 Frequency : only with SLES10
513 Description: Use OFED "KMP" provided by Novell
514 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
515 Lustre is now built against this package. That means you need to
516 install the ofed-kmp package from Novell for the patchless client.
517 Note that the ofed-kmp that Novell ships may not exactly match the
518 kernel version but should still be compatible.
520 Severity : enhancement
522 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
523 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
525 --------------------------------------------------------------------------------
527 2010-07-15 Oracle, Inc.
529 * Server support for kernels:
530 2.6.18-164.11.1.el5 (RHEL 5)
531 2.6.18-164.11.1.0.1.el5 (OEL 5)
532 * Client support for unpatched kernels:
533 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
534 2.6.18-164.11.1.el5 (RHEL 5),
535 2.6.18-164.11.1.0.1.el5 (OEL 5)
536 2.6.16.60-0.42.8 (SLES 10),
538 2.6.29.4-167.fc11 (FC11)
539 * Recommended e2fsprogs version: 1.41.10-sun2
541 Severity : enhancement
543 Description: Distribute lfs_migrate script with lustre
544 Details : The lfs_migrate mentioned in the manual is now included in the
545 lustre tree in lustre/scripts and will distributed in lustre RPMs.
547 Severity : enhancement
549 Description: Failover nid list added to import proc file.
551 Severity : enhancement
553 Description: "som_preview" mount option required to use SOM
554 Details : The clients will need to use som_preview mount option
558 Description: Lustre does not do 1MB IOs to HW RAID
559 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
561 Severity : enhancement
563 Description: lfs quota output is not very convenient for awk/sed-parsing
564 Details : Some positions in lfs quota output table could be empty or
565 non-empty which made it hard to parse it with scripts, now
566 a dash is put instead of space where there is not supposed
571 Description: Update to OFED 1.5.1
572 Details : OFED 1.5.1 will be used in environments that don't use
573 vendor supplied OFED.
577 Description: Remove lnet. prefix from lctl params display
578 Details : Also renamed lnet.memused to lnet_memused to
583 Description: Use conf_param -d to remove permanant settings.
587 Description: Add list_param -R
588 Details : lctl list_param now supports recursive mode.
591 Bugzilla : 21829, 22312
592 Description: Fix llobdstat.
593 Details : 1. Now updating the write_bytes counter
594 2. Port script fixes from 1.8.
598 Description: decrease the usage of memory on clients.
599 Details : 1. On clients, recycle dentries and inodes unused.
600 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
605 Description: llobdstat fix and enhancement
606 Details : add a counter to set a limit to how many samples will be returned
607 fix a wildcard in the path to limit to obdfilter stats only
611 Description: lfs check only allowed for root
612 Details : Code cleanups and added a regresion test of non-root use of
617 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
618 Details : Fix hr->hr_index overload.
622 Description: Fix FIEMAP functionality.
626 Description: FC11 patchless client support added
630 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
631 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
635 Description: New acc-sm script and various test-framework enhancments.
639 Description: Previously landed patch was reverted.
643 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
647 Description: Add ldiskfs maxdirsize mount option.
651 Description: Fix mkfs.lustre support for 16TB LUNs.
655 Description: Print mmp_check_interval
656 Details : print mmp_check_interval and make it possible to abort
657 mount operation in case it takes too long.
661 Description: Add cascading_rw to lustre mpi tests.
665 Description: Store SOM data in EA.
666 Details : Also several SOM related cleanups.
670 Description: Add refcount to md_open_data
674 Description: Use refcount on lov_request_set to prevent memory corruption.
678 Description: Statahead: use parent dir's i_mutex for synchronization.
682 Description: loadgen improvements.
686 Description: Enable peer health detection in ksocklnd and ko2iblnd.
690 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
694 Description: Fix race on obd_devcie:md_stats
698 Description: Fix SLES 11 client builds.
699 Details : Resolved uninitalized pointer gcc warnings.
703 Description: Allow non-root access for "lfs check".
704 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
708 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
710 Severity : enhancement
712 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
713 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
717 Description: Sanity test_27p fails.
718 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
719 Sanity tests 27 were cleaned up and improved.
723 Description: Unify req format on client/servers
724 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
725 buffers and swabbers approach..
730 Description: drop join file stuff.
732 Severity : enhancement
734 Description: Add 2.6.27 and SLES11 (patchless client) support
738 Description: Fix in ptlrpc_expire_one_request() to print the signed time
741 Severity : enhancement
743 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
746 Severity : enhancement
748 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
750 Severity : enhancement
752 Description: correctly handle big reply message.
753 Details : send LNet event if reply is bigger than buffer and adjust this buffer
758 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
760 Severity : enhancement
762 Description: Add support for OEL5.
766 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
768 Severity : enhancement
770 Description: "weak-modules" support
771 Details : Implement "weak-modules" support which enables kernel modules
772 to be used with any kernel that implements the same kABI. In
773 order to achieve this modules are now installed in
774 /lib/modules/$(uname -r)/updates/kernel on all distributions.
778 Description: client could not reconnect to OST because of an active request.
779 Details : The client now retries to reconnect to the same server, if a
780 connect request failed with EBUSY or -EAGAIN. Test case 35b
781 conf-sanity.sh has been created.
785 Description: Don't allow make backward step on assignin osc next id.
786 Details : race between allocation next id and ll_sync thread can be cause
787 of set wrong osc next id and can be kill valid ost objects.
791 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
795 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
799 Description: Error handling in osc_statfs_interpret() has been improved.
800 Details : Check in osc_statfs_interpret() for EBADR.
804 Description: Do not send statfs() requests to OSTs disabled by administrator.
805 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
807 Severity : enhancement
809 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
811 Severity : enhancement
813 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
815 Severity : enhancement
817 Description: Enhancement for lfs(1) command to use numeric uid/gid.
821 Description: Improved cleanup procedure for conf-sanity test_46a.
825 Description: Do not start bulk transfer, if client tried to reconnect.
826 Details : Do not start bulk transfer, if client tried to reconnect,
827 but failed due to active rpcs. Check for exp_abort_active_req
828 has been implemented.
832 Description: Prevent inconsistences between linux and lustre mount structures.
833 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
834 Make the sleep interruptible.
836 Severity : enhancement
838 Description: provide server to client comms path
839 Details : server to client communications path via new LDLM_SET_INFO rpc
843 Description: truncate starts GFP_FS allocation under transaction causing deadlock
844 Details : ldiskfs_truncate calls grab_cache_page which may start page
845 allocation under an open transaction. This may lead to
846 calling prune_icache with consequent lustre reentrance.
850 Description: send correctly lsm on open replay
851 Details : MDS is trust to LSM size on replay open, but client can set wrong size
856 Description: enable client interface failover
857 Details : When a child reconnects from another NID, properly update export
858 nid hash position and ldlm reverse import.
860 Severity : enhancement
862 Description: Communicate OST degraded/readonly state via statfs to MDS
863 Details : Flags in the statfs returned from OSTs indicate whether the
864 OST is in a degraded RAID state, or if the filesystem has
865 turned read-only after a filesystem error is detected.
870 Description: don't panic if EPROTO was hit when reading symlink
871 Details : correctly handling request reference in error cases.
873 Severity : enhancement
875 Description: Add LustreNetLink, a kernel-userspace communcation path.
877 Severity : enhancement
879 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
884 Description: access to llog context before init.
885 Details : move handling CATALOGS file at osc layer and forbid access to llog
889 Frequency : rare, only when client is writing to OST on local node
891 Description: Avoid deadlock for local client writes
892 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
893 memory freeing context. This allows OST threads to set the
894 PF_MEMALLOC flag on task structures in order to allocate memory
895 from reserved pools and complete IO.
896 Use GFP_HIGHUSER for OST allocations for non-local client writes,
897 so that the OST threads generate memory pressure and allow
898 inactive pages to be reclaimed.
900 Severity : enhancement
902 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
907 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
908 Details : move ldlm namespace creation in setup phase to avoid grab
909 _lprocfs_lock with cli_sem held.
913 Description: Temporarily disable grant shrink.
914 Details : Disable the feature for debugging.
917 Frequency : only when formatting test filesystems
919 Description: Unable to run several mkfs.lustre on loop devices at the same time
920 Details : mkfs.lustre returns error 256 on the concurrent loop devices
921 formatting. The solution is to proper handle the error.
923 Severity : enhancement
925 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
927 Severity : enhancement
929 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
934 Description: resolve race between obd_disconnect and class_disconnect_exports
935 Details : if obd_disconnect will be called to already disconnected export he
936 forget release one reference and osc module can't unloaded.
938 Severity : enhancement
940 Description: Allow tuning service thread via /proc
941 Details : For each service a new
942 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
943 created that can be used to set min/max thread counts, and get the
944 current number of running threads.
949 Descriptoin: don't return error if have partially created objects for file.
950 Details : lov_update_create_set uses set->set_success as index for created
951 objects, so if some requests will be failed, they will have hole
952 at end of array and qos_shrink_lsm for allocate correct lsm.
954 Severity : enhancement
956 Description: Update OFED support to 1.4.1
961 Description: Reduce small size read RPC
962 Details : Set read-ahead limit for every file and only do read-ahead when
963 available read-ahead pages are bigger than 1M to avoid small size
966 Severity : enhancement
968 Description: move AT tunable parameters for more consistent usage
969 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
971 Severity : enhancement
973 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
974 Details : allow skip disconnected ost for send statfs request and hide error
980 Description: fix racy locking of mballoc block bitmaps causing BUG
981 Details : The locking of the mballoc buddy bitmap and the in-memory
982 block bitmap was using two different spin locks in some
983 cases. This made it possible to incorrectly access the
984 mballoc bitmap while another process was modifying it,
985 causing a sanity assertion to fail. While no on-disk corruption
986 was reported, there was some risk of this happening.
989 Frequency : rare, on llog test 6
991 Descriptoin: don't allow connect to already connected import
992 Details : allowing connect to already connected import is hide connecting problem.
997 Descriptoin: allow kill process which wait statahead result
998 Details : for some reasons 'ls' can stick in waiting result from statahead,
999 in this case need way for kill this process.
1001 Severity : enhancement
1003 Description: Add state history info file, enhance import info file
1004 Details : Track import connection state changes in a new osc/mdc proc file;
1005 add overview-type data to the osc/mdc import proc file.
1007 Severity : enhancement
1009 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1011 Severity : enhancement
1013 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1018 Descriptoin: OST grant too much space to client even there are not enough space.
1019 Details : Client will shrink its grant cache to OST if there are no write
1020 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1021 this grant cache if there are already not enough avaible space
1022 (left_space < total_clients * 32M).
1025 Frequency : start MDS on uncleanly shutdowned MDS device
1027 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1028 Details : stay in waiting mds<>ost recovery finished produce random bugs
1029 due race between two ll_sync thread for one lov target. send
1030 ACTIVATE event only if connect realy finished and import have
1034 Frequency : rare, connect and disconnect target at same time
1036 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1037 Details : don't call obd_disconnect under lov_lock. this long time
1038 operation and can block ptlrpcd which answer to connect request.
1043 Descriptoin: don't lose wakeup for imp_recovery_waitq
1044 Details : recover_import_no_retry or invalidate_import and import_close can
1045 both sleep on imp_recovery_waitq, but we was send only one wakeup
1049 Frequency : always with long access acl
1051 Descriptoin: mds can't pack reply with long acl.
1052 Details : mds don't control size of acl but they limited by reint/getattr
1055 Severity : enhancement
1057 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1059 Severity : enhancement
1061 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1064 Frequency : start MDS on uncleanly shutdowned MDS device
1066 Descriptoin: aborting recovery hang on MDS
1067 Details : don't throttle destroy RPCs for the MDT.
1070 Frequency : on remount
1072 Description: external journal device not working after the remount
1073 Details : clear dev_rdonly flag for external journal devices in
1079 Description: shutdown vs evict race
1080 Details : client_disconnect_export vs connect request race.
1081 if client will evicted at this time - we start invalidate
1082 thread without referece to import and import can be freed
1086 Frequency : rare, need acl's on inode.
1088 Description: client can't handle ost additional correctly
1089 Details : if ost was added after client connected to mds client can have
1090 hit lnet_try_match_md ... to big messages to wide striped files.
1091 in this case need teach client to handle config events about add
1092 lov target and update client max ea size at that event.
1094 Severity : enhancement
1096 Description: Changelogs
1097 Details : Changelogs are a lightweight mechanism to track filesystem
1098 metadata and namespace changes. The changelog is recorded
1099 permanently on the MDTs, and is periodically "consumed" / purged
1100 when records are no longer needed.
1102 Severity : enhancement
1104 Description: compact fld format with extents
1105 Details : Store range of seq rather than every seq in FLD. Seq
1106 controller update FLD rather than clients. In Case of CMD, mdt0
1107 has FLD, all other metadata server act as non persistent proxy
1108 for FLD queries and cache fld entries in fld cache.
1113 Description: don't skip ost target if they assigned to file
1114 Details : Drop slow OSCs if we can, but not for requested start idx.
1115 This means "if OSC is slow and it is not the requested
1116 start OST, then it can be skipped, otherwise skip it only
1117 if it is inactive/recovering/out-of-space.
1121 Description: more cleanup in mds_lov
1122 Details : not send LOV EA under replay, we can't know about they size at this
1123 time. Don't allow client connect to mds before any ost connected,
1124 for avoid problems with LOV EA size and returning EIO to client.
1126 Severity : enhancement
1128 Description: Interoperability at server side (Disk interoperability)
1130 Severity : enhancement
1132 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1134 Severity : enhancement
1136 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1138 Severity : enhancement
1140 Description: New client IO stack (CLIO).
1142 Severity : enhancement
1144 Description: Commit on sharing. Eliminate inter-client dependencies between
1145 uncommitted transactions by doing transaction commits.
1146 Thereby clients may recovery independently.
1149 Frequency : Create a symlink file with a very long name
1151 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1152 Details : If there is no extra space in the request for early cancels,
1153 ldlm_req_handles_avail() returns 0 instead of a negative value.
1155 Severity : enhancement
1157 Description: Add /proc entry for import status
1158 Details : The mdc, osc, and mgc import directories now have
1159 an import directory that contains useful import data for debugging
1160 connection problems.
1162 Severity : enhancement
1164 Description: Re-disable certain /proc logging
1165 Details : Enable and disable client's offset_stats, extents_stats and
1166 extents_stats_per_process stats logging on the fly.
1169 Frequency : Only on FC kernels 2.6.22+
1171 Description: oops in statahead
1172 Details : Do not drop reference count for the dentry from VFS when lookup,
1173 VFS will do that by itself.
1175 Severity : enhancement
1177 Description: Generic /proc file permissions
1178 Details : Set /Proc file permissions in a more generic way to enable non-
1179 root users operate on some /proc files.
1183 Description: Hitting mdc_commit_close() ASSERTION
1184 Details : Properly handle request reference release in
1185 ll_release_openhandle().
1189 Description: quota recovery deadlock during mds failover
1190 Details : This patch includes att18982, att18236, att18237 in bz14840.
1192 1. fix osts hang when mds does failover with quotaon
1193 2. prevent watchdog storm when osts threads wait for the
1198 Frequency : only patchless client
1199 Description: add workaround for race between add/remove dentry from hash
1201 Severity : enhancement
1203 Description: Allow OST glimpses to return PW locks
1207 Description: LBUG when llog conf file is full
1208 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1212 Description: Prevent import from entering FULL state when server in recovery
1216 Description: service mount cannot take device name with ":"
1217 Details : Only when device name contains ":/" will mount treat it as
1223 Description: replace ptlrpcd with the statahead thread to interpret the async
1224 statahead RPC callback
1228 Frequency : on recovery
1229 Description: I/O failures after umount during fail back
1230 Details : if client reconnected to restarted server we need join to recovery
1231 instead of find server handler is changed and process self
1232 eviction with cancel all locks.
1234 Severity : enhancement
1236 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1238 Severity : enhancement
1240 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1242 Severity : enhancement
1244 Description: Upcall on Lustre log has been dumped
1245 Details : Allow for a user mode script to be called once a Lustre log has
1246 been dumped. It passes the filename of the dumped log to the
1247 script, the location of the script can be specified via
1248 /proc/sys/lnet/debug_log_upcall.
1253 Description: avoid idr_remove called for id which is not allocated.
1254 Details : Move assigment s_dev for clustered nfs to end of initialization,
1255 to avoid problem with error handling.
1260 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1261 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1262 into CONN_USED_HASH and this prodice warning when put connection
1263 again in unused hash.
1266 Severity : enhancement
1268 Description: File striping can now be set to use an arbitrary pool of OSTs.
1270 Severity : enhancement
1272 Description: Export bytes_read/bytes_write count on OSC/OST.
1276 Description: Early reply size mismatch, MGC loses connection
1277 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1278 the connect flags are properly negotiated.
1283 Description: quotas are not honored with O_DIRECT
1284 Details : all writes with the flag O_DIRECT will use grants which leads to
1285 this problem. Now using OBD_BRW_SYNC to guard this.
1289 Description: add quota statistics
1290 Details : 1. sort out quota proc entries and proc code.
1291 2. add quota statistics
1293 Severity : enhancement
1295 Description: enable quota support for HEAD.
1299 Description: Properly propagate oinfo flags from lov to osc for statfs
1300 Details : restore missing copy oi_flags to lov requests.
1302 Severity : enhancement
1304 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1305 : l_getgroups(8), lst(8), routerstat(8)
1306 Details : included man pages for llobdstat(8), llstat(8),
1307 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1309 Severity : enhancement
1311 Description: configure's --enable-quota should check the
1312 : kernel .config for CONFIG_QUOTA
1313 Details : configure is terminated if --enable-quota is passed but
1314 : no quota support is in kernel
1318 Description: Remove portals compatibility
1319 Details : Remove portals compatibility, not interoperable with releases
1324 Description: Resolve device initialization race
1325 Details : Prevent proc handler from accessing devices added to the
1326 obd_devs array but yet be intialized.
1328 Severity : enhancement
1330 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1332 Severity : enhancement
1334 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1339 Description: Using wrong pointer in osc_brw_prep_request
1340 Details : Access to array[-1] can produce panic if kernel compiled with
1341 CONFIG_PAGE_ALLOC enabled
1345 Description: Client runs out of low memory
1346 Details : Consider only lowmem when counting initial number of llap pages
1350 Description: *optional* service tags registration
1351 Details : if the "service tags" package is installed on a Lustre node
1352 When the filesystem is mounted, a local-node service tag will
1353 be created. See http://inventory.sun.com/ for more information
1354 about the Service Tags asset management system.
1358 Description: Kernel BUG tries to release flock
1359 Details : Lustre does not destroy flock lock before last reference goes
1360 away. So always drop flock locks when client is evicted and
1361 perform unlock regardless of successfulness of speaking to MDS.
1365 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1369 Description: sanity test 65a fails if stripecount of -1 is set
1370 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1375 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1376 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1379 Severity : enhancement
1381 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1385 Description: do not process already freed flock
1386 Details : flock can possibly be freed by another thread before it reaches
1387 to ldlm_flock_completion_ast.
1391 Description: LBUG during stress test
1392 Details : Need properly lock accesses the flock deadlock detection list.
1396 Description: oops in page fault handler
1397 Details : kernel page fault handler can return two special 'pages' in error
1398 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1402 Description: timeout with invalidate import.
1403 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1404 should be handled by ptlrpcd. This produce long age waiting and
1405 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1407 Severity : enhancement
1409 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1413 Description: enable MGS and MDT services start separately
1414 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1415 a co-located MGS without starting the MGS, which is a complement
1416 to 'nosvc' mount option.
1420 Frequency : after recovery
1421 Description: precreate to many object's after del orphan.
1422 Details : del orphan st in oscc last_id == next_id and this triger growing
1423 count of precreated objects. Set flag LOW to skip increase count
1424 of precreated objects.
1428 Frequency : rare, on clear nid stats
1429 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1430 Details : when clean nid stats sometimes try destroy live entry,
1431 and this produce panic in free.
1435 Description: Stack overflow during MDS log replay
1436 ease stack pressure by using a thread dealing llog_process.
1440 Description: wait until IO finished before start new when do lock cancel.
1441 Details : VM protocol want old IO finished before start new, in this case
1442 need wait until PG_writeback is cleared until check dirty flag and
1443 call writepages in lock cancel callback.
1445 Severity : enhancement
1447 Description: using special macro for print time and cleanup in includes.
1451 Description: mds_mfd_close() ASSERTION(rc == 0)
1452 Details : In mds_mfd_close(), we need protect inode's writecount change
1453 within its orphan write semaphore to prevent possible races.
1457 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1461 Frequency : rare, on shutdown ost
1462 Description: don't hit live lock with umount ost.
1463 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1464 use shrink_dcache_sb instead.
1468 Description: don't panic with use echo client
1469 Details : echo client pass NULL as client nid pointer and this produce null
1470 pointer dereference.
1474 Description: fix build on ppc32
1475 Details : compile code with -m64 flag produce wrong object file for ppc32.
1479 Description: add message levels for liblustreapi
1483 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1484 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1485 will result in an -EAGAIN and in the exit call path we call
1486 alloc_rr() which will with increasing aggressiveness attempt to
1487 aquire precreated objects on the minimum number of required OSCs.
1491 Description: Use old size assignment to avoid deadlock
1492 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1493 the scheduling while holding a spinlock. We do not need locking
1494 for size in ll_update_inode() because size is only updated from
1495 the MDS for directories or files without objects, so there is no
1496 other place to do the update, and concurrent access to such inodes
1497 are protected by the inode lock.
1501 Description: resolve "_IOWR redefined" build error on SLES10
1505 Description: dump the memory debugging after all modules are unloaded to
1506 suppress false negative in conf_sanity test 39
1508 Severity : enhancement
1510 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1515 Description: fix SLES kernel versioning
1516 Details : the kernel version for our SLES 10 kernel did not include a "-"
1517 before the "smp" at the end. while this was not a problem in
1518 general, it did mean that software trying to use the kernel
1519 version to try to detect a vendor specific kernel would fail.
1520 this was most evident by the OFED build scripts.
1524 Description: Don't update lov_desc members until making sure they are valid
1525 Details : When updating lov_desc members via proc fs, need fix their
1526 validities before doing the real update.
1530 Description: don't put request into delay list while invalidate in flight.
1531 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1532 import in flight. this produce timeout for invalidate and sometimes
1533 can cause stale data.
