3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.18.1.el7 (RHEL7.7)
7 3.10.0-1127.19.1.el7 (RHEL7.8)
8 3.10.0-1160.2.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.6.3.el8 (RHEL8.2)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.180-94.100 (SLES12 SP3)
17 4.4.0-131 (Ubuntu 16.04)
18 4.15.0-32 (Ubuntu 18.04)
19 5.4.0-48 (Ubuntu 20.04)
20 vanilla linux 5.4.0 (ZFS + ldiskfs)
21 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
22 * Client primary kernels built and tested during release cycle:
23 3.10.0-1062.18.1.el7 (RHEL7.7)
24 3.10.0-1127.19.1.el7 (RHEL7.8)
25 3.10.0-1160.2.1.el7 (RHEL7.9)
26 4.12.14-95.48 (SLES12 SP4)
27 4.12.14-122.20 (SLES12 SP5)
28 4.12.14-197.37 (SLES15 SP1)
29 4.15.0-48 (Ubuntu 18.04)
30 * Other clients known to build on these kernels at some point (others may also work):
31 3.10.0-862.14.4.el7 (RHEL7.5)
32 3.10.0-957.27.2.el7 (RHEL7.6)
33 4.14.0-49.13.1.el7a (RHEL7.5)
34 4.14.0-115.2.2.el7a (RHEL7.6)
35 4.18.0-80.11.2.el8 (RHEL8.0)
36 4.18.0-147.8.1.el8 (RHEL8.1)
37 4.18.0-193.6.3.el8 (RHEL8.2)
38 4.4.120-92.70 (SLES12 SP2)
39 4.4.180-94.100 (SLES12 SP3)
40 4.4.0-131 (Ubuntu 16.04)
41 5.4.0-37 (Ubuntu 20.04)
42 5.4.0 (vanilla kernel.org)
43 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
44 * Recommended ZFS version: 0.8.4
45 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
46 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
47 stack overflow. For more information, please refer to bugzilla 17630.
48 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
49 the re-exporting nodes to be mounted with "32bitapi" mount option
51 --------------------------------------------------------------------------------
55 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
56 for currently supported client and server kernel versions.
57 * Server primary kernels built and tested during release cycle:
58 3.10.0-1062.1.1.el7 (RHEL7.7)
59 * Other server kernels known to build and work at some point (others may also work):
60 3.10.0-862.14.4.el7 (RHEL7.5)
61 3.10.0-957.27.2.el7 (RHEL7.6)
62 4.18.0-80.7.1.el8 (RHEL8.0)
63 4.4.120-92.70 (SLES12 SP2)
64 4.4.180-94.100 (SLES12 SP3)
65 4.4.0-131 (Ubuntu 16.04)
66 4.15.0-32 (Ubuntu 18.04)
67 vanilla linux 4.6.7 (ZFS only)
68 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
69 * Client primary unpatched kernels built and tested during release cycle:
70 3.10.0-1062.1.1.el7 (RHEL7.7)
71 4.12.14-95.29 (SLES12 SP4)
72 4.15.0-48 (Ubuntu 18.04)
73 * Other clients known to build on unpatched kernels at some point (others may also work):
74 3.10.0-862.14.4.el7 (RHEL7.5)
75 3.10.0-957.27.2.el7 (RHEL7.6)
76 4.14.0-49.13.1.el7a (RHEL7.5)
77 4.14.0-115.2.2.el7a (RHEL7.6)
78 4.18.0-80.7.1.el8 (RHEL8.0)
79 4.4.120-92.70 (SLES12 SP2)
80 4.4.180-94.100 (SLES12 SP3)
81 4.4.0-131 (Ubuntu 16.04)
82 vanilla linux 4.15.0 (ZFS only)
83 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
84 * Recommended ZFS / SPL version: 0.7.13
85 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
86 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
87 stack overflow. For more information, please refer to bugzilla 17630.
88 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
89 the re-exporting nodes to be mounted with "32bitapi" mount option
90 * Note for Ubuntu 14.04.5:
91 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
92 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
94 --------------------------------------------------------------------------------
98 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
99 for currently supported client and server kernel versions.
100 * Server known to build on the following kernels (others may also work):
101 2.6.32-696.18.7.el6 (RHEL6.9)
102 3.10.0-862.14.4.el7 (RHEL7.5)
103 3.10.0-957.el7 (RHEL7.6)
104 4.4.120-92.70 (SLES12 SP2)
105 4.4.162-94.69 (SLES12 SP3)
106 4.4.0-131 (Ubuntu 16.04)
107 4.15.0-32 (Ubuntu 18.04)
108 vanilla linux 4.6.7 (ZFS only)
109 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
110 * Client known to build on unpatched kernels (others may also work):
111 2.6.32-696.18.7.el6 (RHEL6.9)
112 2.6.32-754.6.3.el6 (RHEL6.10)
113 3.10.0-862.14.4.el7 (RHEL7.5)
114 3.10.0-957.el7 (RHEL7.6)
115 4.14.0-49.13.1.el7a (RHEL7.5)
116 4.14.0-115.2.2.el7a (RHEL7.6)
117 4.4.120-92.70 (SLES12 SP2)
118 4.4.162-94.69 (SLES12 SP3)
119 4.4.0-131 (Ubuntu 16.04)
120 4.15.0-32 (Ubuntu 18.04)
121 vanilla linux 4.15.0 (ZFS only)
122 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
123 * Recommended ZFS / SPL version: 0.7.9
124 * Tested with ZFS / SPL version: 0.7.9
125 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
126 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
127 stack overflow. For more information, please refer to bugzilla 17630.
128 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
129 the re-exporting nodes to be mounted with "32bitapi" mount option
130 * Note for Ubuntu 14.04.5:
131 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
132 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
134 --------------------------------------------------------------------------------
136 2018-04-03 Intel Corporation
138 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
139 for currently supported client and server kernel versions.