1537 Frequency : on ppc only
1538 Description: not convert ost objects for directory because it's not exist.
1539 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1543 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1545 Severity : enhancement
1547 Description: Update to support 2.6.22.14 vanilla kernel.
1551 Frequency : rare, on recovery
1552 Description: read procfs can produce deadlock in some situation
1553 Details : Holding lprocfs lock which send rpc can produce block for destroy
1554 obd objects and this also block reconnect with -EALREADY. This
1555 isn't fix all lprocfs bugs - but make it rare.
1557 Severity : enhancement
1559 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1562 Frequency : frequent on X2 node
1564 Description: mdc_set_open_replay_data LBUG
1565 Details : Set replay data for requests that are eligible for replay.
1569 Description: lustre_mgs: operation 101 on unconnected MGS
1570 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1571 MGC, and late on MGC cannot successfully connect to MGS and a lot
1572 of the error messages complaining that MGS is not connected.
1576 Frequency : on network error
1577 Description: panic with double free request if network error
1578 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1579 but it's true only for synchronus enqueue, for async enqueue
1580 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1583 Severity : enhancement
1585 Description: client-side metadata stat-ahead during readdir(directory readahead)
1586 Details : perform client-side metadata stat-ahead when the client detects
1587 readdir and sequential stat of dir entries therein
1590 Frequency : on start mds
1592 Description: Implement get_info(last_id) in obdfilter.
1595 Frequency : occasional
1597 Description: Correctly check stale fid, not start epoch if ost not support SOM
1598 Details : open with flag O_CREATE need set old fid in op_fid3 because
1599 op_fid2 was overwritten with new generated fid, but mds can answer
1600 with one of these two fids and both is not stale. Setattr
1601 incorrectly started an epoch and assume will be called
1602 done_writing, but without SOM done_writing ever being called.
1605 Frequency : rare, depends on device drivers and load
1607 Description: MDS or OSS nodes crash due to stack overflow
1608 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1609 In some cases, in conjunction with device drivers that use a lot
1610 of stack the MDS (or possibly OSS) service threads could overflow
1611 the stack. One change which was identified to consume additional
1612 stack has been reworked to avoid the extra stack usage.
1615 Frequency : occasional
1617 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1618 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1619 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1620 oscc_has_objects and see OSC has no objests and call
1621 oscc_internal_create to resend create request.
1623 Severity : enhancement
1625 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1627 Severity : enhancement
1629 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1632 Frequency : very rare
1634 Description: Fix replay if there is an un-replied request and open
1635 Details : In some cases, older replay request will revert the
1636 mcd->mcd_last_xid on MDS which is used to record the client's
1637 latest sent request.
1639 Severity : enhancement
1641 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1643 Severity : enhancement
1645 Description: Add rhel5 support to HEAD.
1647 Serverity : enhancement
1649 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1654 Description: Don't allow skipping OSTs if index has been specified.
1655 Details : Don't allow skipping OSTs if index has been specified, make
1656 locking in internal create lots better.
1660 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1661 Details : access to bitfield in structure is always rounded to long
1662 and this produce problem with not atomic change any bit.
1666 Description: Lustre make rpms failed.
1667 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1668 builds Lustre rpms from tarball.
1671 Frequency : rare on shutdown ost
1673 Description: If llog cancel was not send before clean_exports phase, this can
1674 produce deadlock in llog code.
1675 Details : If llog thread has last reference to obd and call class_import_put
1676 this produce deadlock because llog_cleanup_commit_master wait when
1677 last llog_commit_thread exited, but this never success because was
1678 called from llog_commit_thread.
1682 Description: allow userland application know is lost one of stripes.
1683 Details : fill lvb_blocks with error code on ost and return it to
1684 application if error flag found.
1688 Description: NULL lov_tgts causing MDS oops
1689 Details : more safe checks for NULL lov_tgts for avoid oops.
1691 Severity : enhancement
1693 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1697 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1698 Details : If an OST is down the MDS will hang indefinitely in
1699 obd_statfs_rqset() waiting for the statfs data. While for
1700 MDS QOS usage of statfs, it should not stuck in waiting.
1702 Severity : enhancement
1704 Description: remote_acl support
1705 Details : Support ACL-based permission check for remote user.
1706 Support setfacl/getfacl for remote user with the utils
1707 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1708 the system "{s,g}etfacl" utils.
1710 Severity : enhancement
1712 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1714 Severity : enhancement
1716 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1720 Description: make mgs_setparam() handle fsname containing dash
1721 Details : fsname containing a dash does not work with lctl conf_param
1723 Severity : enhancement
1725 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1727 Severity : enhancement
1729 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1733 Frequency : rare, at shutdown
1734 Description: access already free / zero obd_namespace.
1735 Details : if client_disconnect_export was called without force flag set,
1736 and exist connect request in flight, this can produce access to
1737 NULL pointer (or already free pointer) when connect_interpret
1738 store ocd flags in obd_namespace.
1742 Frequency : only at startup
1743 Description: not alloc memory with spinlock held.
1744 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1745 if any spinlock held.
1747 Severity : enhancement
1749 Description: make lustre randomly fail allocating memory
1750 Details : Make lustre randomly failed allocating memory for testing purpose.
1752 Severity : enhancement
1754 Description: lost problems with lov objid file
1755 Details : Fixes some scability and access to not inited memory problems
1756 in work with lov objdid file.
1761 Description: lfs find does not continue on file error
1762 Details : Continue other files processing when a file/dir is absent.
1766 Description: Inconsistent usage of lustre_pack_reply()
1767 Details : Standardize the usage of lustre_pack_reply() such that it
1768 always generate a CERROR on failure.
1771 Frequency : occasional
1773 Description: MDS hang or stay in waiting lock
1774 Details : If client receive lock with CBPENDING flag ldlm need send lock
1775 cancel as separate rpc, to avoid situation when cancel request
1776 can't processed due all i/o threads stay in wait lock.
1780 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1784 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1788 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1792 Description: Update to RHEL4 latest kernel.
1794 Severity : enhancement
1796 Description: Build SLES10 patchless client fails
1797 Details : The configure was broken by run ./configure with
1798 --with-linux-obj=.... argument for patchless client. When the
1799 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1800 can't search header adequately. Use absolute path such as
1801 -I($LINUX)/include instead.
1805 Description: interrupt oig_wait produce painc on resend.
1806 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1807 private set, and this produce situation when rq_ptlrpcd_data not
1808 copyed to new allocated request and triggered LBUG on assert
1809 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1810 ptlrpcd set if request is changed and can be safety changed to use
1815 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1816 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1817 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1818 these macros and replace them with their difinitions instead.
1822 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1823 Details : Update which_patch & target file for SLES10 latest kernel.
1826 Description: Remove client patches from SLES 10 kernel.
1827 Details : This causes SLES 10 clients to behave as patchless clients
1828 even on a Lustre-patched (server) kernel.
1830 Severity : enhancement
1832 Description: self-adjustable client's lru lists
1833 Details : use adaptive algorithm for managing client cached locks lru
1834 lists according to current server load, other client's work
1835 pattern, memory activities, etc. Both, server and client
1836 side namespaces provide number of proc tunables for controlling
1839 Severity : enhancement
1841 Description: light-weight GSS support
1842 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1843 and reduce performance overhead.
1845 Severity : enhancement
1847 Description: Linux keyring support
1848 Details : Support using service of Linux keyring for Lustre GSS internal
1849 context refresh/cache mechanism.
1853 Description: Fix errors in lfs documentation
1854 Details : Fixes man pages
1858 Description: don't use GFP_* in generic Lustre code.
1859 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1863 Description: obdclass is limited by single OBD_ALLOC(idarray)
1864 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1868 Description: warnings with build patchless client with vanila 2.6.19 and up
1869 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1870 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1874 Frequency : Only for SLES
1875 Description: sanity_quota fail test_1
1876 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1877 file, which makes TSTID[2] inunique.
1881 Frequency : Only for SLES9
1882 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1883 Details : Update target/ChangeLog/which_patch .
1887 Frequency : Only for test_10 in sanity-quota.sh
1888 Description: a bug in quota test code
1889 Details : checking a test flag in a wrong place
1893 Frequency : Only for test_13 in sanity-quota.sh
1894 Description: a bug in quota test script
1895 Details : shouldn't check overfull allocated quota
1899 Description: jbd statistics
1900 Details : Port older jbd statistics patch for sles10
1902 Severity : enhancement
1904 Description: Add printing inode info into message about error in writepage.
1909 Description: reply_lock_interpret crash due to race with it and lock cancel.
1910 Details : Do not replay locks that are being cancelled. Do not reference
1911 locks by their address during replay, just by their handle.
1915 Frequency : When flocks are used.
1916 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1917 Details : Flock locks might destroy just granted lock if it could be merged
1918 with another existing flock, this is done in completion handler,
1919 so teach ldlm_cli_enquque_fini that this is a valid case for
1925 Description: Oops in read and write path when failing to allocate lock.
1926 Details : Check if lock allocation failed and return error back.
1930 Description: lstripe command fails for valid OST index
1931 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1932 instead of lov->desc.ld_active_tgt_count.
1936 Description: sanity.sh failed test 103
1937 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1942 Description: df doesn't work properly if diskfs blocksize != 4K
1943 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1947 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1948 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1950 Severity : enhancement
1952 Description: Adaptive timeouts
1953 Details : RPC timeouts adapt to changing server load and network
1954 conditions to reduce resend attempts and improve recovery time.
1958 Description: llapi_file_create() does not allow some changes
1959 Details : add llapi_file_open() that allows specifying the mode and
1960 open flags, and also returns an open file handle.
1964 Description: merge and cleanup kernel patches.
1965 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1969 Description: Add journal checksum support.(Kernel part)
1970 Details : The journal checksum feature adds two new flags i.e
1971 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1972 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1973 indicates that the commit block contains the checksum for
1974 the blocks described by the descriptor blocks. Now commit
1975 record can be sent to disk without waiting for descriptor
1976 blocks to be written to disk. This behavior is controlled
1977 using JBD2_FEATURE_ASYNC_COMMIT flag.
1981 Description: OSS needs mutliple precreate threads
1982 Details : Add ability to start more than one create thread per OSS.
1986 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1987 Details : execute lfs setstripe on client
1991 Description: mds_obd_create error creating tmp object
1992 Details : When the user sets quota on root, llog will be affected and can't
1993 create files and write files.
1996 Frequency : Always on ia64 patchless client, and possibly others.
1998 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1999 Details : This allows the patchless client to be loaded on architectures
2000 without this export.
2004 Description: RedHat Update kernel for RHEL5
2005 Details : Add the kernel config file for RHEL5.
2009 Description: RedHat Update kernel for RHEL5
2010 Details : Modify the kernel config file more closer RHEL5.
2014 Description: Build failure against Centos5 (RHEL5)
2015 Details : Define PAGE_SIZE when it isn't present.
2019 Description: test_7 on sanity-quota.sh can't be used on separate machines
2020 Details : add facet to handle it
2024 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2025 Details : with NFS, the anon dentry's parent was set to itself in
2026 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2027 corresponding dentry other than use rec->ur_name.
2029 Severity : enhancement
2031 Description: lfs setstripe enhancement
2032 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2036 Description: enable data checksumming by default
2037 Details : enable checksum by default, allow --disable-checksum
2038 configure option and "-o nochecksum" mount option. Checksums
2039 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2043 Description: lustre support for RHEL5
2044 Details : Add support for RHEL5.
2048 Description: Client eviction due to failover config
2049 Details : after a connection loss, the lustre client should attempt to
2050 reconnect to the last active server first before trying the
2051 other potential connections.
2055 Description: when mds and osts use different quota unit(32bit and 64bit),
2056 quota will be released repeatly.
2057 Details : void sending multiple quota reqs to mds, which will keep the
2058 status between the reqs.
2062 Description: osts not allocated evenly to files
2063 Details : change the condition to increase offset_idx
2066 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2068 Description: Data corruption for OSTs that are formatted larger than 2TB
2070 Details : When generating the bio request for lustre file writes the
2071 sector number would overflow a temporary variable before being
2072 used for the IO. The data reads correctly from Lustre (which
2073 will overflow in a similar manner) but other file data or
2074 filesystem metadata may be corrupted in some cases.
2078 Description: Tune the kernel for good SCSI performance.
2079 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2080 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2085 Description: rewrite ext2-derived code in obdclass/uuid.c
2086 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2087 from scratch preserving functionality.
2091 Description: "lfs find -obd UUID" prints directories
2092 Details : "lfs find -obd UUID" will return all directory names instead
2093 of just file names. It is incorrect because the directories
2094 do not reside on the OSTs.
2098 Description: use i_size_read and i_size_write in 2.6 port
2099 Details : replace inode->i_size access with i_size_read/write()
2102 Frequency : when removing large files
2104 Description: scheduling issue during removal of large Lustre files
2105 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2106 It causes scheduling issues when removing large files (17TB in the
2110 Frequency : only with liblustre clients on XT3
2112 Description: evictions taking too long
2113 Details : allow llrd to evict clients directly on OSTs
2116 Frequency : only on ppc
2118 Description: /proc/fs/lustre/devices broken on ppc
2119 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2120 We should make sure the type of 'index' is loff_t and then cast
2121 explicitly as needed below. Do not assign an explicitly cast
2125 Frequency : only for rhel5
2127 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2128 Details : Modify the target file & which_kernel.
2130 Severity : enhancement
2132 Description: omit set fsid for export NFS
2133 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2134 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2138 Frequency : Always, for kernels after 2.6.16
2139 Description: Fix warning idr_remove called for id=.. which is not allocated.
2140 Details : Last kernels save old s_dev before kill super and not allow
2141 to restore from callback - restore it before call kill_anon_super.
2145 Description: buffer overruns could theoretically occur
2146 Details : llapi_semantic_traverse() modifies the "path" argument by
2147 appending values to the end of the origin string, and a
2148 overrun may occur. Adding buffer overrun check in liblustreapi.
2152 Description: Fix error on 'ls .' at the top of the Lustre mount.
2153 Details : Don't revalidate dentry if it is a root dentry.
2157 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2158 Details : Modify vm-tunables-rhel4.patch.
2162 Description: Kernel config for 2.6.18-vanilla.
2163 Details : Modify targets/2.6-vanilla.target.in.
2164 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2165 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2166 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2167 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2171 Description: improve handling recoverable errors
2172 Details : if request processig with error which can be recoverable on server
2173 request should be resend, otherwise page released from cache and
2178 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2179 Details : Modify target file & which_patch.
2180 A flaw was found in the IA32 system call emulation provided
2181 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2182 value could be stored in the %RAX register, which could trigger an
2183 out-of-bounds system call table access. An untrusted local user
2184 could exploit this flaw to run code in the kernel
2185 (ie a root privilege escalation). (CVE-2007-4573).
2189 Description: O_DIRECT bypasses client statistics.
2190 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2191 read_ahead_stats were updated. copied the stats section from
2192 osc_send_oap_rpc() into async_internal().
2196 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2200 Description: change order of libsysio includes
2201 Details : '#include sysio.h' should always come before '#include xtio.h'
2205 Description: handle "serious error: objid * already exists" more gracefully
2206 Details : If LAST_ID value on disk is smaller than the objects existing in
2207 the O/0/d* directories, it indicates disk corruption and causes an
2208 LBUG(). If the object is 0-length, then we should use the existing
2209 object. This will help to avoid a full fsck in most cases.
2211 Severity : enhancement
2213 Description: adapt the lustre_config script to support the upgrade case
2214 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2215 server targets to 1.6 in parallel.
2219 Description: To avoid grant space > avaible space when the disk is almost
2220 full. Without this patch you might see the error "grant XXXX >
2221 available" or some LBUG about grant, when the disk is almost
2223 Details : In filter_check_grant, for non_grant cache write, we should
2224 check the left space by if (*left > ungranted + bytes), instead
2225 of (*left > ungranted), because only we are sure the left space
2226 is enough for another "bytes", then the ungrant space should be
2227 increase. In client, we should update cl_avail_grant only there
2228 is OBD_MD_FLGRANT in the reply.
2232 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2233 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2234 described in CVE-2007-4573. This problem could allow untrusted
2235 local users to gain root access.
2238 Frequency : when using O_DIRECT and quotas
2240 Description: Incorrect file ownership on O_DIRECT output files
2241 Details : block usage reported by 'lfs quota' does not take into account
2242 files that have been written with O_DIRECT.
2247 Description: touch file failed when fs is not full
2248 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2249 otherwise we can get ENOSP while fs is not full.
2253 Description: parallel lock callbacks
2254 Details : Instead of sending blocking and completion callbacks as separated
2255 requests, adding them to a set and sending in parallel.
2258 Frequency : only for Cray XT3
2259 Bugzilla : 12829/13455
2260 Description: Changing primary group doesn't change the group lustre assigns to
2262 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2263 overridden with the primary group provided by the group upcall,
2264 whereas the supplied fsgid can be trusted if it is in the list of
2265 supplementary groups returned by the group upcall.
2267 Severity : enhancement
2269 Description: Allow masking D_WARNING, D_ERROR messages from console
2270 Details : Console messages can now be disabled via lnet.printk.
2274 Description: User code with malformed file open parameter crashes client node
2275 Details : Before packing join_file req, all the related reference should be
2276 checked carefully in case some malformed flags cause fake
2277 join_file req on client.
2281 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2282 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2283 mode is equal to the granted mode. However, it should also check
2284 that LDLM_FL_CP_REQD is not set.
2288 Description: Heavy nfs access might result in deadlocks
2289 Details : After ELC code landed, it is now improper to enqueue any mds
2290 locks under och_sem, because enqueue might want to decide to
2291 cancel open locks for same inode we are holding och_sem for.
2295 Description: Client eviction while running blogbench
2296 Details : A lot of unlink operations with concurrent I/O can lead to a
2297 deadlock causing evictions. To address the problem, the number of
2298 oustanding OST_DESTROY requests is now throttled to
2299 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2300 callbacks are processed in priority.
2304 Description: enable ACLs on MDS by default
2305 Details : ACLs must be enabled on MDS by default.
2308 Frequency : PPC/PPC64 only
2310 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2311 Details : fix duplicated definitions between asm-ppc64/types.h and
2312 lustre_types.h on PPC.
2315 Frequency : PPC/PPC64 only
2317 Description: asm-ppc/segment.h does not exist
2318 Details : fix compile issue on PPC.
2322 Description: better handle error messages in extents code
2325 Frequency : RHEL4 only
2327 Description: mkfs is very slow on IA64/RHEL4
2328 Details : A performance regression has been discovered in the MPT Fusion
2329 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2330 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2331 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2333 Severity : enhancement
2335 Description: SNMP support enhancement
2336 Details : Adding total number of sampled request for an MDS node in snmp
2339 Severity : enhancement
2341 Description: Optimize ldlm waiting list processing for PR extent locks
2342 Details : When processing waiting list for read extent lock and meeting read
2343 lock that is same or wider to it that is not contended, skip
2344 processing rest of the list and immediatelly return current
2345 status of conflictness, since we are guaranteed there are no
2346 conflicting locks in the rest of the list.
2350 Description: Time out and refuse to reconnect
2351 Details : When the failover node is the primary node, it is possible
2352 to have two identical connections in imp_conn_list. We must
2353 compare not conn's pointers but NIDs, otherwise we can defeat
2354 connection throttling.
2358 Description: port llog fixes from b1_6 into HEAD
2359 Details : Port llog reference couting and some llog cleanups from b1_6
2360 (bug 10800) into HEAD, for protect from panic and access to already
2361 free llog structures.
2365 Description: Detect stride IO mode in read-ahead
2366 Details : When a client does stride read, read-ahead should detect that and
2367 read-ahead pages according to the detected stride pattern.
2371 Description: data checksumming impacts single node performance
2372 Details : add support for several checksum algorithm. Currently, only CRC32
2373 and Adler-32 are supported. The checksum type can be changed on
2374 the fly via /proc/fs/lustre/osc/*/checksum_type.
2378 Description: use adler32 for page checksums
2379 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2384 Description: build for x2 fails
2385 Details : fix compile issue on Cray systems.
2389 Description: Properly match for duplicate locks
2390 Details : Due to different lock order from skiplists code, we need to
2391 traverse entire list for now
2394 Frequency : only on PPC/SLES10
2396 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2397 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2398 be defined. Add a hack in mkfs_lustre.c to work around this compile
2403 Description: LASSERT on MDS when client holding flock lock dies
2404 Details : ldlm pool logic depends on number of granted locks equal to
2405 number of released locks which is not true for flock locks, so
2406 just exclude such locks from consideration.
2410 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2411 Details : Use fsfilt_sync() for both the whole filesystem sync and
2412 individual file sync to eliminate dangerous inode locking
2413 with I_LOCK that can lead to a deadlock.
2417 Description: performance in 1.6.3
2418 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2419 sure that 1MB requests are not fragmented by the block layer.
2421 Severity : enhancement
2423 Description: organize the server-side client stats on per-nid basis
2424 Details : Change the structure of stats under obdfilter and mds to
2432 The "uuid"s file would list the uuids of _active_ exports.
2433 And the clear entry is to clear all stats and stale nids.
2435 Severity : enhancement
2437 Description: eliminate client locks in face of contention
2438 Details : file contention detection and lockless i/o implementation
2439 for contended files.
2443 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2446 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2448 Description: MDS may lose file striping (and hence file data) in some cases
2449 Details : If there are additional extended attributes stored on the MDS,
2450 in particular ACLs, SELinux, or user attributes (if user_xattr
2451 is specified for the client mount options) then there is a risk
2452 of attribute loss. Additionally, the Lustre file striping
2453 needs to be larger than default (e.g. striped over all OSTs),
2454 and an additional attribute must be stored initially in the
2455 inode and then increase in size enough to be moved to the
2456 external attribute block (e.g. ACL growing in size) for file
2461 Description: skiplist implementation simplification
2462 Details : skiplists are used to group compatible locks on granted list
2463 that was implemented as tracking first and last lock of each
2464 lock group the patch changes that to using doubly linked lists
2468 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2469 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2470 because LASSERT on both the data supplied by a client, and the
2471 data on disk is dangerous and incorrect.
2473 Severity : enhancement
2475 Description: Slow truncate/writes to huge files at high offsets.
2476 Details : Directly associate cached pages to lock that protect those pages,
2477 this allows us to quickly find what pages to write and remove
2478 once lock callback is received.
2482 Description: more ldlm soft lockups
2483 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2484 starve other threads from the resource lock for a long time in
2485 case of long waiting queue, so change the debug level from
2486 D_OTHER to the less frequently used D_INFO.
2488 Severity : enhancement
2490 Description: add -gid, -group, -uid, -user options to lfs find
2494 Description: Hung threads in invalidate_inode_pages2_range
2495 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2496 once one is completed. As a result, some RPCs are stuck in the
2497 queue and are never sent.
2501 Description: filter threads hungs on waiting journal commit
2502 Details : Cleanup filter group llog code, then only filter group llog will
2503 be only created in the MDS/OST syncing process.
2507 Description: Procfs and llog threads access destoryed import sometimes.
2508 Details : Sync the import destoryed process with procfs and llog threads by
2509 the import refcount and semaphore.
2511 Severity : enhancement
2513 Description: openlock cache of b1_6 port to HEAD
2518 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2519 Details : If insertion of an extent fails, then discard the inode
2520 preallocation and free data blocks else it can lead to duplicate
2525 Description: don't always update ctime in ext3_xattr_set_handle()
2526 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2527 In some cases the ctime should not be updated, for example for
2528 2.0->1.8 compatibility it is necessary to delete an xattr and it
2529 should not update the ctime.
2533 Bugzilla : 15713/16362
2534 Description: Assertion in iopen_connect_dentry in 1.6.3
2535 Details : looking up an inode via iopen with the wrong generation number can
2536 populate the dcache with a disconneced dentry while the inode
2537 number is in the process of being reallocated. This causes an
2538 assertion failure in iopen since the inode's dentry list contains
2539 both a connected and disconnected dentry.
2543 Description: assertion failure in ldlm_handle2lock()
2544 Details : fix a race between class_handle_unhash() and class_handle2object()
2545 introduced in lustre 1.6.5 by bug 13622.
2550 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2551 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2556 Description: oops in sd_iostats_seq_show()
2557 Details : unloading/reloading the scsi low level driver triggers a kernel
2558 bug when trying to access the sd iostat file.
2563 Description: Kernel panics during QLogic driver reload
2564 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2565 patch, causing memory corruption.
2570 Description: journal_dev option does not work in b1_6
2571 Details : pass mount option during pre-mount.
2573 Severity : enhancement
2575 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2576 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2577 extent information of a file. It can be used to map logical blocks
2578 in a file to physical blocks in the block device.
2582 Description: LDLM soft lockups - improvement
2583 Details : It is be possible to send the lock handle along with each read
2584 or write request because the client is already doing a lock match
2585 itself so there isn't any reason the OST should have to re-do that
2589 Frequency : only on Cray X2
2591 Description: X2 build failures
2592 Details : fix build failures on Cray X2.
2596 Description: xid & resent requests
2597 Details : Initialize RPC XID from clock at startup (randomly if clock is
2600 Severity : enhancement
2602 Description: Add lustre_start utility to start or stop multiple Lustre servers
2607 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2608 Details : In case of memory pressure, list_del() can be called twice on
2609 req->rq_history_list, causing a kernel oops.
2613 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2614 Details : fix stack overflow in the distributed lock manager by defering
2615 export eviction after a failed AST to the elt thread instead of
2616 handling it in the dlm interpret routine.
2620 Description: Convert some comments to new format.
2621 Details : Update documenting comments to match doxygen conventions.
2625 Description: Grammar fixes.
2626 Details : A couple of trivial spelling fixes.
2630 Description: OSD_COUNTERS-mandatory
2631 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2632 to simplify the code.
2636 Description: simplify cmm_device freeing logic.
2637 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2641 Description: Add lockdep support to dt_object_operations locking interface.
2642 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2643 parameter indicating lock ordering. Update mdd code to use new
2648 Description: Introduce failloc constants for lockless IO tests.
2649 Details : Add two new failloc constants to test lockless IO. Only one of
2650 them in implemented---another is checked in yet to be landed
2655 Description: Add lockdep support for inode mutex.
2656 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2657 in the situations where Lustre has to lock more than one inode
2662 Description: Add optional invariants checking support.
2663 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2664 configure switch. This macro is to be used for consistency and
2665 sanity checks that are too expensive to be left in `production'
2670 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2671 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2676 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2677 Details : Use spin_lock_nested() in (the only) situation where more than
2678 one ldlm_lock is locked simultaneously. Also, fix possible
2679 dead-lock in ldlm_lock_change_resource() by enforcing particular
2684 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2685 Details : Instead of specifying each ldlm_lock call-back through separate
2686 parameter, wrap them into struct ldlm_callback_suite.
2690 Description: Kill join_lru obd method and its callers.
2691 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2692 in memory---a requirement implemented through ->o_join_lru() obd
2693 method in HEAD. Remove this method and its users.
2697 Description: Add asynchronous ldlm ENQUEUE completion handler.
2698 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2699 case that stock ldlm_completion_ast() cannot handle as it waits
2700 until lock is granted. Introduce new ldlm_completion_ast_async()
2701 for this. Also comment ldlm_completion_ast().
2705 Description: ldlm_error <-> errno conversion.
2706 Details : Add functions to map (rather arbitrary) between LDLM error codes
2707 and standard errno values. CLIO needs this to prevent LDLM specific
2708 constants from escaping ldlm and osc.
2712 Description: Kill unused ldlm_handle2lock_ns() function.
2713 Details : Kill unused ldlm_handle2lock_ns() function.
2717 Description: Add lu_ref support to ldlm_lock
2718 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2719 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2720 and ldlm_resource. LDLM interface has to be changed, because code that
2721 releases a reference on a lock, has to "know" what reference this is.
2722 In the most frequent case
2724 lock = ldlm_handle2lock(handle);
2726 LDLM_LOCK_PUT(lock);
2728 no changes are required. When any other reference (received _not_ from
2729 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2730 instead of LDLM_LOCK_PUT().
2732 Arguably, changes are pervasive, and interface requires some discipline
2733 for proper use. On the other hand, it was very instrumental in finding
2734 a few leaked lock references.
2738 Description: Add ldlm_lock_addref_try().
2739 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2740 attempts to addref a lock that might be being canceled
2745 Description: Add ldlm_weigh_callback().
2746 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2747 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2748 hard-coded `number of pages' logic.
2752 Description: Add lockdep annotations to llog code.
2753 Details : Use appropriately tagged _nested() locking calls in the places
2754 where llog takes more than one ->lgh_lock lock.
2758 Description: Add loi_kms_set().
2759 Details : Wrap kms updates into a helper function.
2763 Description: Constify instances of struct lsm_operations.
2764 Details : Constify instances of struct lsm_operations.
2768 Description: lu_conf support.
2769 Details : On a server, a file system object is uniquely identified
2770 by a fid, which is sufficient to locate and load all object
2771 state (inode). On a client, on the other hand, more data are
2772 necessary instantiate an object. Change lu_object_find() and
2773 friends to take additional `lu_conf' argument describing object.
2774 Typically this includes layout information.
2778 Description: lu_context fixes.
2779 Details : Introduce new lu_context functions that are needed on the client
2780 side, where some system threads (ptlrpcd) are shared by multiple
2781 modules, and so cannot be stopped during module shutdown.
2785 Description: Add start and stop methods to lu_device_type_operations.
2786 Details : Introduce two new methods in lu_device_type_operations, that are
2787 invoked when first instance of a given type is created and last one
2788 is destroyed respectively. This is need by CLIO.
2792 Description: Add lu_ref support to struct lu_device.
2793 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2794 track leaked references.
2798 Description: Introduce lu_kmem_descr.
2799 Details : lu_kmem_descr and its companion interface allow to create
2800 and destroy a number of kmem caches at once.
2804 Description: Fix lu_object finalization race.
2805 Details : Fix a race between lu_object_find() finding an object and its
2806 concurrent finalization. This race is (most likely) not possible
2807 on the server, but might happen on the client.
2811 Description: Introduce lu_ref interface.
2812 Details : lu_ref is a debugging module allowing to track references to
2813 a given object. It is quite cpu expensive, and has to be
2814 explicitly enabled with --enable-lu_ref. See usage description
2819 Description: Factor lu_site procfs stats into a separate function.
2820 Details : Separate lu_site stats printing code into a separate function
2821 that can be reused on a client.
2825 Description: Constify instances of struct {lu,dt,md}_device_operations.
2826 Details : Constify instances of struct {lu,dt,md}_device_operations.
2830 Description: Introduce struct md_site and move meta-data specific parts of
2831 struct lu_site here.
2832 Details : Move md-specific fields out of struct lu_site into special struct
2833 md_site, so that lu_site can be used on a client.
2837 Description: Kill mdd_lov_destroy().
2838 Details : Remove unused mdd code.
2842 Description: Add st_block checking to multistat.c.
2843 Details : Add st_block checking to multistat.c.
2847 Description: Add lu_ref support to struct obd_device.
2848 Details : Add lu_ref tracking to obd_device.
2852 Description: Kill obd_set_fail_loc().
2853 Details : Remove unused code.
2857 Description: Add special type for ptlrpc_request interpret functions.
2858 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2859 there. Actual usage will be in CLIO.
2863 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2864 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2865 doesn't work with lockdep.
2869 Description: Add rwv.c test program.
2870 Details : New testing program exercising readv(2) and writev(2) (Qian).
2874 Description: Add sendfile.c test program.
2875 Details : New testing program exercising sendfile(2) (Jay).
2879 Description: Ratelimit a message that can be very frequent.
2880 Details : Ratelimit a memory allocation failure message that can
2885 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2886 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2890 Description: Liblustre build fixes.
2891 Details : Liblustre build fixes.
2895 Description: libcfs: add cfs_{need,cond}_resched() interface.
2896 Details : libcfs: add cfs_{need,cond}_resched() definition and
2897 implementations for Linux, NT, and liblustre.
2899 Severity : enhancement
2901 Description: More exported tunables for mballoc
2902 Details : Add support for tunable preallocation window and new tunables for
2903 large/small requests
2907 Description: Detect corruption of block bitmap and checking for preallocations
2908 Details : Checks validity of on-disk block bitmap. Also it does better
2909 checking of number of applied preallocations. When corruption is
2910 found, it turns filesystem readonly to prevent further corruptions.
2914 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2915 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2916 window to 0. If it does hit cache miss, and read-pattern is still
2917 stride-io mode, does not reset the stride window, but also does not
2918 increase the stride window length in this case.
2922 Frequency : only for big-endian servers
2923 Description: Check if system is big-endian while mounting fs with extents feature
2924 Details : Mounting a filesystem with extents feature will fail on big-endian
2925 systems since ext3-based ldiskfs is not supported on big-endian
2926 systems. This can be over-riden with "bigendian_extents" mount option.
2928 Severity : enhancement
2930 Description: The root squash functionality
2931 Details : A security feature, which is to prevent users from being able
2932 to mount lustre on their desktop, run as root, and delete
2933 all of the files in the filesystem. The goal is accomplished by
2934 remapping user id (UID) and group id (GID) of the root user to
2935 a UID and GID specified by the system administartor via Lustre
2936 configuration management server (MGS). The functionality also
2937 allows to specify sets of clients for which the remapping does
2942 Description: Excessive recovery window
2943 Details : With AT enabled, the recovery window can be excessively long (6000+
2944 seconds). To address this problem, we no longer use
2945 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2946 timeout no longer depends on the service time, it is set to
2947 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2948 time via pb_service_time.
2952 Description: Watchdog triggered on MDS failover
2953 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2954 the OSTs know that the MDS "UUID" can be reused for the same export
2955 from a different NID, so we do not need to wait for the export to be
2959 Frequency : rare, only if using MMP with Linux RAID
2961 Description: MMP doesn't work with Linux RAID
2962 Details : While using HA for Lustre servers with Linux RAID, it is possible
2963 that MMP will not detect multiple mounts. To make this work we
2964 need to unplug the device queue in RAID when the MMP block is being
2965 written. Also while reading the MMP block, we should read it from
2966 disk and not the cached one.
2968 Severity : enhancement
2970 Description: open file using fid
2971 Details : A file can be opened using just its fid, like
2972 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2975 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2978 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2979 Details : add proper locking for creation of jbd_4k slab cache
2983 Description: MMP check in ext3_remount() fails without displaying any error
2984 Details : When multiple mount protection fails during remount, proper error
2987 Severity : enhancement
2989 Description: Allow stripe size to be up to 4G-64k
2990 Details : Fix math logic to allow large stripe sizes.
2994 Description: add check for >8TB ldiskfs filesystems
2995 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2996 Don't allow >8TB ldiskfs filesystems to be mounted without
2997 force_over_8tb mount option
3002 Descriptoin: OSS complains "too many blocks" and crashes
3003 Details : In filter_commitrw_write():
3004 1. lquota_chkquota();
3005 2. obdo_from_inode();
3006 3. lquota_pending_commit();
3007 "2." has changed uid and gid in this case, then caused
3008 lquota_chkquota() and lquota_pending_commit() to use
3009 different uid and gid. That is the root of the bug.
3013 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3014 Details : Detect and print out the 'stripe count = -1' case properly.
3018 Description: Out or order replies might be lost on replay
3019 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3020 request with tid smaller than one currently being added, add it
3021 to the start, not end of the list.
3023 --------------------------------------------------------------------------------
3025 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3027 * Support for kernels:
3028 2.6.5-7.283 (SLES 9),
3029 2.6.9-55.EL (RHEL 4),
3030 2.6.16.46-0.14 (SLES 10),
3031 2.6.18.8 vanilla (kernel.org)
3032 * Client support for unpatched kernels:
3033 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3034 2.6.16 - 2.6.22 vanilla (kernel.org)
3035 * Due to recently discovered recovery problems, we do not recommend
3036 using patchless RHEL 4 clients with this or any earlier release.
3037 * Recommended e2fsprogs version: 1.39.cfs8
3038 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3039 * Starting with this release, the ldiskfs backing filesystem required
3040 by Lustre is now in its own package, lustre-ldiskfs. This package
3041 should be installed. It is versioned separately from Lustre and
3042 may be released separately in future.
3046 Description: block reactivating mgc import until all deactivates complete
3047 Details : Fix race when failing back MDT/MGS to itself (testing)
3050 Frequency : at statup only
3052 Description: mds_lov_synchronize race leads to various problems
3053 Details : simultaneous MDT->OST connections at startup can cause the
3054 sync to abort, leaving the OSC in a bad state.
3056 Severity : enhancement
3058 Description: add optional extra BUILD_VERSION info
3059 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3060 to override the lustre version.
3063 Frequency : 2.6.18 servers only
3065 Description: ll_kern_mount() doesn't release the module reference
3066 Details : The ldiskfs module reference count never drops down to 0
3067 because ll_kern_mount() doesn't release the module reference.
3072 Description: server LBUG when using old ost_num_threads parameter
3073 Details : Accept the old ost_num_threads parameter but warn that it
3074 is deprecated, and fix an off-by-one error that caused an LBUG.
3079 Description: Transient SCSI error results in persistent IO issue
3080 Details : iobuf->dr_error is not reinitialized to 0 between two
3084 Frequency : sometimes when underlying device returns I/O errors
3086 Description: OSTs not going read-only during write failures
3087 Details : OSTs are not remounted read-only when the journal commit threads
3088 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3089 instead of the ext3 wrappers.
3093 Description: poor connect scaling with increasing client count
3094 Details : Don't run filter_grant_sanity_check for more than 100 exports
3095 to improve scaling for large numbers of clients.
3098 Frequency : SLES10 only
3100 Description: sanity-quota.sh quotacheck failed: rc = -22
3101 Details : Quotas cannot be enabled on SLES10.
3104 Frequency : liblustre clients only
3106 Description: getdirentries does not give error when run on compute nodes
3107 Details : getdirentries does not fail when the size specified as an argument
3108 is too small to contain at least one entry
3110 Severity : enhancement
3112 Description: Add LNET router traceability for debug purposes
3113 Details : If a checksum failure occurs with a router as part of the
3114 IO path, the NID of the last router that forwarded the bulk data
3115 is printed so it can be identified.
3120 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3121 Details : Due to a race condition, liblustre clients were occasionally
3122 evicted incorrectly.
3124 Severity : enhancement
3126 Description: lfs setstripe use optional parameters instead of postional
3129 Severity : enhancement
3131 Description: Nanosecond timestamp support for ldiskfs
3132 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3133 resolution timestamps. There is not yet support for this at
3134 the Lustre filesystem level.
3137 Frequency : during server recovery
3139 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3140 Details : request with rq_no_resend flag not awake l_wait_event if they get
3144 Frequency : nfs export on patchless client
3146 Description: connectathon hang when test nfs export over patchless client
3147 Details : Disconnected dentry cannot be found with lookup, so we do not need
3148 to unhash it or make it invalid
3151 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3152 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3153 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3158 Description: open req refcounting wrong on reconnect
3159 Details : If reconnect happened between getting open reply from server and
3160 call to mdc_set_replay_data in ll_file_open, we will schedule
3161 replay for unreferenced request that we are about to free.
3162 Subsequent close will crash in variety of ways.
3163 Check that request is still eligible for replay in
3164 mdc_set_replay_data().
3169 Description: disable writes to filesystem when reading health_check file
3170 Details : the default for reading the health_check proc file has changed
3171 to NOT do a journal transaction and write to disk, because this
3172 can cause reads of the /proc file to hang and block HA state
3173 checking on a healthy but otherwise heavily loaded system. It
3174 is possible to return to the previous behaviour during configure
3175 with --enable-health-write.
3177 Severity : enhancement
3179 Description: 64-bit inode version
3180 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3181 to the inode. This will be required for version-based recovery.
3186 Description: MDS fails to start if a duplicate client export is detected
3187 Details : in some rare cases it was possible for a client to connect to
3188 an MDS multiple times. Upon recovery the MDS would detect this
3189 and fail during startup. Handle this more gracefully.
3191 Severity : enhancement
3193 Description: Add -o localflock option to simulate old noflock
3195 Details : This will achieve local-only flock/fcntl locks
3201 Description: log_commit_thread vs filter_destroy race leads to crash
3202 Details : Take import reference before releasing llog record semaphore
3207 Description: Wrong request locking in request set processing
3208 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3209 imp_delayed_list, in this place should be used imp_lock.
3212 Frequency : when reconnection
3214 Description: Grant Leak when osc reconnect to OST
3215 Details : When osc reconnect ost, OST(filter) should check whether it
3216 should grant more space to client by comparing fed_grant and
3217 cl_avail_grant, and return the granted space to client instead
3218 of "new granted" space, because client will call osc_init_grant
3219 to update the client grant space info.
3222 Frequency : when client reconnect to OST
3224 Description: Grant Leak when osc do resend and replay bulk write
3225 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3226 bulk write request, because the grant info while be sync between
3227 OSC and OST when reconnect, and we should ignore the grant info
3228 these of resend/replay write req.
3233 Description: Grant space more than available left space sometimes.
3234 Details : When then OST is about to be full, if two bulk writing from
3235 different clients came to OST. Accord the avaliable space of the
3236 OST, the first req should be permitted, and the second one
3237 should be denied by ENOSPC. But if the seconde arrived before
3238 the first one is committed. The OST might wrongly permit second
3239 writing, which will cause grant space > available space.
3242 Frequency : when client is evicted
3244 Description: Grant might be wrongly erased when osc is evicted by OST
3245 Details : when the import is evicted by server, it will fork another
3246 thread ptlrpc_invalidate_import_thread to invalidate the
3247 import, where the grant will be set to 0. While the original
3248 thread will update the grant it got when connecting. So if
3249 the former happened latter, the grant will be wrongly errased
3250 because of this race.
3255 Description: Checking Stale with correct fid
3256 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3257 to check whether it is stale, because sometimes, we want the
3258 enqueue happened anyway, and op_data.fid2 will not be initialized.
3260 Severity : enhancement
3262 Description: update patchless client
3263 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3266 Frequency : only with 2.4 kernel
3268 Description: random memory corruption
3269 Details : size of struct ll_inode_info is to big for union inode.u and this
3270 can be cause of random memory corruption.
3275 Description: Memory leak in recovery
3276 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3277 It should also check obd_fail before fsfilt_start, otherwise if
3278 fsfilt_start return -EROFS,(failover mds during mds recovery).
3279 then the req will return with repmsg->transno = 0 and rc = EROFS.
3280 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3281 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3282 be freed no matter whether fsfilt_commit success or not.
3285 Frequency : only with huge count clients
3287 Description: Prevents from taking the superblock lock in llap_from_page for
3289 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3290 for llap_from_page call in ll_removepage prevents from taking the
3291 superblock lock for a soon died page.
3296 Description: Not check open intent error before release open handle
3297 Details : in some rare cases, the open intent error is not checked before
3298 release open handle, which may cause
3299 ASSERTION(open_req->rq_transno != 0), because it tries to release
3300 the failed open handle.
3305 Description: Set cat log bitmap only after create log success.
3306 Details : in some rare cases, the cat log bitmap is set too early. and it
3307 should be set only after create log success.
3311 Description: Accessing a block bevice can re-enable I/O when Lustre is
3312 tearing down a device.
3313 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3318 Description: service threads may hog cpus when there are a lot of requests
3319 Details : Insert cond_resched to give other threads a chance to use some CPU
3324 Description: the cat log was not initialized in recovery
3325 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3326 unlink log on mds will not be initialized until mds post
3327 recovery. And also in mds post recovery, the unlink log will
3328 initialization will be done asynchronausly, so there will be race
3329 between add unlink log and unlink log initialization.
3333 Description: brw_stats were being printed incorrectly
3334 Details : brw_stats were being printed as log2 but all of them were not
3335 recorded as log2. Also remove some code duplication arising from
3336 filter_tally_{read,write}.
3340 Frequency : rare, only in recovery.
3341 Description: ASSERTION(req->rq_type != LI_POISON) failed
3342 Details : imp_lock should be held while iterating over imp_sending_list for
3343 prevent destroy request after get timeout in ptlrpc_queue_wait.
3347 Description: replay-single.sh test 52 fails
3348 Details : A lock's skiplist need to be cleanup when it being unlinked
3349 from its resource list.
3353 Description: Short directio read returns full requested size rather than
3355 Details : Direct I/O operations should return actual amount of bytes
3356 transferred rather than requested size.
3358 Severity : enhancement
3360 Description: metadata RPC reduction (e.g. for rm performance)
3361 Details : decrease the amount of synchronous RPC between clients and servers
3362 by canceling conflicing lock before the operation on the client
3363 and packing thier handles into the main operation RPC to server.