140 * Server known to build on patched kernels:
141 2.6.32-431.29.2.el6 (RHEL6.5)
142 2.6.32-504.30.3.el6 (RHEL6.6)
143 2.6.32-573.26.1.el6 (RHEL6.7)
144 2.6.32-642.15.1.el6 (RHEL6.8)
145 2.6.32-696.18.7.el6 (RHEL6.9)
146 3.10.0-693.21.1.el7 (RHEL7.4)
147 3.0.101-0.47.71 (SLES11 SP3)
148 3.0.101-107 (SLES11 SP4)
149 3.12.74-60.64.40 (SLES12 SP1)
150 4.4.114-92.64 (SLES12 SP2)
151 4.4.114-94.11 (SLES12 SP3)
152 3.13.0-101 (Ubuntu 14.04, ZFS only)
153 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
154 4.4.0-116 (Ubuntu 16.04)
155 vanilla linux 4.6.7 (ZFS only)
156 * Client known to build on unpatched kernels:
157 2.6.32-431.29.2.el6 (RHEL6.5)
158 2.6.32-504.30.3.el6 (RHEL6.6)
159 2.6.32-573.26.1.el6 (RHEL6.7)
160 2.6.32-642.15.1.el6 (RHEL6.8)
161 2.6.32-696.18.7.el6 (RHEL6.9)
162 3.10.0-693.21.1.el7 (RHEL7.4)
163 3.0.101-0.47.71 (SLES11 SP3)
164 3.0.101-107 (SLES11 SP4)
165 3.12.74-60.64.40 (SLES12 SP1)
166 4.4.114-92.64 (SLES12 SP2)
167 4.4.114-94.11 (SLES12 SP3)
168 3.13.0-101 (Ubuntu 14.04)
169 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
170 4.4.0-116 (Ubuntu 16.04)
172 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
173 * Recommended ZFS / SPL version: 0.7.6
174 * Tested with ZFS / SPL version: 0.7.6
175 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
176 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
177 stack overflow. For more information, please refer to bugzilla 17630.
178 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
179 the re-exporting nodes to be mounted with "32bitapi" mount option
180 * Note for Ubuntu 14.04.5:
181 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
182 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
184 --------------------------------------------------------------------------------
186 2017-07-05 Intel Corporation
188 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
189 for currently supported client and server kernel versions.
190 * Server known to build on patched kernels:
191 2.6.32-431.29.2.el6 (RHEL6.5)
192 2.6.32-504.30.3.el6 (RHEL6.6)
193 2.6.32-573.26.1.el6 (RHEL6.7)
194 2.6.32-642.15.1.el6 (RHEL6.8)
195 2.6.32-696.3.1.el6 (RHEL6.9)
196 3.10.0-514.21.1.el7 (RHEL7.3)
197 3.0.101-0.47.71 (SLES11 SP3)
198 3.0.101-100 (SLES11 SP4)
199 3.12.74-60.64.40 (SLES12 SP1)
200 4.4.59-92.17 (SLES12 SP2)
201 vanilla linux 4.6.7 (ZFS only)
202 * Client known to build on unpatched kernels:
203 2.6.32-431.29.2.el6 (RHEL6.5)
204 2.6.32-504.30.3.el6 (RHEL6.6)
205 2.6.32-573.26.1.el6 (RHEL6.7)
206 2.6.32-642.15.1.el6 (RHEL6.8)
207 2.6.32-696.3.1.el6 (RHEL6.9)
208 3.10.0-514.21.1.el7 (RHEL7.3)
209 3.0.101-0.47.71 (SLES11 SP3)
210 3.0.101-100 (SLES11 SP4)
211 3.12.74-60.64.40 (SLES12 SP1)
212 4.4.59-92.17 (SLES12 SP2)
214 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
215 * Recommended ZFS / SPL version: 0.6.5.9
216 * Tested with ZFS / SPL version: 0.6.5.9
217 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
218 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
219 stack overflow. For more information, please refer to bugzilla 17630.
220 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
221 the re-exporting nodes to be mounted with "32bitapi" mount option
223 --------------------------------------------------------------------------------
225 12-07-2016 Intel Corporation
227 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
228 for currently supported client and server kernel versions.
229 * Server known to build on patched kernels:
230 2.6.32-431.29.2.el6 (RHEL6.5)
231 2.6.32-504.30.3.el6 (RHEL6.6)
232 2.6.32-573.26.1.el6 (RHEL6.7)
233 2.6.32-642.11.1.el6 (RHEL6.8)
234 3.10.0-514.el7 (RHEL7.3)
235 3.0.101-0.47.71 (SLES11 SP3)
236 3.0.101-80 (SLES11 SP4)
237 3.12.62-60.64.8 (SLES12 SP1)
238 vanilla linux 4.5.7 (ZFS only)
239 * Client known to build on unpatched kernels:
240 2.6.32-431.29.2.el6 (RHEL6.5)
241 2.6.32-504.30.3.el6 (RHEL6.6)
242 2.6.32-573.26.1.el6 (RHEL6.7)
243 2.6.32-642.11.1.el6 (RHEL6.8)
244 3.10.0-514.el7 (RHEL7.3)
245 3.0.101-0.47.71 (SLES11 SP3)
246 3.0.101-80 (SLES11 SP4)
247 3.12.62-60.64.8 (SLES12 SP1)
249 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
250 * Recommended ZFS / SPL version: 0.6.5.8
251 * Tested with ZFS / SPL version: 0.6.5.8
252 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
253 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
254 stack overflow. For more information, please refer to bugzilla 17630.
255 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
256 the re-exporting nodes to be mounted with "32bitapi" mount option
258 --------------------------------------------------------------------------------
260 02-29-2016 Intel Corporation
262 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
263 for currently supported client and server kernel versions.
264 * Server known to build on patched kernels:
265 2.6.32-431.29.2.el6 (RHEL6.5)
266 2.6.32-504.30.3.el6 (RHEL6.6)
267 2.6.32-573.12.1.el6 (RHEL6.7)
268 3.10.0-327.3.1.el7 (RHEL7.2)
269 3.0.101-0.47.71 (SLES11 SP3)
270 3.0.101-68 (SLES11 SP4)
271 vanilla linux 4.2.1 (ZFS only)
272 * Client known to build on unpatched kernels:
273 2.6.32-431.29.2.el6 (RHEL6.5)
274 2.6.32-504.30.3.el6 (RHEL6.6)
275 2.6.32-573.12.1.el6 (RHEL6.7)
276 3.10.0-327.3.1.el7 (RHEL7.2)
277 3.0.101-0.47.71 (SLES11 SP3)
278 3.0.101-68 (SLES11 SP4)
281 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
282 * Recommended ZFS / SPL version: 0.6.4.2
283 * Tested with ZFS / SPL version: 0.6.4.2
284 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
285 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
286 stack overflow. For more information, please refer to bugzilla 17630.
287 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
288 the re-exporting nodes to be mounted with "32bitapi" mount option
290 --------------------------------------------------------------------------------
292 03-10-2015 Intel Corporation
294 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
295 for currently supported client and server kernel versions.