3365 Severity : enhancement
3367 Description: Async OSC create to avoid the blocking unnecessarily.
3368 Details : If an OST has no remaining object, system will block on the
3369 creation when it needs to create a new object on this OST. Now,
3370 ways use pre-created objects when available, instead of blocking on
3371 an empty osc while others are not empty. If we must block, we
3372 block for the shortest possible period of time.
3376 Description: improve handling recoverable errors
3377 Details : if request processig with error which can be recoverable on server
3378 request should be resend, otherwise page released from cache and
3381 Severity : enhancement
3383 Description: refine locking for avoid write wrong info into lov_objid file
3384 Details : fix possible races with add new target and write/update data in
3387 --------------------------------------------------------------------------------
3389 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3394 Frequency : on some architectures
3396 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3397 Details : Uninitialized flags sometimes cause configuration commands to
3401 Frequency : patchless clients only
3403 Description: missing __iget() symbol export
3404 Details : The __iget() symbol export is missing. To avoid the need for
3405 this on patchless clients the deathrow inode reaper is turned
3406 off, and we depend on the VM to clean up old inodes. This
3407 dependency was during via the fix for bug 12181.
3409 --------------------------------------------------------------------------------
3411 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3413 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3414 INTEROPERATE with older versions automatically. In many cases a
3415 special upgrade step is needed. Please read the
3416 user documentation before upgrading any part of a live system.
3417 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3418 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3419 * WARNING: Lustre configuration and startup changes are required with
3420 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3424 Severity : enhancement
3426 Description: MountConf
3427 Details : Lustre configuration is now managed via mkfs and mount
3428 commands instead of lmc and lconf. New obd types (MGS, MGC)
3429 are added for dynamic configuration management. See
3430 https://mail.clusterfs.com/wikis/lustre/MountConf for
3433 Severity : enhancement
3435 Description: dynamic OST addition
3436 Details : OSTs can now be added to a live filesystem
3438 Severity : enhancement
3440 Description: startup order invariance
3441 Details : MDTs and OSTs can be started in any order. Clients only
3442 require the MDT to complete startup.
3444 Severity : enhancement
3446 Description: parallel, asynchronous orphan cleanup
3447 Details : orphan cleanup is now performed in separate threads for each
3448 OST, allowing parallel non-blocking operation.
3450 Severity : enhancement
3452 Description: optimized stripe assignment
3453 Details : stripe assignments are now made based on ost space available,
3454 ost previous usage, and OSS previous usage, in order to try
3455 to optimize storage space and networking resources.
3457 Severity : enhancement
3459 Description: Permanently set tunables
3460 Details : All writable /proc/fs/lustre tunables can now be permanently
3461 set on a per-server basis, at mkfs time or on a live system.
3463 Severity : enhancement
3465 Description: Lustre message v2
3466 Details : Add lustre message format v2.
3468 Severity : enhancement
3470 Description: client OST exclusion list
3471 Details : Clients can be started with a list of OSTs that should be
3472 declared "inactive" for known non-responsive OSTs.
3476 Description: SPEC SFS validation failure on NFS v2 over lustre.
3477 Details : Changes the blocksize for regular files to be 2x RPC size,
3478 and not depend on stripe size.
3480 Severity : enhancement
3482 Description: Multiple MD RPCs in flight.
3483 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3484 To avoid overly-overloading MDS, introduce a limit on number of
3485 MDS RPCs in flight for a single client and add /proc controls
3486 to adjust this limit.
3488 Severity : enhancement
3490 Description: client read/write statistics
3491 Details : Add client read/write call usage stats for performance
3492 analysis of user processes.
3493 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3494 file access. extents_stats shows chunk size distribution.
3495 extents_stats_per_process show chunk size distribution per
3498 Severity : enhancement
3500 Description: mds statistics
3501 Details : Add detailed mds operations statistics in
3502 /proc/fs/lustre/mds/*/stats.
3506 Description: Failure of copying files with lustre special EAs.
3507 Details : Client side always return success for setxattr call for lustre
3508 special xattr (currently only "trusted.lov").
3513 Description: Refcount LNET uuids
3514 Details : The global LNET uuid list grew linearly with every startup;
3515 refcount repeated list entries instead of always adding to
3518 Severity : enhancement
3520 Description: Dynamic service threads
3521 Details : Within a small range, start extra service threads
3522 automatically when the request queue builds up.
3525 Frequency : mixed-endian client/server environments
3527 Description: mixed-endian crashes
3528 Details : The new msg_v2 system had some failures in mixed-endian
3531 Severity : enhancement
3533 Description: Easy OST removal
3534 Details : OSTs can be permanently deactivated with e.g. 'lctl
3535 conf_param lustre-OST0001.osc.active=0'
3537 Severity : enhancement
3539 Description: MGS proc entries
3540 Details : Added basic proc entries for the MGS showing what filesystems
3543 Severity : enhancement
3545 Description: provide MGS failover
3546 Details : Added config lock reacquisition after MGS server failover.
3548 Severity : enhancement
3550 Description: add Linux 2.4 support
3551 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3555 Description: a large application tries to do I/O to the same resource and dies
3556 in the middle of it.
3557 Details : Check the req->rq_arrival time after the call to
3558 ost_brw_lock_get(), but before we do anything about
3559 processing it & sending the BULK transfer request. This
3560 should help move old stale pending locks off the queue as
3561 quickly as obd_timeout.
3564 Frequency : when an incorrect nid is specified during startup
3566 Description: ptlrpc connect to non-existant node causes kernel crash
3567 Details : LNET can't be re-entered from an event callback, which
3568 happened when we expire a message after the export has been
3569 cleaned up. Instead, hand the zombie cleanup off to another
3572 Severity : enhancement
3574 Description: plain/inodebits lock performance improvement
3575 Details : Grouping plain/inodebits in granted list by their request modes
3576 and bits policy, thus improving the performance of search through
3580 Frequency : only if OST filesystem is corrupted
3582 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3583 Details : for a short time RPCs with bulk IO are in the replay list,
3584 but replay of bulk IOs is unimplemented. If the OST filesystem
3585 is corrupted due to disk cache incoherency and then replay is
3586 started it is possible to trip an assertion. Avoid putting
3587 committed RPCs into the replay list at all to avoid this issue.
3590 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3592 Description: System hang on startup
3593 Details : This bug allowed the liblustre (e.g. catamount) client to
3594 return to the app before handling all startup RPCs. This
3595 could leave the node unresponsive to lustre network traffic
3596 and manifested as a server ptllnd timeout.
3598 Severity : enhancement
3600 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3601 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3602 boolean to control whether to print peer debug info when a
3603 client's RPC times out.
3606 Frequency : only for kernels with patches from Lustre below 1.4.3
3608 Description: Remove old rdonly API
3609 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3612 Frequency : only for devices with external journals
3614 Description: Set external device read-only also
3615 Details : During a commanded failover stop, we set the disk device
3616 read-only while the server shuts down. We now also set any
3617 external journal device read-only at the same time.
3620 Frequency : when upgrading from 1.4 while trying to change parameters
3622 Description: The wrong (new) MDC name was used when setting parameters for
3623 upgraded MDT's. Also allows changing of OSC (and MDC)
3624 parameters if --writeconf is specified at tunefs upgrade time.
3627 Frequency : when setting specific ost indicies
3629 Description: QOS code breaks on skipped indicies
3630 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3631 created with --index need not be sequential.
3633 Severity : enhancement
3635 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3636 Details : The uninit_groups feature works in conjunction with the kernel
3637 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3638 up the pass1 processing of e2fsck. This is a read-only feature
3639 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3640 mount filesystems that have had this feature enabled.
3642 Severity : enhancement
3644 Description: Improve multi-block allocation algorithm to avoid fragmentation
3645 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3646 allocation locality and avoid filesystem fragmentation.
3648 ------------------------------------------------------------------------------
3650 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3652 * Support for kernels:
3653 2.4.21-47.0.1.EL (RHEL 3)
3654 2.6.5-7.283 (SLES 9)
3655 2.6.9-42.0.10.EL (RHEL 4)
3656 2.6.12.6 vanilla (kernel.org)
3657 2.6.16.27-0.9 (SLES 10)
3658 * Recommended e2fsprogs version: 1.39.cfs5
3660 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3664 Frequency : occasional, depends on client load and configuration
3665 Bugzilla : 12181, 12203
3666 Description: data loss for recently-modified files
3668 Details : In some cases it is possible that recently written or created
3669 files may not be written to disk in a timely manner (this should
3670 normally be within 30s unless client IO load is very high).
3671 The problem appears as zero-length files or files that are a
3672 multiple of 1MB in size after a client crash or client eviction
3673 that are missing data at the end of the file.
3675 This problem is more likely to be hit on clients where files are
3676 repeatedly created and unlinked in the same directory, clients
3677 have a large amount of RAM, have many CPUs, the filesystem has
3678 many OSTs, the clients are rebooted frequently, and/or the files
3679 are not accessed by other nodes after being written.
3681 The presence of the problem can be detected by looking at
3682 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3683 smaller than the second (nr_unused) then dirty files will not
3684 be flushed automatically to disk. "sync; sleep 10" should be
3685 run several times on the node before unmounting it to update
3686 Lustre (this is also safe to run on nodes without this problem).
3688 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3689 that can cause this same problem, so customers using that kernel
3690 also need to update the kernel in addition to Lustre. In order
3691 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3693 It is normal that files written just before a client crash (less
3694 than 30s) may not yet have been flushed to disk, even for local
3698 Frequency : frequent on thin XT3 nodes
3700 Description: UUID collision on thin XT3 Linux nodes
3701 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3702 randomly, since we relied on an insufficiently-seeded PRNG.
3707 Description: OSS hangs after "All ost request buffers busy"
3708 Details : A deadlock between quota and journal operations caused OSS
3709 hangs after printing "All ost request buffers busy."
3712 Frequency : always on liblustre builds
3714 Description: Cleanup compiler warnings on liblustre
3717 Frequency : always on liblustre builds on XT3
3719 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3723 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3724 > 32000 subdirectories
3725 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3726 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3727 subdirectory count crosses 32000. This will aid e2fsck to
3728 correctly handle more than 32000 subdirectories.
3731 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3733 Description: System hang on startup
3734 Details : This bug allowed the liblustre (e.g. catamount) client to
3735 return to the app before handling all startup RPCs. This
3736 could leave the node unresponsive to lustre network traffic
3737 and manifested as a server ptllnd timeout.
3739 Severity : enhancement
3741 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3742 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3743 boolean to control whether to print peer debug info when a
3744 client's RPC times out.
3749 Description: make O_SYNC working on 2.6 kernels
3750 Details : 2.6 kernels use different method for mark pages for write,
3751 so need add a code to lustre for O_SYNC work.
3756 Description: Failure to close file and release space on NFS
3757 Details : Put inode details into lock acquired in ll_intent_file_open.
3758 Use mdc_intent_lock in ll_intent_open to properly
3759 detect all kind of errors unhandled by mdc_enqueue.
3764 Description: proc file read during shutdown sometimes raced obd removal,
3766 Details : Add lock to prevent obd access after proc file removal.
3769 Frequency : Only for files larger than 4GB on 32-bit clients.
3771 Description: improperly doing page alignment of locks
3772 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3773 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3778 Description: RPCs being resent when they shouldn't be
3779 Details : Some RPCs that should not be resent are being resent. This
3780 can cause inconsistencies in the RPC state machine. Do not
3781 resend such requests.
3784 Frequency : rare, only with NFS export
3786 Description: Crash on NFS re-export node
3787 Details : under very unusual load conditions an assertion is hit in
3788 ll_intent_file_open()
3791 Frequency : only if OST filesystem is corrupted
3793 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3794 Details : for a short time RPCs with bulk IO are in the replay list,
3795 but replay of bulk IOs is unimplemented. If the OST filesystem
3796 is corrupted due to disk cache incoherency and then replay is
3797 started it is possible to trip an assertion. Avoid putting
3798 committed RPCs into the replay list at all to avoid this issue.
3803 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3804 Details : Large single O_DIRECT read and write calls can fail to allocate
3805 a sufficiently large buffer to process the request. In case of
3806 allocation failure the allocation is retried with a smaller
3807 buffer and broken into smaller requests.
3809 Severity : enhancement
3811 Description: Add -o localflock option to simulate old noflock behaviour.
3812 Details : This will achieve local-only flock/fcntl locks coherentness.
3817 Description: versioning check is incomplete
3818 Details : Checking the version difference of client vs. server, report
3819 error if the gap is too big.
3824 Description: add support PG_writeback bit
3825 Details : add support for PG_writeback bit for Lustre, for more carefull
3826 work with page cache in 2.6 kernel. This also fix some deadlocks
3827 and remove hack for work O_SYNC with 2.6 kernel.
3829 Severity : enhancement
3831 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3832 Details : The uninit_groups feature works in conjunction with the kernel
3833 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3834 up the pass1 processing of e2fsck. This is a read-only feature
3835 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3836 mount filesystems that have had this feature enabled.
3838 Severity : enhancement
3840 Description: Improve multi-block allocation algorithm to avoid fragmentation
3841 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3842 allocation locality and avoid filesystem fragmentation.
3844 ------------------------------------------------------------------------------
3846 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3848 * Support for kernels:
3849 2.6.9-42.0.3.EL (RHEL 4)
3850 2.6.5-7.276 (SLES 9)
3851 2.4.21-47.0.1.EL (RHEL 3)
3852 2.6.12.6 vanilla (kernel.org)
3853 2.6.16.21-0.8 (SLES10)
3854 * Recommended e2fsprogs version: 1.39.cfs2-0
3856 * The backwards-compatible /proc/sys/portals symlink has been removed
3857 in this release. Before upgrading, please ensure that you change
3858 any configuration scripts or /etc/sysctl.conf files that access
3859 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3860 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3861 in advance of the upgrade on any system running Lustre 1.4.6 or
3862 newer, since /proc/sys/lnet was added in that version.
3863 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3867 Frequency : only when quota is used
3869 Description: avoid scanning export list for quota master
3870 Details : Change the algorithms to avoid scanning export list in order
3871 to improve the efficiency.
3874 Frequency : MDS failover only, very rarely
3876 Description: "went back in time" messages on mds failover
3877 Details : The greatest transno may be lost when the current operation
3878 finishes with an error (transno==0) and the client's last_rcvd
3879 record is over-written. Save the greatest transno in the
3880 mds_last_transno for this case.
3883 Frequency : always for specific kernels and striping counts
3885 Description: client may get "Matching packet too big" without ACL support
3886 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3887 when trying to access files in certain configurations. The
3888 clients should in fact be denied when mounting because they do
3889 not understand ACLs.
3892 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3894 Description: many clients connecting with IO in progress causes connect timeouts
3895 Details : Avoid synchronous journal commits to avoid delays caused by many
3896 clients connecting/disconnecting when bulk IO is in progress.
3897 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3898 OST_IO_PORTAL to avoid delays behind potentially many pending
3902 Frequency : occasionally with multiple writers to a single file
3904 Description: shared writes to file may result in wrong size reported by stat()
3905 Details : Allow growing of kms when extent lock is cancelled
3908 Frequency : always with random mmap IO to multi-striped file
3910 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3911 Details : Only if the hole is at the end of OST object so that kms is too
3912 small. Fix is to increase kms accordingly in ll_nopage.
3915 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3917 Description: writes to a missing object would leak memory on the OST
3918 Details : If there is an inconsistency between the MDS and OST filesystems,
3919 such that the MDS references an object that doesn't exist, writes
3920 to that object will leak memory due to incorrect cleanup in the
3921 error handling path, eventually running out of memory on the OST.
3926 Description: Creating too long symlink causes lustre errors
3927 Details : Check symlink and name lengths before sending requests to MDS.
3930 Frequency : only if flock is enabled (not on by default)
3932 Description: posix locks not released on fd closure on 2.6.9+
3933 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3934 kernels, this caused such locks not to be released on fd close and
3935 then assertions on fs unmount about still used locks.
3938 Frequency : MDS failover only, very rarely
3940 Description: clients may get ASSERTION(granted_lock != NULL)
3941 Details : When request was taking a long time, and a client was resending
3942 a getattr by name lock request. The were multiple lock requests
3943 with the same client lock handle and
3944 mds_getattr_name->fixup_handle_for_resent_request found one of the
3945 lock handles but later failed with ASSERTION(granted_lock != NULL).
3950 Description: handle->h_buffer_credits > 0, assertion failure
3951 Details : h_buffer_credits is zero after truncate, causing assertion
3952 failure. This patch extends the transaction or creates a new
3956 Frequency : NFS re-export or patchless client
3957 Bugzilla : 11179, 10796
3958 Description: Crash on NFS re-export node (__d_move)
3959 Details : We do not want to hash the dentry if we don't have a lock.
3960 But if this dentry is later used in d_move, we'd hit uninitialised
3961 list head d_hash, so we just do this to init d_hash field but
3962 leave dentry unhashed.
3965 Frequency : NFS re-export or patchless client
3967 Description: NFS exports has problem with symbolic link
3968 Details : lustre client didn't properly install dentry when re-exported
3969 to NFS or running patchless client.
3972 Frequency : NFS re-export or patchless client
3974 Description: Various nfs/patchless fixes.
3975 Details : fixes reuse disconected alias for lookup process - this fixes
3976 warning "find_exported_dentry: npd != pd",
3977 fix permission error with open files at nfs.
3978 fix apply umask when do revalidate.
3981 Frequency : occasional
3983 Description: Crash on NFS re-export node
3984 Details : calling clear_page() on the wrong pointer triggered oops in
3985 generic_mapping_read().
3988 Frequency : rarely, using O_DIRECT IO
3990 Description: unaligned directio crashes client with LASSERT
3991 Details : check for unaligned buffers before trying any requests.
3994 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3996 Description: stale data returned from RAID cache
3997 Details : If only a small amount of IO is done to the RAID device before
3998 reading it again it is possible to get stale data from the RAID
3999 cache instead of reading it from disk.
4002 Frequency : always for sles10 kernel
4004 Description: sles10 support
4005 Details : ll_follow_link: compile fixes and using of nd_set_link
4006 under newer kernels.
4009 Frequency : depends on arch, kernel and compiler version, always on sles10
4012 Description: recursive or deep enough symlinks cause stack overflow
4013 Details : getting rid of large stack-allocated variable in
4017 Frequency : depends on hardware
4019 Description: lustre write performance loss in the SLES10 kernel
4020 Details : the performance loss is caused by using of write barriers in the
4021 ext3 code. The SLES10 kernel turns barrier support on by
4022 default. The fix is to undo that change for ldiskfs.
4024 ------------------------------------------------------------------------------
4026 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4028 * Support for kernels:
4029 2.6.9-42.0.3EL (RHEL 4)
4030 2.6.5-7.276 (SLES 9)
4031 2.4.21-47.0.1.EL (RHEL 3)
4032 2.6.12.6 vanilla (kernel.org)
4036 Frequency : quota enabled and large files being deleted
4038 Description: releasing more than 4GB of quota at once hangs OST
4039 Details : If a user deletes more than 4GB of files on a single OST it
4040 will cause the OST to spin in an infinite loop. Release
4041 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4046 Description: statfs data retrieved from /proc may be stale or zero
4047 Details : When reading per-device statfs data from /proc, in the
4048 {kbytes,files}_{total,free,avail} files, it may appear
4049 as zero or be out of date.
4052 Frequency : systems with MD RAID1 external journal devices
4054 Description: lconf's call to blkid is confused by RAID1 journal devices
4055 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4056 of returning all block devices that match the journal UUID.
4059 Frequency : always, for aggregate stripe size over 4GB
4061 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4062 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4063 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4066 Frequency : always if "lfs find" used on a local file/directory
4068 Description: "lfs find" segfaults if used on a local file/directory
4069 Details : The case where a directory component was not specified wasn't
4070 handled correctly. Handle this properly.
4073 Frequency : always on ppc64
4075 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4076 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4077 perform endianness conversion.
4080 Frequency : rarely (truncate to non-zero file size after write under load)
4081 Bugzilla : 10730, 10687
4082 Description: Files padded with zeros to next 4K multiple
4083 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4084 it is possible that files that are truncated to a non-zero size
4085 immediately after being written are filled with zero bytes beyond
4086 the truncated size. No file data is lost.
4088 Severity : enhancement
4090 Description: Allow recovery/failover for liblustre clients.
4091 Details : liblustre clients were unaware of failover configurations until
4094 Severity : enhancement
4096 Description: user file locks should fail when not mounting with flock option
4097 Details : Set up an error-returning stub in ll_file_operations.lock field
4098 to prevent incorrect behaviour when client is mounted without
4099 flock option. Also, set up properly f_op->flock field for
4103 Frequency : always on ia64
4105 Description: "lfs df" loops on printing out MDS statfs information
4106 Details : The obd_ioctl_data was not initialized and in some systems
4107 this caused a failure during the ioctl that did not return
4108 an error. Initialize the struct and return an error on failure.
4111 Frequency : SLES 9 only
4113 Description: Error of copying files with lustre special EAs as root
4114 Details : Client side always return success for setxattr call for lustre
4115 special xattr (currently only "trusted.lov").
4118 Frequency : rarely on clusters with both ia64+i386 clients
4120 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4121 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4122 clients are doing concurrent writes to the same file it is
4123 possible that the ia64 clients may LASSERT because the OST
4124 extent locks are not PAGE_SIZE aligned. Ensure that grown
4125 locks are always aligned on the request boundary.
4128 Frequency : specific use, occasional
4130 Description: Overwriting in use executable truncates on-disk binary image
4131 Details : If one node attempts to overwrite an executable in use by
4132 another node, we now correctly return ETXTBSY instead of
4133 truncating the file.
4135 Severity : enhancement
4137 Description: Async OSC create to avoid the blocking unnecessarily.
4138 Details : If an OST has no remaining object, system will block on the creating
4139 when need to create a new object on this OST. Now, ways use
4140 pre-created objects when available, instead of blocking on an
4141 empty osc while others are not empty. If we must block, we block
4142 for the shortest possible period of time.
4147 Description: chmod on Lustre root is propagated to other clients
4148 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4149 invalid by the follow_mount time.
4152 Frequency : liblustre clients only
4154 Description: Race in 'instant cancel' lock handling could lead to such locks
4155 never to be granted in case of SMP MDS
4156 Details : Do not destroy not yet granted but cbpending locks in
4160 Frequency : replay/resend of open
4162 Description: non null lock assetion failure in mds_intent_policy
4163 Details : Trying to replay/resend lockless open requests resulted in
4164 mds_open() returning 0 with no lock. Now it sets a flag if
4165 a lock is going to be returned.
4167 Severity : enhancement
4169 Description: Checksum enhancements
4170 Details : New checksum enhancements allow for resending RPCs that failed
4173 Severity : enhancement
4175 Description: Tunables on number of dirty pages in cacche
4176 Details : Allow to set limit on number of dirty pages cached.
4181 Description: client crash on unmount - lock still has references
4182 Details : In some error handling cases it was possible to leak a lock
4183 reference on a client while accessing a file. This was not
4184 harmful to the client during operation, but would cause the
4185 client to crash when the filesystem is unmounted.
4188 Frequency : specific case, rare
4190 Description: ETXTBSY on mds though file not in use by client
4191 Details : ETXTBSY is no longer incorrectly returned when attempting to
4192 chmod or chown a directory that the user previously tried to
4193 execute or a currently-executing binary.
4196 Frequency : extremely rare except on liblustre-based clients
4198 Description: Lustre space not freed when files are deleted
4199 Details : Clean up open-unlinked files after client eviction. Previously
4200 the unlink was skipped and the files remained as orphans.
4205 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4206 Details : In some cases it was possible to send a blocking callback to
4207 a client doing a glimpse, even though that client didn't get
4208 a lock granted. When the glimpse lock is cancelled on the OST
4209 the freed lock is left on the waiting list and corrupted the list.
4212 Frequency : all core dumps
4214 Description: Broke core dumps to lustre
4215 Details : Negative dentry may be unhashed if parent does not have UPDATE
4216 lock, but some callers, e.g. do_coredump, expect dentry to be
4217 hashed after successful create, hash it in ll_create_it.
4219 ------------------------------------------------------------------------------
4221 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4223 * Support for kernels:
4224 2.6.9-42.0.2.EL (RHEL 4)
4225 2.6.5-7.276 (SLES 9)
4226 2.4.21-40.EL (RHEL 3)
4227 2.6.12.6 vanilla (kernel.org)
4231 Frequency : always on RHEL 3
4233 Description: Number of open files grows over time
4234 Details : The number of open files grows over time, whether or not
4235 Lustre is started. This was due to a filp leak introduced
4236 by one of our kernel patches.
4238 ------------------------------------------------------------------------------
4240 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4242 * Support for kernels:
4243 2.6.9-42.EL (RHEL 4)
4244 2.6.5-7.276 (SLES 9)
4245 2.4.21-40.EL (RHEL 3)
4246 2.6.12.6 vanilla (kernel.org)
4251 Bugzilla : 5719, 9635, 9792, 9684
4252 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4253 Details : If a server is under heavy load and cannot reply to new
4254 connection requests before the client resends the (re)connect,
4255 the connection handling code can behave badly if two service
4256 threads are concurrently handing separate (re)connections from
4257 the same client. Add better locking to the connection handling
4258 code, and ensure that only a single connection will be processed
4259 for a given client UUID, even if the lock is dropped.
4261 Severity : enhancement
4263 Description: add TCP zero-copy support to kernel
4264 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4265 sends if the hardware supports scatter-gather and checksumming.
4266 This allows socklnd to do client-write and server-read more
4267 efficiently and reduce CPU utilization from skbuf copying.
4270 Frequency : only if NFS exporting from client
4272 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4273 Details : If setting attributes on a file created under NFS that had
4274 never been opened it would be possible to oops the client
4275 if the file had no objects.
4278 Frequency : always for liblustre
4280 Description: liblustre client does MDS+OSTs setattr RPC for each write
4281 Details : When doing a write from a liblustre client, the client
4282 incorrectly issued an RPC to the MDS and each OST the file was
4283 striped over in order to update the timestamps. When writing
4284 with small chunks and many clients this could overwhelm the MDS
4285 with RPCs. In all cases it would slow down the write because
4286 these RPCs are unnecessary.
4288 Severity : enhancement
4290 Description: allow number of MDS service threads to be changed at module load
4291 Details : It is now possible to change the number of MDS service threads
4292 running. Adding "options mds mds_num_threads={N}" to the MDS's
4293 /etc/modprobe.conf will set the number of threads for the next
4294 time Lustre is restarted (assuming the "mds" module is also
4295 reloaded at that time). The default number of threads will
4296 stay the same, 32 for most systems.
4301 Description: OST crash if filesystem is unformatted or corrupt
4302 Details : If an OST is started on a device that has never been formatted
4303 or if the filesystem is corrupt and cannot even mount then the
4304 error handling cleanup routines would dereference a NULL pointer.
4309 Description: NULL pointer deref in llap_from_page.
4310 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4311 incorrect) mapping if the page was truncated/reclaimed while it was
4312 searched for. Check for this condition and skip such pages when
4313 doing readahead. Introduce extra check to llap_from_page() to
4314 verify page->mapping->host is non-NULL (so page is not anonymous).
4317 Frequency : Sometimes when using sys_sendfile
4319 Description: "page not covered by a lock" warnings from ll_readpage
4320 Details : sendfile called ll_readpage without right page locks present.
4321 Now we introduced ll_file_sendfile that does necessary locking
4322 around call to generic_file_sendfile() much like we do in
4326 Frequency : with certain MDS communication failures at client mount time
4328 Description: NULL pointer deref after failed client mount
4329 Details : a client connection request may delayed by the network layer
4330 and not be sent until after the PTLRPC layer has timed out the
4331 request. If the client fails the mount immediately it will try
4332 to clean up before the network times out the request. Add a
4333 reference from the request import to the obd device and delay
4334 the cleanup until the network drops the request.
4337 Frequency : occasionally during client (re)connect
4339 Description: assertion failure during client (re)connect
4340 Details : processing a client connection request may be delayed by the
4341 client or server longer than the client connect timeout. This
4342 causes the client to resend the connection request. If the
4343 original connection request is replied in this interval, the
4344 client may trip an assertion failure in ptlrpc_connect_interpret()
4345 which thought it would be the only running connect process.
4348 Frequency : only with obd_echo servers and clients that are rebooted
4350 Description: kernel BUG accessing uninitialized data structure
4351 Details : When running an obd_echo server it did not start the ping_evictor
4352 thread, and when a client was evicted an uninitialized data
4353 structure was accessed. Start the ping_evictor in the RPC
4354 service startup instead of the OBD startup.
4356 Severity : enhancement
4357 Bugzilla : 10193 (patchless)
4358 Description: Remove dependency on various unexported kernel interfaces.
4359 Details : No longer need reparent_to_init, exit_mm, exit_files,
4360 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4363 Frequency : rare (only users of deprecated and unsupported LDAP config)
4365 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4366 Details : LDAP apparently contains 'lustreName' attributes instead of
4367 'name'. A simple remapping of the name is sufficient.
4370 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4372 Description: waiting_locks_callback trips kernel BUG if client is evicted
4373 Details : Running with the dump_on_timeout debug flag turned on makes
4374 it possible that the waiting_locks_callback() can try to dump
4375 the Lustre kernel debug logs from an interrupt handler. Defer
4376 this log dumping to the expired_lock_main() thread.
4378 Severity : enhancement
4380 Description: Support NFS exporting on 2.6 kernels.
4381 Details : Implement non-rawops metadata methods for NFS server to use without
4382 changing NFS server code.
4385 Frequency : very rare (synthetic metadata workload only)
4387 Description: two racing renames might cause an MDS thread to deadlock
4388 Details : Running the "racer" program may cause one MDS thread to rename
4389 a file from being the source of a rename to being the target of
4390 a rename at exactly the same time that another thread is doing
4391 so, and the second thread has already enqueued these locks after
4392 doing a lookup of the target and is trying to relock them in
4393 order. Ensure that we don't try to re-lock the same resource.
4396 Frequency : only very large systems with liblustre clients
4398 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4399 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4400 all OSTs in parallel. This allows the eviction of stale liblustre
4401 clients to proceed much faster than if they were done in series,
4402 and also offers similar improvements for other set_info RPCs.
4407 Description: excessive CPU usage during initial read phase on client
4408 Details : During the initial read phase on a client, it would agressively
4409 retry readahead on the file, consuming too much CPU and impacting
4410 performance (since 1.4.5.8). Improve the readahead algorithm
4411 to avoid this, and also improve some other common cases (read
4412 of small files in particular, where "small" is files smaller than
4413 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4418 Description: MDS crash when receiving packet with unknown intent.
4419 Details : Do not LBUG in unknown intent case, just return -EFAULT
4421 Severity : enhancement
4422 Bugzilla : 9293, 9385
4423 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4424 Details : Do not serialize getattr (non-intent version) and statfs.
4427 Frequency : occasional, when OST network is overloaded/intermittent
4429 Description: client evicted by OST after bulk IO timeout
4430 Details : If a client sends a bulk IO request (read or write) the OST
4431 may evict the client if it is unresposive to its data GET/PUT
4432 request. This is incorrect if the network is overloaded (takes
4433 too long to transfer the RPC data) or dropped the OST GET/PUT
4434 request. There is no need to evict the client at all, since
4435 the pinger and/or lock callbacks will handle this, and the
4436 client can restart the bulk request.
4439 Frequency : Always when mmapping file with no objects
4441 Description: client crashes when mmapping file with no objects
4442 Details : Check that we actually have objects in a file before doing any
4443 operations on objects in ll_vm_open, ll_vm_close and
4449 Description: Request leak when working with deleted CWD
4450 Details : Introduce advanced request refcount tracking for requests
4451 referenced from lustre intent.
4453 Severity : Enhancement
4455 Description: Cache open file handles on client.
4456 Details : MDS now will return special lock along with openhandle, if
4457 requested and client is allowed to hold openhandle, even if unused,
4458 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4459 closing files for every read/write openration.
4461 Severity : Enhancement
4463 Description: Cache open negative dentries on client when possible.
4464 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4465 negative dentries on lock revocation.
4470 Description: Remounting a client read-only wasn't possible with a zconf mount
4471 Details : It wasn't possible to remount a client read-only with llmount.
4473 Severity : enhancement
4474 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4475 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4476 MPICH is included to add Lustre-specific ADIO interfaces.
4477 This is based closely on the UFS ADIO layer and only differs
4478 in file creation, in order to allow the OST striping to be set.
4479 This is user-contributed code and not supported by CFS.
4484 Description: extended inode attributes (immutable, append-only) work improperly
4485 when 2.4 and 2.6 kernels are used on client/server or vice versa
4486 Details : Introduce kernel-independent values for these flags.
4488 Severity : enhancement
4491 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4492 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4493 granularity. Now, Fractional MB granularity is be supported,
4494 this is very useful for low memory system.
4496 Severity : enhancement
4498 Description: Getattr by fid
4499 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4500 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4501 have locks and avoids getattr rpc requests for it on every stat.
4504 Frequency : Always, for filesystems larger than 2TB
4506 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4507 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4508 filesystems larger than 2TB to be mounted with mballoc enabled.
4511 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4513 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4514 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4515 filesystems larger than 2TB is mounted then the kernel will
4516 silently corrupt the start of the filesystem. CONFIG_LBD is
4517 enabled for all CFS-supported kernels, but the possibility of
4518 this happening with a modified kernel config exists.
4520 Severity : enhancement
4522 Description: add client O_DIRECT support for 2.6 kernels
4523 Details : It is now possible to do O_DIRECT reads and writes to files
4524 in the Lustre client mountpoint on 2.6 kernel clients.
4526 Severity : enhancement
4528 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4529 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4530 parallel, not waiting for response from every OST before sending
4531 a rpc to the next OST.
4536 Description: setattr vs write race when updating file timestamps
4537 Details : Client processes that update a file timestamp into the past
4538 right after writing to the file (e.g. tar) it is possible that
4539 the updated file modification time can be reset to the current
4540 time due to a race between processing the setattr and write RPC.
4542 Severity : enhancement
4544 Description: Bring 'lfs find' closer in line with regular Linux find.
4545 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4546 -print0 options and obtains all the needed info through the lustre
4549 Severity : enhancement
4551 Description: support up to 1024 configured devices on one node
4552 Details : change obd_dev array from statically allocated to dynamically
4553 allocated structs as they are first used to reduce memory usage
4558 Description: Flush dirty partially truncated pages during truncate
4559 Details : Immediatelly flush partially truncated pages in filter_setattr,
4560 this way we completely avoid having any pages in page cache on OST
4561 and can retire ugly workarounds during writes to flush such pages.
4566 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4567 Details : protect inode from truncation within vfs_unlink() context
4568 just take a reference before calling vfs_unlink() and release it
4569 when parent's i_sem is free.
4574 Description: last_id value checked outside lock on OST caused LASSERT failure
4575 Details : If there were multiple MDS->OST object precreate requests in
4576 flight, it was possible that the OST's last object id was checked
4577 outside a lock and incorrectly tripped an assertion. Move checks
4578 inside locks, and discard old precreate requests.
4581 Frequency : always, if extents are used on OSTs
4583 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4584 Details : OSTs using the extents format would not zero the high 16 bits of
4585 the index physical block number. This is not a problem for any
4586 OST filesystems smaller than 16TB, and no kernels support ext3
4587 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4588 new/modified files) and can be fixed for existing filesystems
4589 with e2fsprogs-1.39-cfs1.
4594 Description: import connection selection may be incorrect if timer wraps
4595 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4596 import connections to be ignored if the 32-bit jiffies counter
4597 wraps. Use a 64-bit jiffies counter.
4600 Frequency : very large clusters immediately after boot
4602 Description: LNET request buffers exhausted under heavy short-term load
4603 Details : If a large number of client requests are generated on a service
4604 that has previously never seen so many requests it is possible
4605 that the request buffer growth cannot keep up with the spike in
4606 demand. Instead of dropping incoming requests, they are held in
4607 the LND until the RPC service can accept more requests.
4610 Frequency : Sometimes during replay
4612 Description: Assertion failure in ll_local_open after replay.
4613 Details : If replay happened on an open request reply before we were able
4614 to set replay handler, reply will become not swabbed tripping the
4615 assertion in ll_local_open. Now we set the handler right after
4616 recognising of open request
4619 Frequency : very rare
4621 Description: kernel reports "badness in vsnprintf"
4622 Details : Reading from the "recovery_status" /proc file in small chunks
4623 may cause a negative length in lprocfs_obd_rd_recovery_status()
4624 call to vsnprintf() (which is otherwise harmless). Exit early
4625 if there is no more space in the output buffer.
4627 Severity : enhancement
4629 Description: clear OBD RPC statistics by writing to them
4630 Details : It is now possible to clear the OBD RPC statistics by writing
4631 to the "stats" file.
4636 Description: Client mtime is not the same on different clients after utimes
4637 Details : In some cases, the client was using the utimes() syscall on
4638 a file cached on another node. The clients now validate the
4639 ctime from the MDS + OSTs to determine which one is right.
4644 Description: Inability to activate failout mode
4645 Details : lconf script incorrectly assumed that in pythong string's numeric
4646 value is used in comparisons.
4649 Frequency : always with multiple stripes per file
4651 Description: Inefficient object allocation for mutli-stripe files
4652 Details : When selecting which OSTs to stripe files over, for files with
4653 a stripe count that divides evenly into the number of OSTs,
4654 the MDS is always picking the same starting OST for each file.
4655 Return the OST selection heuristic to the original design.
4660 Description: mount failures may take full timeout to return an error
4661 Details : Under some heavy load conditions it is possible that a
4662 failed mount can wait for the full obd_timeout interval,
4663 possibly several minutes, before reporting an error.
4664 Instead return an error as soon as the status is known.
4666 Frequency : quota enabled and large files being deleted
4668 Description: releasing more than 4GB of quota at once hangs OST
4669 Details : If a user deletes more than 4GB of files on a single OST it
4670 will cause the OST to spin in an infinite loop. Release
4671 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4676 Description: statfs data retrieved from /proc may be stale or zero
4677 Details : When reading per-device statfs data from /proc, in the
4678 {kbytes,files}_{total,free,avail} files, it may appear
4679 as zero or be out of date.
4682 Frequency : systems with MD RAID1 external journal devices
4684 Description: lconf's call to blkid is confused by RAID1 journal devices
4685 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4686 of returning all block devices that match the journal UUID.
4689 Frequency : always, for aggregate stripe size over 4GB
4691 Description: assertion fails when trying to use 4GB stripe size
4692 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4693 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4696 Frequency : always on ppc64
4698 Description: the first write on an ext3 filesystem with mballoc got stuck
4699 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4700 perform endianness conversion.
4702 ------------------------------------------------------------------------------
4704 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4706 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4707 INTEROPERATE with older versions automatically. Please read the
4708 user documentation before upgrading any part of a live system.
4709 * WARNING: Lustre networking configuration changes are required with
4710 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4713 * Support for kernels:
4714 2.6.9-22.0.2.EL (RHEL 4)
4715 2.6.5-7.244 (SLES 9)
4716 2.6.12.6 vanilla (kernel.org)
4719 Severity : enhancement
4720 Bugzilla : 7981/8208
4721 Description: Introduced Lustre Networking (LNET)
4722 Details : LNET is new networking infrastructure for Lustre, it includes
4723 a reorganized network configuration mode (see the user
4724 documentation for full details) as well as support for routing
4725 between different network fabrics. Lustre Networking Devices
4726 (LNDs) for the supported network fabrics have also been
4727 created for this new infrastructure.
4729 Severity : enhancement
4730 Description: Introduced Access control lists
4731 Details : clients can set ACLs on files and directories in order to have
4732 more fine-grained permissions than the standard Unix UGO+RWX.
4733 The MDS must be started with the "-o acl" mount option.
4735 Severity : enhancement
4736 Description: Introduced filesystem quotas
4737 Details : Administrators may now establish per-user quotas on the
4740 Severity : enhancement
4742 Description: Configuration change for the XT3
4743 The PTLLND is now used to run Lustre over Portals on the XT3
4744 The configure option(s) --with-cray-portals are no longer used.
4745 Rather --with-portals=<path-to-portals-includes> is used to
4746 enable building on the XT3. In addition to enable XT3 specific
4747 features the option --enable-cray-xt3 must be used.
4752 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4753 Details : A race between allocating a new llcd and re-getting the llcd_lock
4754 allowed another thread to grab newly-allocated llcd.
4756 Severity : enhancement
4758 Description: 2.6 OST async journal commit and locking fix to improve performance
4759 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4760 2.6 kernels are now async as they already were in 2.4 kernels so
4761 that they can commit concurrently with the network bulk transfer.
4762 For block-allocated files the filter allocation semaphore is held
4763 to avoid filesystem fragmentation during allocation. BKL lock
4764 removed for 2.6 xattr operations where it is no longer needed.
4769 Description: lconf incorrectly determined whether two IP networks could talk
4770 Details : In some more complicated routing and multiple-network
4771 configurations, lconf will avoid trying to make a network
4772 connection to a disjoint part of the IP space. It was doing the
4773 math incorrectly for one set of cases.
4778 Description: Fix for potential infinite loop processing records in an llog.
4779 Details : If an llog record is corrupted/zeroed, it is possible to loop
4780 forever in llog_process(). Validate the llog record length
4781 and skip the remainder of the block on error.
4784 Frequency : occasional (liblustre only)
4786 Description: liblustre could not open files whose last component is a symlink
4787 Details : sysio_path_walk() would incorrectly pass the open intent to
4788 intermediate path components.
4791 Frequency : rare (liblustre only with non-standard tuning)
4792 Bugzilla : 7201 (7350)
4793 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4794 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4795 after it has been released, only do this for referenced locks
4797 Severity : enhancement
4799 Description: specifying an (invalid) directory default stripe_size of -1
4800 would reset the directory default striping
4801 Details : stripe_size -1 was used internally to signal directory stripe
4802 removal, now use "all default" to signal dir stripe removal
4803 as a directory striping of "all default" is not useful
4806 Frequency : common for large clusters running liblustre clients
4808 Description: doing an ls when liblustre clients are running is slow
4809 Details : sending a glimpse AST to a liblustre client waits for every AST
4810 to time out, as liblustre clients will not respond. Since they
4811 cannot cache data we refresh the OST lock LVB from disk instead.
4813 Severity : enhancement
4815 Description: doing an ls at the same time as file IO can be slow
4816 Details : enqueue and other "small" requests can be blocked behind many
4817 large IO requests. Create a new OST IO portal for non-IO
4818 requests so they can be processed faster.
4821 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4823 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4824 created file with mode 000
4825 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4826 allow an ftruncate() as a non-root user to a file with mode 000.
4827 Lustre now respects this flag to disable mode checks when
4828 truncating a file owned by the user
4831 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4834 Description: Revoking locks from clients that went dead or catatonic might take
4836 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4837 cancellation of such locks instant on servers without waiting for
4838 any reply from clients. Clients drops these locks when cancel
4839 notification from server is received without replying.
4842 Frequency : liblustre-only, when liblustre client dies or becomes busy
4844 Description: Doing ls on Linux clients can take a long time with active
4846 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4847 granting such locks to it in the first place if possible. Locks
4848 are taken by proxy on the OST during the read or write and
4849 dropped immediately afterward. Add connect flags handling, do
4850 not grant locks to liblustre clients for glimpse ASTs.
4852 Severity : enhancement
4854 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4855 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4856 sequential reads and behaves badly with applications doing random
4857 reads. Improve it by reading ahead at least read region, and
4858 avoiding excessive large RPC for small reads.
4860 Severity : enhancement
4862 Description: Creating more than 1000 files for a single job may cause a load
4863 imbalance on the OSTs if there are also a large number of OSTs.
4864 Details : qos_prep_create() uses an OST index reseed value that is an
4865 even multiple of the number of available OSTs so that if the
4866 reseed happens in the middle of the object allocation it will
4867 still utilize the OSTs as uniformly as possible.
4872 Description: OST or MDS may oops in ping_evictor_main()
4873 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4874 but doesn't restart at beginning of obd_exports_timed list
4877 Severity : enhancement
4879 Description: improve by-nid export eviction on the MDS and OST
4880 Details : allow multiple exports with the same NID to be evicted at one
4881 time without re-searching the exports list.
4884 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4886 Description: MDS may oops in groups_free()
4887 Details : in rare race conditions a newly allocated group_info struct is
4888 freed again, and this can be NULL. The 2.4 compatibility code
4889 for groups_free() checked for a NULL pointer, but 2.6 did not.
4892 Frequency : common for liblustre clients doing little filesystem IO
4893 Bugzilla : 9352, 7313
4894 Description: server may evict liblustre clients accessing contended locks
4895 Details : if a client is granted a lock or receives a completion AST
4896 with a blocking AST already set it would not reply to the AST
4897 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4900 Frequency : lfs setstripe, only systems with more than 160 OSTs
4902 Description: unable to set striping with a starting offset beyond OST 160
4903 Details : llapi_create_file() incorrectly limited the starting stripe
4904 index to the maximum single-file stripe count.