296 * Server known to build on patched kernels:
297 2.6.32-431.29.2.el6 (RHEL6.5)
298 2.6.32-504.8.1.el6 (RHEL6.6)
299 3.0.101-0.46 (SLES11 SP3)
300 * Client known to build on unpatched kernels:
301 2.6.32-431.29.2.el6 (RHEL6.5)
302 2.6.32-504.8.1.el6 (RHEL6.6)
303 3.10.0-123.20.1.el7 (RHEL7)
304 3.0.101-0.46 (SLES11 SP3)
305 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
306 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
307 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
308 stack overflow. For more information, please refer to bugzilla 17630.
309 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
310 the re-exporting nodes to be mounted with "32bitapi" mount option
312 Severity : enhancement
314 Description: control OST-index in IDIF via ROCOMPAT flag.
315 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
316 last_rcvd file. For new formatted OST device, it will be auto set;
317 for the case of upgrading from old OST device, you can enable it
318 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
319 enabled, for new created OST-object, its IDIF-in-LMA will contain
320 the OST-index; for the existing OST-object, the OSD will convert
321 old format IDIF as new format IDIF with OST-index stored in the
322 LMA EA when accessing such OST-object or via OI scrub. Once such
323 flag is enabled, it cannot be reverted back, so the system cannot
324 be downgraded to the orignal incompatible version.
326 --------------------------------------------------------------------------------
328 07-30-2014 Intel Corporation
330 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
331 for currently supported client and server kernel versions.
332 * Server known to build on patched kernels:
333 2.6.32-431.20.3.el6 (RHEL6.5)
334 3.0.101-0.31 (SLES11 SP3)
335 * Client known to build on unpatched kernels:
336 2.6.32-431.20.3.el6 (RHEL6.5)
337 3.0.101-0.31 (SLES11 SP3)
340 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
341 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
342 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
343 stack overflow. For more information, please refer to bugzilla 17630.
344 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
345 the re-exporting nodes to be mounted with "32bitapi" mount option
347 --------------------------------------------------------------------------------
349 10-31-2013 Intel Corporation
351 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
352 * Server builds for kernels:
353 2.6.32-358.18.1.el6 (RHEL6.4)
354 * Client builds for unpatched kernels:
355 2.6.32-358.18.1.el6 (RHEL6.4)
356 3.0.92-0.8 (SLES11 SP3)
358 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
359 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
360 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
361 stack overflow. For more information, please refer to bugzilla 17630.
362 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
363 the re-exporting nodes to be mounted with "32bitapi" mount option
365 --------------------------------------------------------------------------------
367 05-30-2013 Intel Corporation
369 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
370 * Server builds for kernels:
371 2.6.32-358.6.2.el6 (RHEL6.4)
372 * Client builds for unpatched kernels:
373 2.6.32-358.6.2.el6 (RHEL6.4)
374 3.0.74-0.6.6 (SLES11 SP2)
376 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
377 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
378 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
379 stack overflow. For more information, please refer to bugzilla 17630.
380 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
381 the re-exporting nodes to be mounted with "32bitapi" mount option
383 --------------------------------------------------------------------------------
385 9-30-2012 Whamcloud, Inc.
387 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
388 * Server support for kernels:
389 2.6.32-279.5.1.el6 (RHEL6)
390 * Client support for unpatched kernels:
391 2.6.32-279.5.1.el6 (RHEL6)
393 2.6.18-238.19.1.el5 (RHEL5)
394 2.6.16.60-0.69.1 (SLES10)
395 * Recommended e2fsprogs version: 1.41.90.wc4
396 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
397 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
398 stack overflow. For more information, please refer to bugzilla 17630.
399 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
400 the re-exporting nodes to be mounted with "32bitapi" mount option
402 Severity : enhancement
404 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
406 Severity : enhancement
408 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
410 Severity : enhancement
412 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
414 Severity : enhancement
416 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
418 Severity : enhancement
420 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
422 --------------------------------------------------------------------------------
424 04-30-2012 Whamcloud, Inc.
426 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
427 * Server support for kernels:
428 2.6.32-220.4.2.el6 (RHEL6)
429 * Client support for unpatched kernels:
430 2.6.32-220.4.2.el6 (RHEL6)
432 2.6.18-238.19.1.el5 (RHEL5)
433 2.6.16.60-0.69.1 (SLES10)
434 * Recommended e2fsprogs version: 1.41.90.wc4
435 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
436 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
437 stack overflow. For more information, please refer to bugzilla 17630.
438 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
439 the re-exporting nodes to be mounted with "32bitapi" mount option
441 Severity : enhancement
443 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
445 Severity : enhancement
447 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
449 Severity : enhancement
451 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
453 Severity : enhancement
455 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
457 --------------------------------------------------------------------------------
459 09-30-2011 Whamcloud, Inc.
461 * Server support for kernels:
462 2.6.18-238.19.1.el5 (RHEL5)
463 2.6.32-131.6.1.el6 (RHEL6)
464 * Client support for unpatched kernels:
465 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
466 2.6.18-238.19.1.el5 (RHEL5)
467 2.6.32-131.6.1.el6 (RHEL6)
468 2.6.29.4-167.fc11 (FC11)
469 2.6.16.60-0.69.1 (SLES10)
471 * Recommended e2fsprogs version: 1.41.90.wc3
472 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
473 Lustre filesystem with 4K stack may cause a stack overflow. For more
474 information, please refer to bugzilla 17630.
475 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
476 mounted with "32bitapi" mount option
478 Severity : enhancement
479 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
481 Severity : enhancement
482 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
484 Severity : enhancement
485 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
488 Frequency : only with SLES10
490 Description: Use OFED "KMP" provided by Novell
491 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
492 Lustre is now built against this package. That means you need to
493 install the ofed-kmp package from Novell for the patchless client.
494 Note that the ofed-kmp that Novell ships may not exactly match the
495 kernel version but should still be compatible.
497 Severity : enhancement
499 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
500 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
502 --------------------------------------------------------------------------------
504 2010-07-15 Oracle, Inc.
506 * Server support for kernels:
507 2.6.18-164.11.1.el5 (RHEL 5)
508 2.6.18-164.11.1.0.1.el5 (OEL 5)
509 * Client support for unpatched kernels:
510 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
511 2.6.18-164.11.1.el5 (RHEL 5),
512 2.6.18-164.11.1.0.1.el5 (OEL 5)
513 2.6.16.60-0.42.8 (SLES 10),
515 2.6.29.4-167.fc11 (FC11)
516 * Recommended e2fsprogs version: 1.41.10-sun2
518 Severity : enhancement
520 Description: Distribute lfs_migrate script with lustre
521 Details : The lfs_migrate mentioned in the manual is now included in the
522 lustre tree in lustre/scripts and will distributed in lustre RPMs.