4907 Frequency : LDAP users only
4909 Description: lconf did not handle in-kernel recovery with LDAP properly
4910 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4912 Severity : enhancement
4914 Description: bind OST threads to NUMA nodes to improve performance
4915 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4916 node and do their allocations there to localize memory access
4918 Severity : enhancement
4920 Description: llmount can determine client NID directly from Myrinet (GM)
4921 Details : the client NID code from gmnalnid was moved directly into
4922 llmount, removing the need to use this or specifying the
4923 client NID explicitly when mounting GM clients with zeroconf
4926 Frequency : if client is started with down MDS
4928 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4929 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4930 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4931 ptlrpc_import_delayed_req() because we didn't check if the
4932 request was interrupted, and we also didn't break out of the
4933 event loop if there was no timeout
4938 Description: data loss during non-page-aligned writes to a single file from
4939 both multiple nodes and multiple threads on one node at same time
4940 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4941 inconsistency led to false short-reads, that were cached and later
4942 used by ->prepare_write() to fill in partially written page,
4943 leading to data loss.
4946 Frequency : always, if lconf --abort_recovery used
4948 Description: lconf --abort_recovery fails with 'Operation not supported'
4949 Details : lconf was attempting to abort recovery on the MDT device and not
4952 Severity : enhancement
4954 Description: remove cleanup logs
4955 Details : replace lconf-generated cleanup logs with lustre internal
4956 cleanup routines. Eliminates the need for client-cleanup and
4959 Severity : enhancement
4961 Description: add support for EAs (user and system) on lustre filesystems
4962 Details : it is now possible to store extended attributes in the Lustre
4963 client filesystem, and with the user_xattr mount option it
4964 is possible to allow users to store EAs on their files also
4966 Severity : enhancement
4968 Description: Add possibility (config option) to show minimal available OST free
4970 Details : When compiled with --enable-mindf configure option, statfs(2)
4971 (and so, df) will return least minimal free space available from
4972 all OSTs as amount of free space on FS, instead of summary of
4973 free spaces of all OSTs.
4975 Severity : enhancement
4977 Description: do not expand extent locks acquired on OST-side
4978 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4979 by server: they are not cached anyway.
4982 Frequency : when mmap is used/binaries executed from Lustre
4984 Description: Unmmap pages before throwing them away from read cache.
4985 Details : llap_shrink cache now attempts to unmap pages before discarding
4986 them (if unmapping failed - do not discard). SLES9 kernel has
4987 extra checks that trigger if this unmapping is not done first.
4992 Description: lconf didn't resolve symlinks before checking to see whether a
4993 given mountpoint was already in use
4996 Frequency : when migrating failover services
4997 Bugzilla : 6395, 9514
4998 Description: When migrating a subset of services from a node (e.g. failback
4999 from a failover service node) the remaining services would
5000 time out and evict clients.
5001 Details : lconf --force (implied by --failover) sets the global obd_timeout
5002 to 5 seconds in order to quickly disconnect, but this caused
5003 other RPCs to time out too quickly. Do not change the global
5004 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5006 Severity : enhancement
5007 Frequency : if MDS is started with down OST
5008 Bugzilla : 9439,5706
5009 Description: Allow startup/shutdown of an MDS without depending on the
5010 availability of the OSTs.
5011 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5012 Add appropriate locking and lov-osc refcounts for safe
5013 cleaning. Add osc abort_inflight calls in case the
5014 synchronize never started.
5017 Frequency : occasional (Cray XT3 only)
5019 Description: root not authorized to access files in CRAY_PORTALS environment
5020 Details : The client process capabilities were not honoured on the MDS in
5021 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5022 been accessed by an authorized user then root was able to access
5023 the file on the local client also. The root user capabilities
5024 are now allowed on the MDS, as this environment has secure UID.
5027 Frequency : occasional
5029 Description: ldiskfs "too long searching" message happens too often
5030 Details : A debugging message (otherwise harmless) prints too often on
5031 the OST console. This has been reduced to only happen when
5032 there are fragmentation problems on the filesystem.
5037 Description: Division by zero in statfs when all OSCs are inactive
5038 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5039 lov_statfs divides by value returned by lov_get_stripecnt().
5043 Bugzilla : 9489, 3273
5044 Description: First write from each client to each OST was only 4kB in size,
5045 to initialize client writeback cache, which caused sub-optimal
5046 RPCs and poor layout on disk for the first written file.
5047 Details : Clients now request an initial cache grant at (re)connect time
5048 and so that they can start streaming writes to the cache right
5049 away and always do full-sized RPCs if there is enough data.
5050 If the OST is rebooted the client also re-establishes its grant
5051 so that client cached writes will be honoured under the grant.
5056 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5057 Details : Now I/O RPCs go to different portal number and (presumably) fast
5058 lock requests (and glimses) and other RPCs get their own service
5059 threads pool that should be able to service those RPCs
5062 Severity : enhancement
5064 Description: Ability to exchange lustre version between client and servers and
5065 issue warnings at client side if client is too old. Also for
5066 liblustre clients there is ability to refuse connection of too old
5068 Details : New 'version' field is added to connect data structure that is
5069 filled with version info. That info is later checked by server and
5073 Frequency : rare, liblustre only.
5074 Bugzilla : 9296, 9581
5075 Description: Two simultaneous writes from liblustre at offset within same page
5076 might proceed at the same time overwriting eachother with stale
5078 Details : I/O lock withing llu_file_prwv was released too early, before data
5079 actually was hitting the wire. Extended lock-holding time until
5080 server acknowledges receiving data.
5083 Frequency : extremely rare. Never observed in practice.
5085 Description: avoid generating lustre_handle cookie of 0.
5086 Details : class_handle_hash() generates handle cookies by incrementing
5087 global counter, and can hit 0 occasionaly (this is unlikely, but
5088 not impossible, because initial value of cookie counter is
5089 selected randonly). Value of 0 is used as a sentinel meaning
5090 "unassigned handle" --- avoid it. Also coalesce two critical
5091 sections in this function into one.
5093 Severity : enhancement
5095 Description: allow liblustre clients to delegate truncate locking to OST
5096 Details : To avoid overhead of locking, liblustre client instructs OST to
5097 take extent lock in ost_punch() on client's behalf. New connection
5098 flag is added to handle backward compatibility.
5100 Severity : enhancement
5101 Bugzilla : 4928, 7341, 9758
5102 Description: allow number of OST service threads to be specified
5103 Details : a module parameter allows the number of OST service threads
5104 to be specified via "options ost ost_num_threads={N}" in the
5105 OSS's /etc/modules.conf or /etc/modprobe.conf.
5109 Bugzilla : 6146, 9635, 9895
5110 Description: servers crash with bad pointer in target_handle_connect()
5111 Details : In rare cases when a client is reconnecting it was possible that
5112 the connection request was the last reference for that export.
5113 We would temporarily drop the export reference and get a new
5114 one, but this may have been the last reference and the export
5115 was just destroyed. Get new reference before dropping old one.
5117 Severity : enhancement
5118 Frequency : if client is started with failover MDS
5120 Description: Allow multiple MDS hostnames in the mount command
5121 Details : Try to read the configuration from all specified MDS
5122 hostnames during a client mount in case the "primary"
5125 Severity : enhancement
5127 Description: Stop sending data to evicted clients as soon as possible.
5128 Details : Check if the client we are about to send or are sending data to
5129 was evicted already. (Check is done every second of waiting,
5130 for which l_wait_event interface was extended to allow checking
5131 of exit condition at specified intervals).
5134 Frequency : rare, normally only when NFS exporting is done from client
5136 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5138 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5141 Frequency : always, if srand() is called before liblustre initialization
5143 Description: Liblustre uses system PRNG disturbing its usage by user application
5144 Details : Introduce internal to lustre fast and high-quality PRNG for
5145 lustre usage and make liblustre and some other places in generic
5146 lustre code to use it.
5148 Severity : enhancement
5149 Bugzilla : 9477, 9557, 9870
5150 Description: Verify that the MDS configuration logs are updated when xml is
5151 Details : Check if the .xml configuration logs are newer than the config
5152 logs stored on the MDS and report an error if this is the case.
5153 Request --write-conf, or allow starting with --old_conf.
5155 Severity : enhancement
5157 Description: Handle symlinks in the path when checking if Lustre is mounted.
5158 Details : Resolve intermediate symlinks when checking if a client has
5159 mounted a filesystem to avoid duplicate client mounts.
5164 Description: lconf can hit an error exception but still return success.
5165 Details : The lconf command catches the Command error exception at the top
5166 level script context and will exit with the associated exit
5167 status, but doesn't ensure that this exit status is non-zero.
5172 Description: failure of ptlrpc thread startup can cause oops
5173 Details : Starting a ptlrpc service thread can fail if there are a large
5174 number of threads or the server memory is very fragmented.
5175 Handle this without oopsing.
5178 Frequency : always, only if liblustre and non-default acceptor port was used
5180 Description: liblustre cannot connect to servers with non-default acceptor port
5181 Details : tcpnal_set_default_params() was not called and was therefore
5182 ignoring the environment varaible TCPNAL_PORT, as well as other
5183 TCPNAL_ environment variables
5188 Description: two objects could be created on the same OST for a single file
5189 Details : If an OST is down, in some cases it was possible to create two
5190 objects on a single OST for a single file. No problems other
5191 than potential performance impact and spurious error messages.
5195 Bugzilla : 5681, 9562
5196 Description: Client may oops in ll_unhash_aliases
5197 Details : Client dcache may become inconsistent in race condition.
5198 In some cases "getcwd" can fail if the current directory is
5204 Description: Inode refcounting problems in NFS export code
5205 Details : link_raw functions used to call d_instantiate without obtaining
5206 extra inode reference first.
5210 Bugzilla : 9942, 9903
5211 Description: Referencing freed requests leading to crash, memleaks with NFS.
5212 Details : We used to require that call to ll_revalidate_it was always
5213 followed by ll_lookup_it. Also with revalidate_special() it is
5214 possible to call ll_revalidate_it() twice for the same dentry
5215 even if first occurence returned success. This fix changes semantic
5216 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5217 reference on a request referred from the intent.
5218 ll_intent_release() then releases such a request.
5221 Frequency : rare, normally benchmark loads only
5223 Description: unlinked inodes were kept in memory on the client
5224 Details : If a client is repeatedly creating and unlinking files it
5225 can accumulate a lot of stale inodes in the inode slab cache.
5226 If there is no other client load running this can cause the
5227 client node to run out of memory. Instead flush old inodes
5228 from client cache that have the same inode number as a new inode.
5231 Frequency : SLES9 2.6.5 kernel and long filenames only
5232 Bugzilla : 9969, 10379
5233 Description: utime reports stale NFS file handle
5234 Details : SLES9 uses out-of-dentry names in some cases, which confused
5235 the lustre dentry revalidation. Change it to always use the
5239 Frequency : rare, unless heavy write-truncate concurrency is continuous
5240 Bugzilla : 4180, 6984, 7171, 9963, 9331
5241 Description: OST becomes very slow and/or deadlocked during object unlink
5242 Details : filter_destroy() was holding onto the parent directory lock
5243 while truncating+unlinking objects. For very large objects this
5244 may block other threads for a long time and slow overall OST
5245 responsiveness. It may also be possible to get a lock ordering
5246 deadlock in this case, or run out of journal credits because of
5247 the combined truncate+unlink. Solution is to do object truncate
5248 first in one transaction without parent lock, and then do the
5249 final unlink in a new transaction with the parent lock. This
5250 reduces the lock hold time dramatically.
5253 Frequency : rare, 2.4 kernels only
5255 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5256 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5257 before unmounting their filesystem in order to avoid deadlock.
5258 The kernel_locked() function in 2.4 kernels only checks whether
5259 the kernel lock is held, not whether it is this process that is
5260 holding it as 2.6 kernels do.
5265 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5266 Details : The client ptlrpc code may be trying to reconnect to a down
5267 server before a previous connection attempt has timed out.
5268 Increase the reconnect interval to be longer than the connection
5269 timeout interval to avoid sending duplicate connections to
5273 Frequency : echo_client brw_test command
5275 Description: fix echo_client to work with OST preallocated code
5276 Details : OST preallocation code (5137) didn't take echo_client IO path
5277 into account: echo_client calls filter methods outside of any
5278 OST thread and, hence, there is no per-thread preallocated
5279 pages and buffers to use. Solution: hijack pga pages for IO. As
5280 a byproduct, this avoids unnecessary data copying.
5284 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5285 Description: Client can oops in mdc_commit_close() after open replay
5286 Details : It was possible for the MDS to return an open request with no
5287 transaction number in mds_finish_transno() if the client was
5288 evicted, but without actually returning an error. Clients
5289 would later try to replay that open and may trip an assertion
5290 Simplify the client close codepath, and always return an error
5291 from the MDS in case the open is not successful.
5294 Frequency : rare, 2.6 OSTs only
5296 Description: OST may deadlock under high load on fragmented files
5297 Details : If there was a heavy load and highly-fragmented OST filesystems
5298 it was possible to have all the OST threads deadlock waiting on
5299 allocation of biovecs, because the biovecs were not released
5300 until the entire RPC IO was completed. Instead, release biovecs
5301 as soon as they are complete to ensure forward IO progress.
5303 Severity : enhancement
5305 Description: Support for specifying external journal device at mount
5306 Details : If an OST or MDS device is formatted with an external journal
5307 device, this device major/minor is stored in the ext3 superblock
5308 and may not be valid for failover. Allow detecting and
5309 specifying the external journal at mount time.
5314 Description: Mounting an MDS with pending unlinked files may cause oops
5315 Details : target_finish_recovery() calls mds_postrecov() which returned
5316 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5317 considers this an error and immediately begins cleaning up the
5318 lov, just after starting the mds_lov process
5320 Severity : enhancement
5322 Description: Implement 'lfs df' to report actual free space on per-OST basis
5323 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5324 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5325 usage of objects; '-h' to report in human readable format.
5327 ------------------------------------------------------------------------------
5329 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5336 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5337 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5338 at startup during orphan unlinking. Instead, skip buddy update
5339 before setup, as it will be regenerated after recovery is complete.
5344 Description: If an OST is inactive, its locks might reference stale inodes.
5345 Details : lov_change_cbdata() must iterate over all namespaces, even if
5346 they are inactive to clear inode references from the lock.
5348 Severity : enhancement
5349 Frequency : occasional, if non-standard max_dirty_mb used
5351 Description: Client will block write RPCs if not enough grant
5352 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5353 then the client will block writes while waiting for another RPC
5354 to complete instead of consuming its dirty limit. With change
5355 we get improved performance when max_dirty_mb is small.
5357 Severity : enhancement
5358 Bugzilla : 3389, 6253
5359 Description: Add support for supplementary groups on the MDS.
5360 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5361 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5362 lookups for user supplementary groups into a cache.
5366 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5367 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5368 for liblustre O_EXCL is enforced by MDS.
5372 Description: readdir never returns NULL in liblustre.
5373 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5374 dentry in struct dirent.
5378 Frequency : liblustre only, depends on application IO pattern
5379 Description: liblustre clients evicted if not contacting servers
5380 Details : Don't put liblustre clients into the ping_evictor list, so
5381 they will not be evicted by the pinger ever.
5383 Severity : enhancement
5385 Description: Add ability to evict clients by NID from MDS.
5386 Details : By echoing "nid:$NID" string into
5387 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5388 $NID would be instantly evicted from this MDS and from all active
5389 OSTs connected to it.
5393 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5394 Details : lookup_it_finish() used to query file size from OSTs that was not
5399 Description: service threads change working directory to that of init
5400 Details : Starting lustre service threads may pin the working directory
5401 of the parent thread, making that filesystem busy. Threads
5402 now change to the working directory of init to avoid this.
5406 Frequency : during shutdown only
5407 Description: shutdown with a failed MDS or OST can cause unmount to hang
5408 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5413 Frequency : 2.6 only
5414 Description: chmod/chown may include an extra supplementary group
5415 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5416 supplementary group and if none is specified this is used.
5419 Bugzilla : 5479 (6816)
5421 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5422 Details : If lookup is in progress on a file that is unlinked we might try
5423 to revalidate the inode and fail in revalidate after lookup is
5424 complete and ll_file_open() enqueues the open again but
5425 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5428 Frequency : always, if lconf --abort_recovery used
5430 Description: lconf --abort_recovery fails with 'Operation not supported'
5431 Details : lconf was attempting to abort recovery on the MDT device and not
5434 ------------------------------------------------------------------------------
5436 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5441 Frequency : rare (only unsupported configurations with a node running as an
5443 Bugzilla : 6514, 5137
5444 Description: Mounting a Lustre file system on a node running as an OST could
5446 Details : OSTs now preallocates memory needed to write out data at
5447 startup, instead of when needed, to avoid having to
5448 allocate memory in possibly low memory situations.
5449 Specifically, if the file system is mounted on on OST,
5450 memory pressure could force it to try to write out data,
5451 which it needed to allocate memory to do. Due to the low
5452 memory, it would be unable to do so and the node would
5453 become unresponsive.
5455 Severity : enhancement
5457 Description: Addition of lconf --service command line option
5458 Details : lconf now accepts a '--service <arg>' option, which is
5459 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5461 Severity : enhancement
5463 Description: Failover mode is now the default for OSTs.
5464 Details : By default, OSTs will now run in failover mode. To return to
5465 the old behaviour, add '--failout' to the lmc line for OSTs.
5467 Severity : enhancement
5469 Description: Health checks are now provided for MDS and OSTs
5470 Details : Additional detailed health check information on MSD and OSTs
5471 is now provided through the procfs health_check value.
5474 Frequency : occasional, depends on IO load
5476 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5477 after numerous create/delete cycles
5478 Details : The ext3 inode allocation policy would not allocate new inodes
5479 very well on the OSTs because there are no new directories
5480 being created. Instead we look for groups with free space if
5481 the parent directories are nearly full.
5486 Description: Network or server problems during mount may cause partially
5487 mounted clients instead of returning an error.
5488 Details : The config llog parsing code may overwrite the error return
5489 code during mount error handling, returning success instead
5495 Description: MDS can fail to allocate large reply buffers
5496 Details : After long uptimes the MDS can fail to allocate large reply
5497 buffers (e.g. zconf client mount config records) due to memory
5498 fragmentation or consumption by the buffer cache. Preallocate
5499 some large reply buffers so that these replies can be sent even
5500 under memory pressure.
5504 Frequency : rare (liblustre)
5505 Description: fsx running with liblustre complained that using truncate() to
5506 extend the file doesn't work. This patch corrects that issue.
5507 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5508 ATTR_SIZE and lsm use in llu_setattr_raw.
5512 Frequency : rare, only 2.6 kernels
5513 Description: Unusual file access patterns on the MDS may result in inode
5514 data being lost in very rare circumstances.
5515 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5516 code in the 2.6 kernel caused the inode and/or EA data not to
5517 be read from disk, causing single-file corruption.
5521 Frequency : rare, only 2.6 filesystems using extents
5522 Description: Heavy concurrent write and delete load may cause data corruption.
5523 Details : It was possible under high-load situations to have an extent
5524 metadata block in the block device cache from a just-unlinked
5525 file overwrite a newly-allocated data block. We now unmap any
5526 metadata buffers that alias just-allocated data blocks.
5530 Frequency : filesystems with default stripe_count larger than 77
5531 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5532 Details : lconf specifies an inode size of 4096 bytes when the default
5533 stripe_count is larger than 77. This conflicts with the default
5534 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5535 this case to avoid pinning too much memory for large EAs.
5537 ------------------------------------------------------------------------------
5539 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5544 Frequency : rare (extremely heavy IO load with hundreds of clients)
5546 Description: Client is evicted, gets IO error writing to file
5547 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5548 caused clients to be evicted for AST timeouts. The fixes in
5549 bug 5192 mean we no longer need to have such short AST timeouts
5550 so ldlm_timeout has been increased.
5553 Frequency : occasional during --force or --failover shutdown under load
5554 Bugzilla : 5949, 4834
5555 Description: Server oops/LBUG if stopped with --force or --failover under load
5556 Details : a collection of import/export refcount and cleanup ordering
5557 issues fixed for safer force cleanup
5560 Frequency : only filesystems larger than 120 OSTs
5561 Bugzilla : 5990, 6223
5562 Description: lfs getstripe would oops on a very large filesystem
5563 Details : lov_getconfig used kfree on vmalloc'd memory
5566 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5567 Bugzilla : 6242, 6243
5568 Description: reading from files that had been truncated to a non-zero size
5569 but never opened returned no data
5570 Details : ll_file_read() reads zeros from no-object files to EOF
5575 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5576 Details : There's little harm in aborting MDS/OSS recovery and letting it
5577 try again, so I removed the LASSERT and return an error instead.
5579 Severity : enhancement
5581 Description: New debugging infrastructure for tracking down data corruption
5582 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5583 (b) cover more of the client-side code path, and (c) try to narrow
5584 down where problems occurred
5588 Bugzilla : 3819, 4364, 4397, 6313
5589 Description: Racing close and eviction MDS could cause assertion in mds_close
5590 Details : It was possible to get multiple mfd references during close and
5591 client eviction, leading to one thread referencing a freed mfd.
5593 Severity: : enhancement
5594 Bugzilla : 3262, 6359
5595 Description: Attempts to reconnect to servers are now more aggressive.
5596 Details : This builds on the enhanced upcall-less recovery that was added
5597 in 1.4.2. When trying to reconnect to servers, clients will
5598 now try each server in the failover group every 10 seconds. By
5599 default, clients would previously try one server every 25 seconds.
5604 Description: After recovery, certain operations trigger a failed
5605 assertion on a client.
5606 Details : Failing over an mds, using lconf -d --failover, while a
5607 client was doing a readdir() call would cause the client to
5608 LBUG after recovery completed and the readdir() was resent.
5610 Severity : enhancement
5612 Description: Default groups are now added by lconf
5613 Details : You can now run lconf --group <servicename> without having to
5614 manually add groups with lmc.
5617 Frequency : occasional
5619 Description: Nodes with an elan id of 0 trigger a failed assertion
5622 Frequency : always when accessing e.g. tty/console device nodes
5624 Description: tty and some other devices nodes cannot be used on lustre
5625 Details : file's private_data field is used by device data and lustre
5626 values in there got lost. New field was added to struct file to
5627 store fs-specific private data.