524 Severity : enhancement
526 Description: Failover nid list added to import proc file.
528 Severity : enhancement
530 Description: "som_preview" mount option required to use SOM
531 Details : The clients will need to use som_preview mount option
535 Description: Lustre does not do 1MB IOs to HW RAID
536 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
538 Severity : enhancement
540 Description: lfs quota output is not very convenient for awk/sed-parsing
541 Details : Some positions in lfs quota output table could be empty or
542 non-empty which made it hard to parse it with scripts, now
543 a dash is put instead of space where there is not supposed
548 Description: Update to OFED 1.5.1
549 Details : OFED 1.5.1 will be used in environments that don't use
550 vendor supplied OFED.
554 Description: Remove lnet. prefix from lctl params display
555 Details : Also renamed lnet.memused to lnet_memused to
560 Description: Use conf_param -d to remove permanant settings.
564 Description: Add list_param -R
565 Details : lctl list_param now supports recursive mode.
568 Bugzilla : 21829, 22312
569 Description: Fix llobdstat.
570 Details : 1. Now updating the write_bytes counter
571 2. Port script fixes from 1.8.
575 Description: decrease the usage of memory on clients.
576 Details : 1. On clients, recycle dentries and inodes unused.
577 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
582 Description: llobdstat fix and enhancement
583 Details : add a counter to set a limit to how many samples will be returned
584 fix a wildcard in the path to limit to obdfilter stats only
588 Description: lfs check only allowed for root
589 Details : Code cleanups and added a regresion test of non-root use of
594 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
595 Details : Fix hr->hr_index overload.
599 Description: Fix FIEMAP functionality.
603 Description: FC11 patchless client support added
607 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
608 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
612 Description: New acc-sm script and various test-framework enhancments.
616 Description: Previously landed patch was reverted.
620 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
624 Description: Add ldiskfs maxdirsize mount option.
628 Description: Fix mkfs.lustre support for 16TB LUNs.
632 Description: Print mmp_check_interval
633 Details : print mmp_check_interval and make it possible to abort
634 mount operation in case it takes too long.
638 Description: Add cascading_rw to lustre mpi tests.
642 Description: Store SOM data in EA.
643 Details : Also several SOM related cleanups.
647 Description: Add refcount to md_open_data
651 Description: Use refcount on lov_request_set to prevent memory corruption.
655 Description: Statahead: use parent dir's i_mutex for synchronization.
659 Description: loadgen improvements.
663 Description: Enable peer health detection in ksocklnd and ko2iblnd.
667 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
671 Description: Fix race on obd_devcie:md_stats
675 Description: Fix SLES 11 client builds.
676 Details : Resolved uninitalized pointer gcc warnings.
680 Description: Allow non-root access for "lfs check".
681 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
685 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
687 Severity : enhancement
689 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
690 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
694 Description: Sanity test_27p fails.
695 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
696 Sanity tests 27 were cleaned up and improved.
700 Description: Unify req format on client/servers
701 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
702 buffers and swabbers approach..
707 Description: drop join file stuff.
709 Severity : enhancement
711 Description: Add 2.6.27 and SLES11 (patchless client) support
715 Description: Fix in ptlrpc_expire_one_request() to print the signed time
718 Severity : enhancement
720 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
723 Severity : enhancement
725 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
727 Severity : enhancement
729 Description: correctly handle big reply message.
730 Details : send LNet event if reply is bigger than buffer and adjust this buffer
735 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
737 Severity : enhancement
739 Description: Add support for OEL5.
743 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
745 Severity : enhancement
747 Description: "weak-modules" support
748 Details : Implement "weak-modules" support which enables kernel modules
749 to be used with any kernel that implements the same kABI. In
750 order to achieve this modules are now installed in
751 /lib/modules/$(uname -r)/updates/kernel on all distributions.
755 Description: client could not reconnect to OST because of an active request.
756 Details : The client now retries to reconnect to the same server, if a
757 connect request failed with EBUSY or -EAGAIN. Test case 35b
758 conf-sanity.sh has been created.
762 Description: Don't allow make backward step on assignin osc next id.
763 Details : race between allocation next id and ll_sync thread can be cause
764 of set wrong osc next id and can be kill valid ost objects.
768 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
772 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
776 Description: Error handling in osc_statfs_interpret() has been improved.
777 Details : Check in osc_statfs_interpret() for EBADR.
781 Description: Do not send statfs() requests to OSTs disabled by administrator.
782 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
784 Severity : enhancement
786 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
788 Severity : enhancement
790 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
792 Severity : enhancement
794 Description: Enhancement for lfs(1) command to use numeric uid/gid.
798 Description: Improved cleanup procedure for conf-sanity test_46a.
802 Description: Do not start bulk transfer, if client tried to reconnect.
803 Details : Do not start bulk transfer, if client tried to reconnect,
804 but failed due to active rpcs. Check for exp_abort_active_req
805 has been implemented.
809 Description: Prevent inconsistences between linux and lustre mount structures.
810 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
811 Make the sleep interruptible.
813 Severity : enhancement
815 Description: provide server to client comms path
816 Details : server to client communications path via new LDLM_SET_INFO rpc
820 Description: truncate starts GFP_FS allocation under transaction causing deadlock
821 Details : ldiskfs_truncate calls grab_cache_page which may start page
822 allocation under an open transaction. This may lead to
823 calling prune_icache with consequent lustre reentrance.
827 Description: send correctly lsm on open replay
828 Details : MDS is trust to LSM size on replay open, but client can set wrong size
833 Description: enable client interface failover
834 Details : When a child reconnects from another NID, properly update export
835 nid hash position and ldlm reverse import.
837 Severity : enhancement
839 Description: Communicate OST degraded/readonly state via statfs to MDS
840 Details : Flags in the statfs returned from OSTs indicate whether the
841 OST is in a degraded RAID state, or if the filesystem has
842 turned read-only after a filesystem error is detected.
847 Description: don't panic if EPROTO was hit when reading symlink
848 Details : correctly handling request reference in error cases.
850 Severity : enhancement
852 Description: Add LustreNetLink, a kernel-userspace communcation path.
854 Severity : enhancement
856 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
861 Description: access to llog context before init.
862 Details : move handling CATALOGS file at osc layer and forbid access to llog
866 Frequency : rare, only when client is writing to OST on local node
868 Description: Avoid deadlock for local client writes
869 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
870 memory freeing context. This allows OST threads to set the
871 PF_MEMALLOC flag on task structures in order to allocate memory
872 from reserved pools and complete IO.