5630 Frequency : when exporting Lustre via NFS
5632 Description: NFSD failed occasionally when looking up a path component
5633 Details : NFSD is looking up ".." which was broken in ext3 directories
5634 that had grown large enough to become hashed.
5637 Frequency : Clusters with multiple interfaces not on the same subnet
5639 Description: Nodes will repeatedly try to reconnect to an interface which it
5640 cannot reach and report an error to the log.
5641 Details : Extra peer list entries will be created by lconf with some peers
5642 unreachable. lconf now validates the peer before adding it.
5645 Frequency : Only if a default stripe is set on the filesystem root.
5647 Description: Setting a default stripe on the filesystem root prevented the
5648 filesystem from being remounted.
5649 Details : The client was sending extra request flags in the root getattr
5650 request and did not allocate a reply buffer for the dir EA.
5653 Frequency : occasional, higher if lots of files are accessed by one client
5654 Bugzilla : 6159, 6097
5655 Description: Client trips assertion regarding lsm mismatch/magic
5656 Details : While revalidating inodes the VFS looks up inodes with ifind()
5657 and in rare cases can find an inode that is being freed.
5658 The ll_test_inode() code will free the lsm during ifind()
5659 when it finds an existing inode and then the VFS later attaches
5660 this free lsm to a new inode.
5664 Bugzilla : 6422, 7030
5665 Description: MDS deadlock between mkdir and client eviction
5666 Details : Creating a new file via mkdir or mknod (starting a transaction
5667 and getting the ns lock) can deadlock with client eviction
5668 (gets ns lock and trying to finish a synchronous transaction).
5671 Frequency : occasional
5672 Description: While starting a server, the fsfilt_ext3 module could not be
5674 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5675 kernels. Previously, lconf would still use the ext3 name
5676 when trying to load modules. Now, it will correctly use
5677 ext3 on 2.4 and ldiskfs on 2.6.
5679 Severity : enhancement
5680 Description: The default stripe count has been changed to 1
5681 Details : The interpretation of the default stripe count (0, to lfs
5682 or lmc) has been changed to mean striping across a single
5683 OST, rather than all available. For general usage we have
5684 found a stripe count of 1 or 2 works best.
5686 Severity : enhancement
5687 Description: Add support for compiling against Cray portals.
5688 Details : Conditional compiling for some areas that are different
5692 Frequency : occasional
5693 Bugzilla : 6409, 6834
5694 Description: Creating files with an explicit stripe count may lead to
5695 a failed assertion on the MDS
5696 Details : If some OSTs are full or unavailable, creating files may
5697 trigger a failed assertion on the MDS. Now, Lustre will
5698 try to use other servers or return an error to the
5702 Frequency : occasional
5704 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5705 caused lock timeout during readahead (since 1.4.2).
5706 Details : Processes doing readahead might match a lock that hasn't been
5707 granted yet if there are overlapping and conflicting lock
5708 requests. The readahead process waits on ungranted lock
5709 (original lock is CBPENDING), while OST waits for that process
5710 to cancel CBPENDING read lock and eventually evicts client.
5712 Severity : enhancement
5714 Description: Initial enabling of flock support for clients
5715 Details : Implements fcntl advisory locking and file status functions.
5716 This feature is provided as an optional mount flag (default
5717 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5718 locking are implemented yet, and those that are are not guaranteed
5719 to be completely correct in production environments.
5720 mount -t lustre -o [flock|noflock] ...
5723 Frequency : occasional
5725 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5726 assertion in the ext3 JBD (journaling) layer.
5727 Details : The b_committed_data struct is protected by the big kernel lock
5728 in 2.4 kernels, serializing journal_commit_transaction() and
5729 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5730 access to this struct. In 2.6 kernels there is finer grained
5731 locking to improve SMP performance of the JBD layer.
5735 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5740 Description: merge and cleanup kernel patches.
5743 -----------------------------------------------------------------------------
5745 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5747 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5748 versions of Lustre. Please update all servers and clients to
5749 version 1.4.2 or later at the same time. You must also run
5750 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5751 to update the configuration logs.
5753 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5754 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5755 - swab llog records of type '0' so we get proper header size/idx (5861)
5756 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5757 - fix rename of one directory over another leaking an inode (5953)
5758 - avoid SetPageDirty on 2.6 (5981)
5759 - don't re-add just-being-destroyed locks to the waiting list (5653)
5760 - when creating new directories, inherit the parent's custom
5761 striping settings if present parent (3048)
5762 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5763 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5764 - don't hold client locks on temporary worklist from l_lru (5666)
5765 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5766 - automatically evict dead clients (5921)
5767 - Update file size properly in create+truncate+fstat case (6196)
5768 - Do not unhash mountpoint dentries, do not allow removal of
5770 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5771 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5772 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5773 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5774 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5775 - Correctly handle reads of files with no objects (6243)
5776 - lctl recover will also mark a device active if deactivate used (5933)
5778 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5779 - allow --write-conf on an MDS with different nettype than client (5619)
5780 - don't write config llogs to MDS for mounts not from that MDS (5617)
5781 - lconf should create multiple TCP connections from a client (5201)
5782 - init scripts are now turned off by default; run chkconfig --on
5783 lustre and chkconfig --on lustrefs to use them
5784 - upcalls are no longer needed for clients to recover to failover
5786 - add --abort-recovery option to lconf to abort recovery on device
5788 - add support for an arbitrary number of OSTs (3026)
5789 - Quota support protocol changes.
5790 - forward compatibility changes to wire structs (6007)
5791 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5792 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5793 - improved "lustre status" script
5794 - initialize blocksize for non-regular files (6062)
5795 - added --disable-server and --disable-client configure options (5782)
5796 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5797 - Vanilla 2.4.29 support
5798 - increase maximum number of obd devices to 520 (6242)
5799 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5800 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5801 SLES 9 2.6.5 kernels
5802 - limit stripes per file to 160 (the maximum EA size) (6093)
5804 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5807 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5808 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5809 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5810 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5811 - small ext3 extents cleanups and fixes (5733)
5812 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5813 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5814 - avoid needless client->OST connect, fix handle mismatch (5317)
5815 - fix DLM error path that led to out-of-sync client, long delays (5779)
5816 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5817 - fix several locking issues related to i_size (5492,5624,5654,5672)
5818 - don't move pending lock onto export if it is already evicted (5683)
5819 - fix kernel oops when creating .foo in unlinked directory (5548)
5820 - fix deadlock in obdfilter statistics vs. object create (5811)
5821 - use time_{before,after} to avoid timer jiffies wrap (5882)
5822 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5823 - Do not leak request if server was not able to process it (5154)
5824 - If mds_open unable to find parent dir, make that negative lookup(5154)
5825 - don't create new directories with extent-mapping (5909, 5936)
5827 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5828 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5829 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5830 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5831 - support for RHEL4 kernel on i686 (5773)
5832 - provide error messages when incompatible logs are encountered (5898)
5834 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5835 * version 1.4.0.10 (1.4.1 release candidate 1)
5837 - don't keep a lock reference when lock is not granted (4238)
5838 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5839 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5840 - drop import inflight refcount on signal_completed_replay error (5255)
5841 - unlock page after async write error during send (3677)
5842 - handle missing objects in filter_preprw_read properly (5265)
5843 - no transno return for symlink open, don't save no-trasno open (3440)
5844 - don't try to complete elan receive that already failed (4012)
5845 - free RPC server reply state on error (5406)
5846 - clean up thread from ptlrpc_start_thread() on error (5160)
5847 - readahead could read extra page into cache that wasn't ejected (5388)
5848 - prevent races in class_attach/setup/cleanup/detach (5260)
5849 - don't dereference de->d_inode after l_dput of de (5458)
5850 - use "int" for stripe value returned from lock_to_stripe (5544)
5851 - mballoc allocation and error-checking fixes in 2.6 (5504)
5852 - block device patches to fix I/O request sizes in 2.6 (5482)
5853 - look up hostnames for IB nals (5602)
5854 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5855 - don't start multiple acceptors for the same port (5277)
5856 - fix incorrect LASSERT in mds_getattr_name (5635)
5857 - export a proc file for general "ping" checking (5628)
5858 - fix "lfs check" to not block when the MDS is down (5628)
5860 - service request history (4965)
5861 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5862 - create an "evict_client" /proc entry on OSTs, like the MDS has
5863 - fix mount usage message, return errors per mount(8) (5168)
5864 - change grep [] to grep "[]" in tests so they work in more UMLs
5865 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5866 - remove ancient LOV_MAGIC_V0 EA support (5047)
5867 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5868 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5869 - allow readahead allocations to fail when low on memory (5383)
5870 - mmap locking landed again, after considerable improvement (2828)
5871 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5873 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5876 - send OST transaction number in read/write reply to free req (4966)
5877 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5878 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5879 - the watchdog thread now runs as interruptible (5246)
5880 - flock/lockf fixes (but it's still disabled, pending 5135)
5881 - don't use EXT3 constants in llite code (5094)
5882 - memory shortage at startup could cause assertion (5176)
5884 - reorganization of lov code
5885 - single portals codebase
5887 - add extents/mballoc support (5025)
5888 - direct I/O reads in the obdfilter (4048)
5889 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5891 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5894 - send OST transaction number in read/write reply to free req (4966)
5895 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5896 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5897 - the watchdog thread now runs as interruptible (5246)
5898 - handle missing objects in filter_preprw_read properly (5265)
5899 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5900 - drop import inflight refcount on signal_completed_replay error (5255)
5901 - unlock page after async write error during send (3677)
5902 - return original error code on reconstructed replies (3761)
5903 - no transno return for symlink open, don't save no-trasno open (3440)
5905 - add pid to ldlm debugging output (4922)
5906 - bump the watchdog timeouts -- we can't handle 30sec yet
5907 - extra debugging for orphan dentry/inode bug (5259)
5909 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5913 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5914 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5915 - fix NULL dereference in /proc/sys/portals/routes (4827)
5916 - allow failed mdc_close() operations to be interrupted (4561)
5917 - stop precreate on OST before MDS would time out on it (4778)
5918 - don't send partial-page writes before EOF from client (4410)
5919 - discard client grant for sub-page writes on large-page clients (4520)
5920 - don't free dentries not owned by NFS code, check generation (4806)
5921 - fix lsm leak if mds_create_objects() fails (4801)
5922 - limit debug_daemon file size, always print CERROR messages (4789)
5923 - use transno after validating reply (3892)
5924 - process timed out requests if import state changes (3754)
5925 - update mtime on OST during writes, return in glimpse (4829)
5926 - add mkfsoptions to LDAP (4679)
5927 - use ->max_readahead method instead of zapping global ra (5039)
5928 - don't interrupt __l_wait_event() during strace
5930 - add software watchdogs to catch hung threads quickly (4941)
5931 - make lustrefs init script start after nfs is mounted
5932 - fix CWARN/ERROR duplication (4930)
5933 - return async write errors to application if possible (2248)
5934 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5935 - print NAL number in %x format (4645)
5936 - update barely-supported suse-2.4.21-171 series (4842)
5937 - support for sles 9 %post scripts
5938 - support for building 2.6 kernel-source packages
5939 - support for sles km_* packages
5941 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5944 - ignore -ENOENT errors in osc_destroy (3639)
5945 - notify osc create thread that OSC is being cleaned up (4600)
5946 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5947 - reconstruct ost_handle() like mds_handle() (4657)
5948 - create a new thread to do import eviction to avoid deadlock (3969)
5949 - let lconf resolve symlinked-to devices (4629)
5950 - don't unlink "objects" from directory with default EA (4554)
5951 - hold socknal file ref over connect in case target is down (4394)
5952 - allow more than 32000 subdirectories in a single directory (3244)
5953 - fix blocks count for O_DIRECT writes (3751)
5954 - OST returns ENOSPC from object create when no space left (4539)
5955 - don't send truncate RPC if file size isn't changing (4410)
5956 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5957 - bind to privileged port in socknal and tcpnal (3689)
5959 - rate limit CERROR/CWARN console message to avoid overload (4519)
5960 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5961 - basic mmap support (3918)
5962 - kernel patch series update from b1_4 (4711)
5964 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5967 - avoid crash during MDS cleanup with OST shut down (2775)
5968 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5969 - don't use bad inodes on the MDS (3744)
5970 - dynamic object preallocation to improve recovery speed (4236)
5971 - don't hold spinlock over lock dumping or change debug flags (4401)
5972 - don't zero obd_dev when it is force cleaned (3651)
5973 - print grants to console if they go negative (4431)
5974 - "lctl deactivate" will stop automatic recovery attempts (3406)
5975 - look for existing locks in ldlm_handle_enqueue() (3764)
5976 - don't resolve lock handle twice in recovery avoiding race (4401)
5977 - revalidate should check working dir is a directory (4134)
5979 - don't always mark "slow" obdfilter messages as errors (4418)
5981 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5984 - don't close LustreDB during write_conf until it is done (3860)
5985 - fix typo in lconf for_each_profile (3821)
5986 - allow dumping logs from multiple threads at one time (3820)
5987 - don't allow multiple threads in OSC recovery (3812)
5988 - fix debug_size parameters (3864)
5989 - fix mds_postrecov to initialize import for llog ctxt (3121)
5990 - replace config semaphore with spinlock (3306)
5991 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5992 - don't allow enqueue to complete on a destroyed export (3822)
5993 - down write_lock before checking llog header bitmap (3825)
5994 - recover from lock replay timeout (3764)
5995 - up llog sem before sending rpc (3652)
5996 - reduce ns lock hold times when setting kms (3267)
5997 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5998 - fix NULL deref and obd_dev leak on setup error (3312)
5999 - replace some LBUG about llog ops with error handling (3841)
6000 - don't match INVALID dentries from d_lookup and spin (3784)
6001 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6002 - fix invalid assertion in ptlrpc_set_wait (3880)
6004 - add libwrap support for the TCP acceptor (3996)
6005 - add /proc/sys/portals/routes for non-root route listing (3994)
6006 - allow setting MDS UUID in .xml (2580)
6007 - print the stack of a process that LBUGs (4228)
6009 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6012 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6013 - make sure to unset replay flag from failed open requests (3440)
6014 - if default stripe count is 0, use OST count for inode size (3636)
6015 - update parent mtime/ctime on client for create/unlink (2611)
6016 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6017 - free recovery state on server during a forced cleanup (3571)
6018 - unregister_reply for resent reqs (3063)
6019 - loop back devices mounting and status check on 2.6 (3563)
6020 - fix resource-creation race that can provoke i_size == 0 (3513)
6021 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6022 - more debugging for page-accounting assertion (3746)
6023 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6024 - avoid deadlock after precreation failure (3758)
6025 - fix race and lock order deadlock in orphan handling (3450, 3750)
6026 - add validity checks when grabbing inodes from l_ast_data (3599)
6028 - add /proc/.../recovery_status to obdfilter (3428)
6029 - lightweight CDEBUG infrastructure, debug daemon (3668)
6030 - change default OSC RPC parameters to be better on small clusters
6031 - turn off OST read cache for files smaller than 32MB
6032 - install man pages and include them in rpms (3100)
6033 - add new init script for (un)mounting lustre filesystems (2593)
6034 - run chkconfig in %post for init scripts (3701)
6035 - drop scimac NAL (unmaintained)
6037 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6040 - clean kiobufs before and after use (3485)
6041 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6042 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6043 - change init script to fail more clearly if not run as root (1528)
6044 - allow clients to reconnect during replay (1742)
6045 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6046 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6047 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6048 - fix cond_resched (3554)
6049 - only evict unfinished clients after recovery (3515)
6050 - allow bulk resend, prevent data loss (3570)
6051 - dynamic ptlrpc request buffer allocation (2102)
6052 - don't allow unlinking open directory if it isn't empty (2904)
6053 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6054 - remove extraneous obd dereference causing LASSERT failure (3334)
6055 - don't use get_cycles() when creating temp. files on the mds (3156)
6056 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6057 - handle EEXIST for set-stripe, set proper directory name (3336)
6059 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6060 - fix ksocknal_fmb_callback() error messages (2918)
6062 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6065 - don't copy lvb into (possibly NULL) reply on error (2983)
6066 - don't deref dentry after dput, don't free lvb on error (2922)
6067 - use the kms to determine writeback rpc length (2947)
6068 - increment oti_logcookies when osc is inactive (2948)
6069 - update client's i_blocks count via lvb messages (2543)
6070 - handle intent open/close of special files properly (1557)
6071 - mount MDS with errors=remount-ro, like obdfilter (2009)
6072 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6073 - don't use cancelling-locks' kms values (2947)
6074 - use highest lock extent for kms, not last one (2925)
6075 - don't dereference ERR_PTR() dentry in error handling path (3107)
6076 - fix thread race in portals_debug_dumplog() (3122)
6077 - create lprocfs device entries at setup instead of at attach (1519)
6078 - common AST error handler, don't evict client on completion race (3145)
6079 - zero nameidata in detach_mnt in 2.6 (3118)
6080 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6081 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6082 - initialize RPC timeout timer earlier for 2.6 (3219)
6083 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6084 - print nal/nid for unknown nid (3258)
6085 - additional checks for oscc recovery before doing precreate (3284)
6086 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6087 - don't crash in mdc_close for bad permissions on open (3285)
6088 - zero i_rdev for non-device files (3147)
6089 - clear page->private before handing to FS, better assertion (3119)
6090 - tune the read pipeline (3236)
6091 - fix incorrect decref of invalidated dentry (2350)
6092 - provide read-ahead stats and refine rpc in flight stats (3328)
6093 - don't hold journal transaction open across create RPC (3313)
6094 - update atime on MDS at close time (3265)
6095 - close LDAP connection when recovering to avoid server load (3315)
6096 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6097 - don't leak open file on MDS after open resend (3325)
6098 - serialize filter_precreate and filter_destroy_precreated (3329)
6099 - loop device shouldn't call sync_dev() for nul device (3092)
6100 - clear page cache after eviction (2766)
6101 - resynchronize MDS->OST in background (2824)
6102 - refuse to mount the same filesystem twice on same mountpoint (3394)
6103 - allow llmount to create routes for mounting behind routers (3320)
6104 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6105 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6106 - fix rare problem with rename on htree directories (3417)
6108 - allow default OST striping configuration per directory (1414)
6109 - fix compilation for qswnal for 2.6 kernels (3125)
6110 - increase maximum number of MDS request buffers for large systems
6111 - change liblustreapi to be useful for external progs like lfsck (3098)
6112 - increase local configuration timeout for slow disks (3353)
6113 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6115 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6118 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6119 - don't overwrite extent policy data in reply if lock was blocked (2901)
6120 - drop filter export grants atomically with removal from device (2663)
6121 - del obd_self_export from work_list in class_disconnect_exports (2908)
6122 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6123 - swab reply message in mdc_close, other PPC fixes (2464)
6124 - fix destroying of named logs (2325)
6125 - overwrite old logs when running lconf --write_conf (2264)
6126 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6127 - fix race in target_handle_connect (2898)
6128 - mds_reint_create() should take same inode create lock (2926)
6129 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6130 - don't close files for self_export to avoid uninitialized obd (2936)
6131 - allow MDS with the same name as client node (2939)
6132 - hold dentry reference for closed log files for unlink (2325)
6133 - reserve space for all logs during transactions (2059)
6134 - don't evict page beyond end of stripe extent (2925)
6135 - don't oops on a deleted current working directory (2399)
6136 - handle hard links to targets without a parent properly (2517)
6137 - don't dereference NULL lock when racing during eviction (2867)
6138 - don't grow lock extents when lots of conflicting locks (2919)
6140 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6143 - account for cache space usage on clients to avoid data loss (974)
6144 - lfsck support in lustre kernel code (2349)
6145 - reduce journal credits needed for BRW writes (2370)
6146 - orphan handling to avoid losing space on client/server crashes
6147 - ptlrpcd can be blocked, stopping ALL progress (2477)
6148 - use lock value blocks to assist in proper KMS, faster stat (1021)
6149 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6150 - recovery for initial connections (2355)
6151 - fixes for mds_cleanup_orphans (1934)
6152 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6153 - block all file creations until orphan recovery completes (1901)
6154 - client remove rq_connection from request struct (2423)
6155 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6156 - recovery timer race (2670)
6157 - mdc_close recovey bug (2532)
6158 - ptlrpc cleanup bug (2710)
6159 - mds timeout on local locks (2588)
6160 - namespace lock held during RPCs (2431)
6161 - handle interrupted sync write properly (2503)
6162 - don't try to handle a message that hasn't been replied to (2699)
6163 - client assert failure during cleanup after abort recovery (2701)
6164 - leak mdc device after failed mount (2712)
6165 - ptlrpc_check_set allows timedout requests to complete (2714)
6166 - wait for inflight reqs when ptlrpcd finishes (2710)
6167 - make sure unregistered services are removed from the srv_list
6168 - reset bulk XID's when resending them (caught by 1138 test)
6169 - unregister_bulk after timeout
6170 - fix lconf error (2694)
6171 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6172 - readahead locks pages, leaves pending causing memory pressure (2673)
6173 - increase OST request buffers to 4096 on large machines (2729)
6174 - fix up permission of existing directories in simple_mkdir (2661)
6175 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6176 - don't assign transno to errored transactions (2742)
6177 - don't delete objects on OST if given a bogus objid from MDS (2751)
6178 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6179 - if rq_no_resend, then timeout request after recovery (2432)
6180 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6181 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6182 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6183 - MDS recovery shouldn't delete objects that it has given out (2730)
6184 - if enqueue arrives after completion, don't clobber LVB (2819)
6185 - don't unlock pages twice when trigger_group_io returns error (2814)
6186 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6187 - don't write pages to disk if there was an error (1450)
6188 - don't ping imports that have recovery disabled (2676)
6189 - take buffered bytes into account when balancing socknal conn (2817)
6190 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6191 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6192 - remove little-endian swabbing of llog records (1987)
6193 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6194 - retry reposting request buffers if they fail (1191)
6195 - grow extent at grant time to avoid granting a revoked lock (2809)
6196 - lock revoke doesn't evict page if covered by a second lock (2765)
6197 - disable VM readahead to avoid reading outside lock extents (2805)
6199 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6200 - updated kernel patches for hp-2.4.20 kernel (2681)
6202 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6205 - fix truncated write corruption (2366)
6206 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6208 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6209 - ensure len(uuid) < 37 in lmc (1171)
6210 - fix ia64 OOPS in llog_test (2255)
6211 - zero end of page at obdfilter for partial page writes (2648)
6212 - don't leave stale dentries around after renames (bug 2428)
6213 - fix timeouts when evicting a client with a single lock held (2642)
6214 - set deadline for the initial HELLO message to drain (2634)
6215 - print out dotted-quad IP addresses in the socknal (2302)
6217 - additional debugging for MDS client eviction problem (2443)
6218 - fix mkfsoptions support for osts (2603, 2604)
6220 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6223 - add series for the vanilla 2.6.0 kernel
6224 - add series for the vanilla 2.4.24 kernel
6225 - add series for a cray x86/64 UL kernel drop
6226 - fix xattr patches for the vanilla 2.4.19 series
6228 - generate true UUIDs in lmc (1171)
6229 - have portals stack dumping break in UML (2466)
6230 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6231 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6232 - wake callback waiting threads on client eviction (2460)
6233 - Add --ptldebug and --subsystem to lmc (1719)
6234 - update assertion to allow safe interrupt allocation
6235 - set rq_no_resend for cancel requests (2432)
6236 - recalculate ptlrpcd timeout after resend (2494)
6237 - call vfs_rmdir when removing pending directories (2368)
6238 - fix renaming a file to itself (2429)
6239 - lmc creates a default one-stripe lov (2454)
6240 - expand procfs space to handle large clusters (2326)
6241 - increase UML stack to avoid overflow
6242 - update lconf's list of debug and subsystem masks
6243 - fix lfs find --obd (2510)
6244 - /proc tunable for disabling filter read caching (2591)
6245 - stop rpm packages from altering slapd.conf (2301)
6246 - disable nagle in the socknal under 0conf (2578)
6247 - choose mds inode size based on stripe count (2572)
6248 - fix kernel-source rpm problems (2516)
6250 - add --disable-doc to avoid pdf generation (2421)
6251 - update documentation, tests, type-os, comments
6252 - avoid format warnings on ia64
6253 - remove the TOE NAL
6254 - tiny code cleanups by removing unused fields
6256 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6259 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6260 - remove the most blatant lies from BUILDING file (2371)
6261 - change default debug level to reasonable production setting
6262 - reduce client side cache size to reduce cache flush time
6263 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6264 - make TCP zerocopy and pinger support enabled by default (2476)
6265 - sync writes completed after process exits caused crashes (2319)
6266 - maintain correct mount count on the MDS (2356)
6267 - backout 1557, because 2316 wasn't really fixed
6268 - better file I/O statistics gathering in /proc
6269 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6270 - another kernel patch to fix zero-copy TCP function export
6271 - don't take duplicate lock when processing re-sent getattr (2420)
6272 - lctl uses obd_self_export instead of creating new conn (2353)
6273 - MDS/OST recovery case which requires object creation asserted (2425)
6274 - move lfs from /usr/sbin to /usr/bin in packages
6275 - fix race between mds_client_add and mds_client_free (2417)
6276 - use kmalloc instead of slabs in portals (2430)
6277 - don't create duplicate records when a failover MDS is present (2442)
6278 - remove unnecessary mount age check (2332)
6279 - don't remove directory inodes from locks prematurely (2451)
6280 - don't break if MDS service name is the same as hostname (2103)
6281 - fix races in client write RPC generation when cache full (2482)
6283 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6286 - remove now-unused request->rq_obd (278)
6287 - if an allocation fails, print out how much memory we've used (1933)
6288 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6289 - add the "configurable stack size" patch to most series files (1256)
6290 - ability to write large log records, for 100+ OST configs (2306)
6291 - fix NULL deref when filter_prep fails (2314)
6292 - fix operator precedence error in filter_sync
6293 - dynamic allocation of socknal TX descriptors (2315)
6294 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6295 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6296 - ensure that CWARN messages in Portals always get to the syslog
6297 - __init/__exit are not for prototype decls (ldlm_init/exit)
6298 - x86-64 compile warning fixes
6299 - fix gateway LMC keyword conflict (2318)
6300 - fix MDS lock inversions in getattr/reint paths (1844)
6301 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6302 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6303 - DLM race condition prevented some lock evictions (2328)
6304 - ENOMEM detection and retry on socknal sends (2230)
6305 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6306 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6307 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6308 - better error messages when a client is rejected during recovery (1505)
6309 - avoid cancelling locks which were never granted, after failure (2330)
6310 - fix i_sem/journal inversion in mds_client_add (2333)
6311 - fix truncate/getattr lock cycle deadlock (2334)
6312 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6313 - fix two rare exit paths which could leak an l_lock() ref (2321)
6314 - fencepost error in MDS/OST orphan recovery (2226)
6315 - make log record alignment 8 bytes (1988)
6316 - lstripe now fails when requested offset > ost_count (2237)
6317 - ensure that all kernel series have a complete list.h (1607)
6318 - fix crashes in special-file operations (2316)
6319 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6321 - allow configurable automake binary, for testing new versions
6322 - small update to the lfs documentation
6324 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6326 * fix negative export reference count in fsfilt_sync (2312)
6328 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6329 * release candidate 0.9.1
6331 - orphans are moved into the PENDING directory for possible recovery
6332 - replayed opens now open by fid for orphan/rename safety (1042)
6333 - last close of an orphan inode generates a transno (683)
6334 - chdir() and mount() now pin the directory entry (1020)
6335 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6336 - discard very old requests without processing them (1502)
6337 - remove some common, well-understood CERRORs (1505)
6338 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6339 - clear "grant" flags in OST replies until OST grant code lands (1644)
6340 - fix read performance by not clobbering i_blksize on client (1598)
6341 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6342 - make LRU size a /proc tunable, clears locks when reduced (707)
6343 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6344 - server locks take references on exports now (1558)
6345 - build fixes for 2.4.20-rh trees (1663)
6346 - return an error from lov_create if all OSCs are inactive (1751)
6347 - fix import levels when a reconnect happens without a timeout (1597)
6348 - exit early from mds_open if we get a lookup error (1749)
6349 - partial page read at EOF wouldn't wait for disk before sending (1642)
6350 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6351 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6352 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6353 - write/truncate lock inversion (1639)
6354 - Don't auto-load obdclass, portals modules during cleanup (1495)
6355 - fix timestamps from jumping to "now" (1763)
6356 - extra journal assertions (1648)
6357 - add an extra multiunlink test (1771)
6358 - fix read_record/write_record API (1776)
6359 - fix leak of offset_extent, possible incorrect i_size later (1772)
6360 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6361 - Debugging for the kqswnal_get_idle_tx problems (1820)
6362 - Allow recovery to be attempted multiple times (1536)
6363 - Write out MDS last_rcvd file after it is first created (1600)
6364 - Fix tx_descriptor leak in failed transmit situations (1827)
6365 - ext3 journaling fixes for assertion failure after IO error (1871)
6366 - class_export_put() on freed export after completion AST error (1896)
6367 - Fix revalidate looping in VFS (1322)
6368 - Don't access a freed export during MDS_REINT timeout (1521)
6369 - Add open-unlink recovery support on the MDS (1673,1764)
6370 - Return an error if no MDS data was read from last_rcvd (1946)
6371 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6372 - Don't setup a disk device that doesn't match exported UUID (317)
6373 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6374 - avoid committing NULL handle in force close
6375 - local.sh is now a one-stripe LOV configuration
6376 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6377 - fix ext3 htree duplicate directory entry corruption (1516)
6378 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6379 - update to new LOV EA format (2097)
6380 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6382 2003-06-15 Phil Schwan <phil@clusterfs.com>
6385 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6386 - per-import/export recovery handling (958, 931, 959)
6387 - multiple last-rcvd slots, for serving multiple FSes (949)
6388 - connections are again shared between multiple imp/exports (963, 964)
6389 - "umount -f" would hang if any requests needed to be sent (393, 978)
6390 - avoid pinning large req buffer by copying for queued messages (989)
6391 - add "uuid" to "lctl device" command to help upcalls (991)
6392 - "open" RPCs with transnos would confuse recovery counters (1037)
6393 - do proper endian conversion of all wire messages (288, 340, 891)
6394 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6395 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6396 - fix ptlrpc_connection leak in target_handle_connect (1174)
6397 - fix import refcounting bug in OST and MDS cleanup (1134)
6398 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6399 - fix very unlikely obd_types race condition (501)
6400 - remove osc_open hack for echo_client (1187)
6401 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6402 - a failure in read_inode2 leads to deadlock (1139)
6403 - cancel ack-locks as soon as transaction is committed (1072)
6404 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6405 - make sure to commitrw after any preprw to avoid deadlock (1162)
6406 - failing to execute a file in a lustre FS would lock inode (1203)
6407 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6408 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6409 - obd_uuid2tgt would walk off the end of the list (1255)
6410 - on IA64 the osc would give portals incorrect bulk size (1258)
6411 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6412 - fix lock inversion caused by new llite matching code (1282)
6413 - limit the number of dirty pages on a client to 10MB (1286)
6414 - timed out locks were not being corrected cancelled (1289)
6415 - fix O_DIRECT above 4GB on IA-32 (1292)
6416 * major user-visible changes
6417 - fail out/fail over policy now controlled by the upcall (993)
6419 - add OBD_PING to check server availability and failure (954)
6420 - lustre messages are now sent in sending host order (288, 340, 891)
6421 - add eadatalen to MDS getattr reply (340)
6422 - OST read replies may contain second buffer, with per-page status (593)
6424 2003-03-11 Phil Schwan <phil@clusterfs.com>
6427 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6428 - failed open()s could cause deadlock; fixed (867, 869)
6429 - stop cancelling OST locks when files are closed (481)
6430 - overlapping XID spaces caused network corruption (851, 853)
6431 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6432 - setattr_raw added, to do single-RPC, server-side setattrs
6433 - lmc/lconf syntax change for OST UUIDs
6434 - fix crashy race condition between ptlrpc_free_req and osc_close
6435 - don't use request in mdc_enqueue if we hit a timeout (889)
6436 - don't set the inode i_size for regular files from the MDS (896)
6437 - handle out of order completion AST (842)
6438 - don't LBUG if a lock request times out after receiving AST (913)
6439 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6440 - if a bad lock AST arrives, send an error instead of dropping entirely
6441 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6442 - fix leak in bulk IO when only partially completed (899, 900, 926)
6443 - fix O_DIRECT for ia64 (55)
6444 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6445 - C-z after timeout could hang a process forever; fixed (977)
6447 - client-side I/O cache (678, 924, 929, 941, 970)
6449 - READPAGE and SETATTRs which don't take server-side locks get
6452 2003-02-11 Phil Schwan <phil@clusterfs.com>
6455 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6456 - Fix fsfilt_extN_readpage() to read a full page of directory
6457 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6458 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6459 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6460 - Most intent operations were removed, in favour of a new RPC mode
6461 that does a single RPC to the server and bypasses most of the VFS
6462 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6463 - Aggressively cancel local locks on DLM servers
6464 - mds_reint_unlink sends EA to the client if it's the last nlink.
6465 client uses that EA to unlink OST objects.
6466 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6467 - recursive symlinks were fixed (439)
6468 - fixed NULL deref in DEBUG_REQ
6469 - filter_update_lastobjid no longer calls sync, which annoyed extN
6470 - fixed multi-client small-writes to a single file problem (445)
6471 - fixed mtime updates during file writes (607)
6472 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6473 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6474 - fixed Linux OST size reporting problem (444, 656)
6475 - OST now updates object mtime with writes or setattr (607, 619)
6476 - client verifies file size before zeroing page past EOF (445)
6477 - OST now writes last allocated objid to disk with allocation (108)
6478 - LOV on echo now works (409)
6480 - mds_reint_unlink sends a new buffer, with the EA included. this
6481 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6482 set if a regular file was being unlinked, and it was the last link
6483 - use PtlGet from the target for bulk writes (315)
6484 - OST now updates object mtime with writes or setattr (607, 619)
6485 - LDLM now has a grant-time callback to revalidate locked items, if
6487 - Many MDS operations were reorganized to combat race conditions
6489 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6490 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6492 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6495 - Fully reactivate OST imports after reconnection (512, others)
6496 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6497 - More graceful error handling for truncating on dead OST (515)
6498 - Don't error out unless we're actually accessing dead stripes (474)
6499 - Fix garbage sizes when stripes are missing (410)
6500 - LRU counters were broken, causing constant lock purge (433, 432)
6501 - garbage on read from stripes with failed OSTs (441)
6502 - mark OSCs as active before reconnecting during recovery (438)
6503 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6504 - lfind did not preserve OST order in output (443)
6505 - symlinks cause hung clients, incorrect data (439)
6506 - stop dereferencing request after dropping refcount (457)
6507 - don't LASSERT(spin_is_locked) on non-SMP (455)
6508 - fixes for many rename() bugs
6509 - fstat didn't correctly synchronize attributes (399)
6510 - server must handle lock cancellation during blocking AST prep (487)
6511 - bulk descriptors were free()d too soon (511)
6512 - fix paths in lconf, which would load incorrect modules (451, 507)
6513 - fix confusing lconf 'host not found' error message (386)
6514 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6515 - fix race condition in mdc_blocking_ast() for inode access (526)
6516 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6517 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6518 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6519 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6520 - fix lstripe in several ways (kernel side) (527)
6521 - fix request leak in ldlm_cli_enqueue (262)
6522 - incorrect OSC was marked inactive after OST failure
6523 - call mds_fs_cleanup before unmounting filesystem (524)
6524 - fix races between taking ns_lock and ldlm_lock_change_resource
6525 - fix races updating LOV export open file list
6526 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6527 - fix recovery NULL deref in ldlm_cli_cancel_unused
6528 - fix some DLM races by using new hash table for lock handles (419)
6529 - permit the client to specify desired inodes, at replay
6530 - duplicate requests when we queue them for replay reintegration
6531 - fix last_rcvd offset calculation
6532 - sync after each recovered transaction, so we always make progress
6533 - never, not always, ERESTART requests without transnos
6534 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6536 - skip replay if the MDS says that the client is already connected
6537 - don't check for a recovery-enabled export to match lctl's UUID
6538 - don't INC_USE_COUNT for phantom exports
6539 - don't crash when cleaning up phantom exports (567)
6540 - don't double-finish or set replay data for errored mdc_open requests
6541 - abort requests when they time out, so we don't get old replies
6542 - send/receive replies for AST messages again
6543 - if the client says that it doesn't have the lock, cancel it on the
6545 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6546 mark it as destroyed, it will all work out when decref is called
6547 - fix module use counts (22, 581)
6549 - ASTs now expect a reply (server cancels lock on error reply)
6551 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6554 - fix many simultaneous client startup (392)
6555 - fix dentry->d_it clobbering
6556 - credentials weren't being shipped for readdir/getattr operations
6557 - remove invalid assertions triggered during some concurrent MD
6559 - proper Lustre versions added (336, 389)
6560 - fix memory leak for create error case (398)
6561 - fix LOV locking bug that would get cli/srv out of sync
6562 - fix echo client over LOV (409)
6563 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6564 - fix double-O_EXCL intent crash (424)
6565 - avoid sending multiple lock CANCELs (352)
6567 - MDS can do multi-client recovery (modulo bugs in new code)
6569 - many updates, edits, cleanups
6571 2002-11-18 Phil Schwan <phil@clusterfs.com>
6574 - fix null d_it dereference (346)
6575 - fix full OST/dbench hang (333)
6576 - fix permission problem with file removal (286)
6577 - fix removal of OSCs from LOV when they fail
6578 - fix NULL deref during bulk timeout (214)
6579 - fix problems related to multiple filesystems on one MDS (241)
6580 - fixed serious subtle metadata locking bugs
6581 - free locks on clients when inodes are removed due to memory
6583 - fix inode pointer in lock data (285)
6584 - partial support for multiple MDS on a single host (241)
6585 - data locks weren't cancelled at clear_inode time (290, 311)
6586 - intent locks could lead to unbounded lock growth (205)
6587 - added a maximum lock count, an LRU list, and a flusher
6588 - fix multiple rename (365)
6589 - properly abstracted the echo client
6590 - OSC locked 1 byte too many; fixed
6591 - rewrote brw callback code:
6592 - fixed recovery bugs related to LOVs (306)
6593 - fixed too-many-pages-in-one-write crash (191)
6594 - fixed (again) crash in sync_io_timeout (214)
6595 - probably fixed callback-related race (385)
6597 - Add capability to MDS protocol
6598 - LDLM cancellations and callbacks on different portals
6600 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6603 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6604 - defer open object destruction to close time (601981, 138)
6605 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6606 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6607 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6608 - don't call obd_setattr in ll_file_release for destroyed objects
6609 * protocol change to lustre_msg: move |version| and add |flags|
6610 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6611 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6612 * added replay of create, unlink, link and rename operations during
6613 MDS failover; recovery should be much more robust now
6614 * remove failed OSCs from LOVs (only lov_create uses this so far)
6615 * the lustre-HOWTO was brought (more) up to date (582544)
6617 2002-10-23 Phil Schwan <phil@clusterfs.com>
6620 - in-use dentries weren't being reused properly (617851)
6621 - prevent multiple LDLM setup (599178)
6622 - fix LOV size calculations for truncate (617853)
6623 - fix client handling of MDS intent errors (POSIX)
6624 - fix permission bug in lovstripe.c test (624321)
6625 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6626 - truncate past end of file could corrupt data
6627 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6628 - a race in recovery could return ETIMEDOUT to apps (623947)
6629 - building outside the source directory was fixed
6630 * the lustre-HOWTO was brought (more) up to date (582544)
6631 * major progress was made on recovery functionality
6633 2002-10-10 Phil Schwan <phil@clusterfs.com>
6636 - recovery deadlock fix
6637 - rm -rf causes LBUG fix (617817)
6638 - file open by multiple tasks fix (618962)
6639 - directory permissions bugs (602707 and 620007)
6640 - journal_stop fixed with locking (611313)
6641 - O_APPEND failures resolved (618273, perhaps 614459)
6642 - lconf PATH fix (619770)
6643 - IA64 build fix (621450)
6644 - RPC buffer sizes scale with amount of memory
6646 2002-10-01 Phil Schwan <phil@clusterfs.com>
6649 - locks would be cancelled without throwing away data pages,
6650 resulting in inconsistent data (605627)
6651 - inode attributes were not always being refreshed (605627, 612449)
6652 - lconf now continues to cleanup after lctl reports an error
6653 - MDS now enforces user permissions (602707)
6654 - lprocfs cleanup fixed, but not yet enabled (614157)
6655 - fixed infinite server hang, should a client not respond to an AST
6656 - avoid going into recovery if user calls readlink() with a buffer
6657 that's too small (613941)
6658 - AST RPCs no longer require replies (614867) -- this may be changed
6659 - don't crash server if client sends an IOV that's too big (611336)
6660 - fixed lock conversion deadlock (611892)
6661 - fixed the following of symlinks (614622)
6662 * recovery: the server can remove locks from a client that dies, other
6663 clients can make progress
6664 * more extN patch fixes
6665 * compile-time configurable ptlrpc buffer allocations
6667 - collaborative read cache document
6668 - Lustre Lite Performance CDR document-in-progress
6670 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6673 - fix typo in patch-2.4.18
6675 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6678 - clear ptlrpc request each time in handle_incoming_request()
6679 - unlink of files now destroys the object on the OST
6681 2002-09-19 Peter Braam <braam@clusterfs.com>
6683 * add hard link support
6684 * change obdfile creation method
6685 * kernel patch changed
6687 2002-09-19 Peter Braam <braam@clusterfs.com>
6690 - stack overflow bug in extN fixed
6692 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6694 * documentation updates
6695 - add man pages for config tools
6696 - update tests/README to describe testing with new config tools
6697 - finish metadata API descriptions
6698 * bug fixes and cleanups
6699 - statfs workaround for 16TB limit
6700 - LOV stripe allocation improved, can stripe on subset of OSTs
6701 - LOV file size/IO offset was wrong for files > 4GB in size
6702 - object EA data was being dropped, caused files to be unreadable
6703 - memory overflow with non-LOV OST caused memory corruption
6704 - fixed regression tests to work with new config tools, obdfilter
6705 - fixed bug when directory size became larger than 1 block
6706 - fixed bug (for single client case) when PWD was deleted
6707 - invalidate local directory pages when doing intent-based ops
6708 - avoid LDLM oops when lock callback contained bad data
6710 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6712 * documentation updates
6713 * bug fixes and cleanups
6714 - configuration tools
6717 - 64-bit compile warnings
6718 - 64-bit internal statfs data
6720 * test_brw on persistent OST devices
6722 * lprocfs (disabled)
6724 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6726 * documentation updates
6727 * bug fixes and cleanups
6728 * configuration tools
6730 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6733 * many small fixes to 0_5_4
6734 * io/network handling
6735 * thinkos in MDS operations
6737 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6740 * crucial basic fixes to 0.5.3
6741 * IOR, Iozone work over Elan
6744 2002-08-07 Phil Schwan <phil@clusterfs.com>
6745 * version 0_5_3, our first alpha
6746 * we use the new Portals iovs
6747 * documentation updates
6748 * bug fixes and cleanups
6749 * small changes in the DLM wire protocol
6751 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6752 * version 0_5_1 with some initial stability,
6753 * locking on MD and file I/O.
6754 * documentation updates
6755 * several bug fixes since 0.5.0
6756 * small changes in wire protocol
6758 2002-07-18 Phil Schwan <phil@clusterfs.com>
6760 * delivered as Lustre Light Alpha
6761 * fixed a crash after handling invalid MDS requests
6762 * fixed directory pages for architectures with non-4k pages sizes
6764 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6765 * release version v0_4_4
6766 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6768 * Fixes a number of interruption problems with OST operations.
6769 * Update documentation for portals header changes
6770 * Move all wire protocol structs/defines to lustre_idl.h
6771 * Fixes symlink length bug.
6772 * Add tcpdump to repository.
6774 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6775 * release version v0_4_3
6776 * Fixes statfs for inodes on extN.
6777 * Fixes bug in runtests which would delete /etc/hosts.
6778 * Use 64-bit object IDs wherever possible (not into VFS though)
6779 Remove ost_get_info, which is unused by lustre, and out of date.
6781 2002-07-03 Peter Braam <braam@clusterfs.com>
6782 * release version v0_4_2 Fixes a lookup error (type not passed)
6783 * move forward to head of Portals
6784 * move forward to latest Lustre kernel
6786 2002-06-25 Peter Braam <braam@clusterfs.com>
6787 * release version v0_4_1. Hopefully stable on single node use.