873 Use GFP_HIGHUSER for OST allocations for non-local client writes,
874 so that the OST threads generate memory pressure and allow
875 inactive pages to be reclaimed.
877 Severity : enhancement
879 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
884 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
885 Details : move ldlm namespace creation in setup phase to avoid grab
886 _lprocfs_lock with cli_sem held.
890 Description: Temporarily disable grant shrink.
891 Details : Disable the feature for debugging.
894 Frequency : only when formatting test filesystems
896 Description: Unable to run several mkfs.lustre on loop devices at the same time
897 Details : mkfs.lustre returns error 256 on the concurrent loop devices
898 formatting. The solution is to proper handle the error.
900 Severity : enhancement
902 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
904 Severity : enhancement
906 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
911 Description: resolve race between obd_disconnect and class_disconnect_exports
912 Details : if obd_disconnect will be called to already disconnected export he
913 forget release one reference and osc module can't unloaded.
915 Severity : enhancement
917 Description: Allow tuning service thread via /proc
918 Details : For each service a new
919 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
920 created that can be used to set min/max thread counts, and get the
921 current number of running threads.
926 Descriptoin: don't return error if have partially created objects for file.
927 Details : lov_update_create_set uses set->set_success as index for created
928 objects, so if some requests will be failed, they will have hole
929 at end of array and qos_shrink_lsm for allocate correct lsm.
931 Severity : enhancement
933 Description: Update OFED support to 1.4.1
938 Description: Reduce small size read RPC
939 Details : Set read-ahead limit for every file and only do read-ahead when
940 available read-ahead pages are bigger than 1M to avoid small size
943 Severity : enhancement
945 Description: move AT tunable parameters for more consistent usage
946 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
948 Severity : enhancement
950 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
951 Details : allow skip disconnected ost for send statfs request and hide error
957 Description: fix racy locking of mballoc block bitmaps causing BUG
958 Details : The locking of the mballoc buddy bitmap and the in-memory
959 block bitmap was using two different spin locks in some
960 cases. This made it possible to incorrectly access the
961 mballoc bitmap while another process was modifying it,
962 causing a sanity assertion to fail. While no on-disk corruption
963 was reported, there was some risk of this happening.
966 Frequency : rare, on llog test 6
968 Descriptoin: don't allow connect to already connected import
969 Details : allowing connect to already connected import is hide connecting problem.
974 Descriptoin: allow kill process which wait statahead result
975 Details : for some reasons 'ls' can stick in waiting result from statahead,
976 in this case need way for kill this process.
978 Severity : enhancement
980 Description: Add state history info file, enhance import info file
981 Details : Track import connection state changes in a new osc/mdc proc file;
982 add overview-type data to the osc/mdc import proc file.
984 Severity : enhancement
986 Description: MDS create should not wait for statfs RPC while holding DLM lock.
988 Severity : enhancement
990 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
995 Descriptoin: OST grant too much space to client even there are not enough space.
996 Details : Client will shrink its grant cache to OST if there are no write
997 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
998 this grant cache if there are already not enough avaible space
999 (left_space < total_clients * 32M).
1002 Frequency : start MDS on uncleanly shutdowned MDS device
1004 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1005 Details : stay in waiting mds<>ost recovery finished produce random bugs
1006 due race between two ll_sync thread for one lov target. send
1007 ACTIVATE event only if connect realy finished and import have
1011 Frequency : rare, connect and disconnect target at same time
1013 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1014 Details : don't call obd_disconnect under lov_lock. this long time
1015 operation and can block ptlrpcd which answer to connect request.
1020 Descriptoin: don't lose wakeup for imp_recovery_waitq
1021 Details : recover_import_no_retry or invalidate_import and import_close can
1022 both sleep on imp_recovery_waitq, but we was send only one wakeup
1026 Frequency : always with long access acl
1028 Descriptoin: mds can't pack reply with long acl.
1029 Details : mds don't control size of acl but they limited by reint/getattr
1032 Severity : enhancement
1034 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1036 Severity : enhancement
1038 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1041 Frequency : start MDS on uncleanly shutdowned MDS device
1043 Descriptoin: aborting recovery hang on MDS
1044 Details : don't throttle destroy RPCs for the MDT.
1047 Frequency : on remount
1049 Description: external journal device not working after the remount
1050 Details : clear dev_rdonly flag for external journal devices in
1056 Description: shutdown vs evict race
1057 Details : client_disconnect_export vs connect request race.
1058 if client will evicted at this time - we start invalidate
1059 thread without referece to import and import can be freed
1063 Frequency : rare, need acl's on inode.
1065 Description: client can't handle ost additional correctly
1066 Details : if ost was added after client connected to mds client can have
1067 hit lnet_try_match_md ... to big messages to wide striped files.
1068 in this case need teach client to handle config events about add
1069 lov target and update client max ea size at that event.
1071 Severity : enhancement
1073 Description: Changelogs
1074 Details : Changelogs are a lightweight mechanism to track filesystem
1075 metadata and namespace changes. The changelog is recorded
1076 permanently on the MDTs, and is periodically "consumed" / purged
1077 when records are no longer needed.
1079 Severity : enhancement
1081 Description: compact fld format with extents
1082 Details : Store range of seq rather than every seq in FLD. Seq
1083 controller update FLD rather than clients. In Case of CMD, mdt0
1084 has FLD, all other metadata server act as non persistent proxy
1085 for FLD queries and cache fld entries in fld cache.
1090 Description: don't skip ost target if they assigned to file
1091 Details : Drop slow OSCs if we can, but not for requested start idx.
1092 This means "if OSC is slow and it is not the requested
1093 start OST, then it can be skipped, otherwise skip it only
1094 if it is inactive/recovering/out-of-space.
1098 Description: more cleanup in mds_lov
1099 Details : not send LOV EA under replay, we can't know about they size at this
1100 time. Don't allow client connect to mds before any ost connected,
1101 for avoid problems with LOV EA size and returning EIO to client.
1103 Severity : enhancement
1105 Description: Interoperability at server side (Disk interoperability)
1107 Severity : enhancement
1109 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1111 Severity : enhancement
1113 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1115 Severity : enhancement
1117 Description: New client IO stack (CLIO).
1119 Severity : enhancement
1121 Description: Commit on sharing. Eliminate inter-client dependencies between
1122 uncommitted transactions by doing transaction commits.
1123 Thereby clients may recovery independently.
1126 Frequency : Create a symlink file with a very long name
1128 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1129 Details : If there is no extra space in the request for early cancels,
1130 ldlm_req_handles_avail() returns 0 instead of a negative value.
1132 Severity : enhancement
1134 Description: Add /proc entry for import status
1135 Details : The mdc, osc, and mgc import directories now have
1136 an import directory that contains useful import data for debugging
1137 connection problems.
1139 Severity : enhancement
1141 Description: Re-disable certain /proc logging
1142 Details : Enable and disable client's offset_stats, extents_stats and
1143 extents_stats_per_process stats logging on the fly.
1146 Frequency : Only on FC kernels 2.6.22+
1148 Description: oops in statahead
1149 Details : Do not drop reference count for the dentry from VFS when lookup,
1150 VFS will do that by itself.
1152 Severity : enhancement
1154 Description: Generic /proc file permissions
1155 Details : Set /Proc file permissions in a more generic way to enable non-
1156 root users operate on some /proc files.
1160 Description: Hitting mdc_commit_close() ASSERTION
1161 Details : Properly handle request reference release in
1162 ll_release_openhandle().
1166 Description: quota recovery deadlock during mds failover
1167 Details : This patch includes att18982, att18236, att18237 in bz14840.
1169 1. fix osts hang when mds does failover with quotaon
1170 2. prevent watchdog storm when osts threads wait for the
1175 Frequency : only patchless client
1176 Description: add workaround for race between add/remove dentry from hash
1178 Severity : enhancement
1180 Description: Allow OST glimpses to return PW locks
1184 Description: LBUG when llog conf file is full
1185 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1189 Description: Prevent import from entering FULL state when server in recovery
1193 Description: service mount cannot take device name with ":"
1194 Details : Only when device name contains ":/" will mount treat it as
1200 Description: replace ptlrpcd with the statahead thread to interpret the async
1201 statahead RPC callback
1205 Frequency : on recovery
1206 Description: I/O failures after umount during fail back
1207 Details : if client reconnected to restarted server we need join to recovery
1208 instead of find server handler is changed and process self
1209 eviction with cancel all locks.
1211 Severity : enhancement
1213 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1215 Severity : enhancement
1217 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1219 Severity : enhancement
1221 Description: Upcall on Lustre log has been dumped
1222 Details : Allow for a user mode script to be called once a Lustre log has
1223 been dumped. It passes the filename of the dumped log to the
1224 script, the location of the script can be specified via
1225 /proc/sys/lnet/debug_log_upcall.
1230 Description: avoid idr_remove called for id which is not allocated.
1231 Details : Move assigment s_dev for clustered nfs to end of initialization,
1232 to avoid problem with error handling.
1237 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1238 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1239 into CONN_USED_HASH and this prodice warning when put connection
1240 again in unused hash.
1243 Severity : enhancement
1245 Description: File striping can now be set to use an arbitrary pool of OSTs.
1247 Severity : enhancement
1249 Description: Export bytes_read/bytes_write count on OSC/OST.
1253 Description: Early reply size mismatch, MGC loses connection
1254 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1255 the connect flags are properly negotiated.
1260 Description: quotas are not honored with O_DIRECT
1261 Details : all writes with the flag O_DIRECT will use grants which leads to
1262 this problem. Now using OBD_BRW_SYNC to guard this.
1266 Description: add quota statistics
1267 Details : 1. sort out quota proc entries and proc code.
1268 2. add quota statistics
1270 Severity : enhancement
1272 Description: enable quota support for HEAD.
1276 Description: Properly propagate oinfo flags from lov to osc for statfs
1277 Details : restore missing copy oi_flags to lov requests.
1279 Severity : enhancement
1281 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1282 : l_getgroups(8), lst(8), routerstat(8)
1283 Details : included man pages for llobdstat(8), llstat(8),
1284 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1286 Severity : enhancement
1288 Description: configure's --enable-quota should check the
1289 : kernel .config for CONFIG_QUOTA
1290 Details : configure is terminated if --enable-quota is passed but
1291 : no quota support is in kernel
1295 Description: Remove portals compatibility
1296 Details : Remove portals compatibility, not interoperable with releases
1301 Description: Resolve device initialization race
1302 Details : Prevent proc handler from accessing devices added to the
1303 obd_devs array but yet be intialized.
1305 Severity : enhancement
1307 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1309 Severity : enhancement
1311 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1316 Description: Using wrong pointer in osc_brw_prep_request
1317 Details : Access to array[-1] can produce panic if kernel compiled with
1318 CONFIG_PAGE_ALLOC enabled
1322 Description: Client runs out of low memory
1323 Details : Consider only lowmem when counting initial number of llap pages
1327 Description: *optional* service tags registration
1328 Details : if the "service tags" package is installed on a Lustre node
1329 When the filesystem is mounted, a local-node service tag will
1330 be created. See http://inventory.sun.com/ for more information
1331 about the Service Tags asset management system.
1335 Description: Kernel BUG tries to release flock
1336 Details : Lustre does not destroy flock lock before last reference goes
1337 away. So always drop flock locks when client is evicted and
1338 perform unlock regardless of successfulness of speaking to MDS.
1342 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1346 Description: sanity test 65a fails if stripecount of -1 is set
1347 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1352 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1353 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1356 Severity : enhancement
1358 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1362 Description: do not process already freed flock
1363 Details : flock can possibly be freed by another thread before it reaches
1364 to ldlm_flock_completion_ast.
1368 Description: LBUG during stress test
1369 Details : Need properly lock accesses the flock deadlock detection list.
1373 Description: oops in page fault handler
1374 Details : kernel page fault handler can return two special 'pages' in error
1375 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1379 Description: timeout with invalidate import.
1380 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1381 should be handled by ptlrpcd. This produce long age waiting and
1382 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1384 Severity : enhancement
1386 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1390 Description: enable MGS and MDT services start separately
1391 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1392 a co-located MGS without starting the MGS, which is a complement
1393 to 'nosvc' mount option.
1397 Frequency : after recovery
1398 Description: precreate to many object's after del orphan.
1399 Details : del orphan st in oscc last_id == next_id and this triger growing
1400 count of precreated objects. Set flag LOW to skip increase count
1401 of precreated objects.
1405 Frequency : rare, on clear nid stats
1406 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1407 Details : when clean nid stats sometimes try destroy live entry,
1408 and this produce panic in free.
1412 Description: Stack overflow during MDS log replay
1413 ease stack pressure by using a thread dealing llog_process.
1417 Description: wait until IO finished before start new when do lock cancel.
1418 Details : VM protocol want old IO finished before start new, in this case
1419 need wait until PG_writeback is cleared until check dirty flag and
1420 call writepages in lock cancel callback.
1422 Severity : enhancement
1424 Description: using special macro for print time and cleanup in includes.
1428 Description: mds_mfd_close() ASSERTION(rc == 0)
1429 Details : In mds_mfd_close(), we need protect inode's writecount change
1430 within its orphan write semaphore to prevent possible races.
1434 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1438 Frequency : rare, on shutdown ost
1439 Description: don't hit live lock with umount ost.
1440 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1441 use shrink_dcache_sb instead.
1445 Description: don't panic with use echo client
1446 Details : echo client pass NULL as client nid pointer and this produce null
1447 pointer dereference.
1451 Description: fix build on ppc32
1452 Details : compile code with -m64 flag produce wrong object file for ppc32.
1456 Description: add message levels for liblustreapi
1460 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1461 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1462 will result in an -EAGAIN and in the exit call path we call
1463 alloc_rr() which will with increasing aggressiveness attempt to
1464 aquire precreated objects on the minimum number of required OSCs.
1468 Description: Use old size assignment to avoid deadlock
1469 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1470 the scheduling while holding a spinlock. We do not need locking
1471 for size in ll_update_inode() because size is only updated from
1472 the MDS for directories or files without objects, so there is no
1473 other place to do the update, and concurrent access to such inodes
1474 are protected by the inode lock.
1478 Description: resolve "_IOWR redefined" build error on SLES10
1482 Description: dump the memory debugging after all modules are unloaded to
1483 suppress false negative in conf_sanity test 39
1485 Severity : enhancement
1487 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1492 Description: fix SLES kernel versioning
1493 Details : the kernel version for our SLES 10 kernel did not include a "-"
1494 before the "smp" at the end. while this was not a problem in
1495 general, it did mean that software trying to use the kernel
1496 version to try to detect a vendor specific kernel would fail.
1497 this was most evident by the OFED build scripts.
1501 Description: Don't update lov_desc members until making sure they are valid
1502 Details : When updating lov_desc members via proc fs, need fix their
1503 validities before doing the real update.
1507 Description: don't put request into delay list while invalidate in flight.
1508 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1509 import in flight. this produce timeout for invalidate and sometimes
1510 can cause stale data.
1514 Frequency : on ppc only
1515 Description: not convert ost objects for directory because it's not exist.
1516 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1520 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1522 Severity : enhancement
1524 Description: Update to support 2.6.22.14 vanilla kernel.
1528 Frequency : rare, on recovery
1529 Description: read procfs can produce deadlock in some situation
1530 Details : Holding lprocfs lock which send rpc can produce block for destroy
1531 obd objects and this also block reconnect with -EALREADY. This
1532 isn't fix all lprocfs bugs - but make it rare.
1534 Severity : enhancement
1536 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1539 Frequency : frequent on X2 node
1541 Description: mdc_set_open_replay_data LBUG
1542 Details : Set replay data for requests that are eligible for replay.
1546 Description: lustre_mgs: operation 101 on unconnected MGS
1547 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1548 MGC, and late on MGC cannot successfully connect to MGS and a lot
1549 of the error messages complaining that MGS is not connected.
1553 Frequency : on network error
1554 Description: panic with double free request if network error
1555 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1556 but it's true only for synchronus enqueue, for async enqueue
1557 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1560 Severity : enhancement
1562 Description: client-side metadata stat-ahead during readdir(directory readahead)
1563 Details : perform client-side metadata stat-ahead when the client detects
1564 readdir and sequential stat of dir entries therein
1567 Frequency : on start mds
1569 Description: Implement get_info(last_id) in obdfilter.
1572 Frequency : occasional
1574 Description: Correctly check stale fid, not start epoch if ost not support SOM
1575 Details : open with flag O_CREATE need set old fid in op_fid3 because
1576 op_fid2 was overwritten with new generated fid, but mds can answer
1577 with one of these two fids and both is not stale. Setattr
1578 incorrectly started an epoch and assume will be called
1579 done_writing, but without SOM done_writing ever being called.
1582 Frequency : rare, depends on device drivers and load
1584 Description: MDS or OSS nodes crash due to stack overflow
1585 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1586 In some cases, in conjunction with device drivers that use a lot
1587 of stack the MDS (or possibly OSS) service threads could overflow
1588 the stack. One change which was identified to consume additional
1589 stack has been reworked to avoid the extra stack usage.
1592 Frequency : occasional
1594 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1595 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1596 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1597 oscc_has_objects and see OSC has no objests and call
1598 oscc_internal_create to resend create request.
1600 Severity : enhancement
1602 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1604 Severity : enhancement
1606 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1609 Frequency : very rare
1611 Description: Fix replay if there is an un-replied request and open
1612 Details : In some cases, older replay request will revert the
1613 mcd->mcd_last_xid on MDS which is used to record the client's
1614 latest sent request.
1616 Severity : enhancement
1618 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1620 Severity : enhancement
1622 Description: Add rhel5 support to HEAD.
1624 Serverity : enhancement
1626 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1631 Description: Don't allow skipping OSTs if index has been specified.
1632 Details : Don't allow skipping OSTs if index has been specified, make
1633 locking in internal create lots better.
1637 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1638 Details : access to bitfield in structure is always rounded to long
1639 and this produce problem with not atomic change any bit.
1643 Description: Lustre make rpms failed.
1644 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1645 builds Lustre rpms from tarball.
1648 Frequency : rare on shutdown ost
1650 Description: If llog cancel was not send before clean_exports phase, this can
1651 produce deadlock in llog code.
1652 Details : If llog thread has last reference to obd and call class_import_put
1653 this produce deadlock because llog_cleanup_commit_master wait when
1654 last llog_commit_thread exited, but this never success because was
1655 called from llog_commit_thread.
1659 Description: allow userland application know is lost one of stripes.
1660 Details : fill lvb_blocks with error code on ost and return it to
1661 application if error flag found.
1665 Description: NULL lov_tgts causing MDS oops
1666 Details : more safe checks for NULL lov_tgts for avoid oops.
1668 Severity : enhancement
1670 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1674 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1675 Details : If an OST is down the MDS will hang indefinitely in
1676 obd_statfs_rqset() waiting for the statfs data. While for
1677 MDS QOS usage of statfs, it should not stuck in waiting.
1679 Severity : enhancement
1681 Description: remote_acl support
1682 Details : Support ACL-based permission check for remote user.
1683 Support setfacl/getfacl for remote user with the utils
1684 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1685 the system "{s,g}etfacl" utils.
1687 Severity : enhancement
1689 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1691 Severity : enhancement
1693 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1697 Description: make mgs_setparam() handle fsname containing dash
1698 Details : fsname containing a dash does not work with lctl conf_param
1700 Severity : enhancement
1702 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1704 Severity : enhancement
1706 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1710 Frequency : rare, at shutdown
1711 Description: access already free / zero obd_namespace.
1712 Details : if client_disconnect_export was called without force flag set,
1713 and exist connect request in flight, this can produce access to
1714 NULL pointer (or already free pointer) when connect_interpret
1715 store ocd flags in obd_namespace.
1719 Frequency : only at startup
1720 Description: not alloc memory with spinlock held.
1721 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1722 if any spinlock held.
1724 Severity : enhancement
1726 Description: make lustre randomly fail allocating memory
1727 Details : Make lustre randomly failed allocating memory for testing purpose.
1729 Severity : enhancement
1731 Description: lost problems with lov objid file
1732 Details : Fixes some scability and access to not inited memory problems
1733 in work with lov objdid file.
1738 Description: lfs find does not continue on file error
1739 Details : Continue other files processing when a file/dir is absent.
1743 Description: Inconsistent usage of lustre_pack_reply()
1744 Details : Standardize the usage of lustre_pack_reply() such that it
1745 always generate a CERROR on failure.
1748 Frequency : occasional
1750 Description: MDS hang or stay in waiting lock
1751 Details : If client receive lock with CBPENDING flag ldlm need send lock
1752 cancel as separate rpc, to avoid situation when cancel request
1753 can't processed due all i/o threads stay in wait lock.
1757 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1761 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1765 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1769 Description: Update to RHEL4 latest kernel.
1771 Severity : enhancement
1773 Description: Build SLES10 patchless client fails
1774 Details : The configure was broken by run ./configure with
1775 --with-linux-obj=.... argument for patchless client. When the
1776 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1777 can't search header adequately. Use absolute path such as
1778 -I($LINUX)/include instead.
1782 Description: interrupt oig_wait produce painc on resend.
1783 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1784 private set, and this produce situation when rq_ptlrpcd_data not
1785 copyed to new allocated request and triggered LBUG on assert
1786 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1787 ptlrpcd set if request is changed and can be safety changed to use
1792 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1793 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1794 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1795 these macros and replace them with their difinitions instead.
1799 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1800 Details : Update which_patch & target file for SLES10 latest kernel.
1803 Description: Remove client patches from SLES 10 kernel.
1804 Details : This causes SLES 10 clients to behave as patchless clients
1805 even on a Lustre-patched (server) kernel.
1807 Severity : enhancement
1809 Description: self-adjustable client's lru lists
1810 Details : use adaptive algorithm for managing client cached locks lru
1811 lists according to current server load, other client's work
1812 pattern, memory activities, etc. Both, server and client
1813 side namespaces provide number of proc tunables for controlling
1816 Severity : enhancement
1818 Description: light-weight GSS support
1819 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1820 and reduce performance overhead.
1822 Severity : enhancement
1824 Description: Linux keyring support
1825 Details : Support using service of Linux keyring for Lustre GSS internal
1826 context refresh/cache mechanism.
1830 Description: Fix errors in lfs documentation
1831 Details : Fixes man pages
1835 Description: don't use GFP_* in generic Lustre code.
1836 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1840 Description: obdclass is limited by single OBD_ALLOC(idarray)
1841 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1845 Description: warnings with build patchless client with vanila 2.6.19 and up
1846 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1847 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1851 Frequency : Only for SLES
1852 Description: sanity_quota fail test_1
1853 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1854 file, which makes TSTID[2] inunique.
1858 Frequency : Only for SLES9
1859 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1860 Details : Update target/ChangeLog/which_patch .
1864 Frequency : Only for test_10 in sanity-quota.sh
1865 Description: a bug in quota test code
1866 Details : checking a test flag in a wrong place
1870 Frequency : Only for test_13 in sanity-quota.sh
1871 Description: a bug in quota test script
1872 Details : shouldn't check overfull allocated quota
1876 Description: jbd statistics
1877 Details : Port older jbd statistics patch for sles10
1879 Severity : enhancement
1881 Description: Add printing inode info into message about error in writepage.
1886 Description: reply_lock_interpret crash due to race with it and lock cancel.
1887 Details : Do not replay locks that are being cancelled. Do not reference
1888 locks by their address during replay, just by their handle.
1892 Frequency : When flocks are used.
1893 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1894 Details : Flock locks might destroy just granted lock if it could be merged
1895 with another existing flock, this is done in completion handler,
1896 so teach ldlm_cli_enquque_fini that this is a valid case for
1902 Description: Oops in read and write path when failing to allocate lock.
1903 Details : Check if lock allocation failed and return error back.
1907 Description: lstripe command fails for valid OST index
1908 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1909 instead of lov->desc.ld_active_tgt_count.
1913 Description: sanity.sh failed test 103
1914 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1919 Description: df doesn't work properly if diskfs blocksize != 4K
1920 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1924 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1925 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1927 Severity : enhancement
1929 Description: Adaptive timeouts
1930 Details : RPC timeouts adapt to changing server load and network
1931 conditions to reduce resend attempts and improve recovery time.
1935 Description: llapi_file_create() does not allow some changes
1936 Details : add llapi_file_open() that allows specifying the mode and
1937 open flags, and also returns an open file handle.
1941 Description: merge and cleanup kernel patches.
1942 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1946 Description: Add journal checksum support.(Kernel part)
1947 Details : The journal checksum feature adds two new flags i.e
1948 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1949 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1950 indicates that the commit block contains the checksum for
1951 the blocks described by the descriptor blocks. Now commit
1952 record can be sent to disk without waiting for descriptor
1953 blocks to be written to disk. This behavior is controlled
1954 using JBD2_FEATURE_ASYNC_COMMIT flag.
1958 Description: OSS needs mutliple precreate threads
1959 Details : Add ability to start more than one create thread per OSS.
1963 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1964 Details : execute lfs setstripe on client
1968 Description: mds_obd_create error creating tmp object
1969 Details : When the user sets quota on root, llog will be affected and can't
1970 create files and write files.
1973 Frequency : Always on ia64 patchless client, and possibly others.
1975 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1976 Details : This allows the patchless client to be loaded on architectures
1977 without this export.
1981 Description: RedHat Update kernel for RHEL5
1982 Details : Add the kernel config file for RHEL5.
1986 Description: RedHat Update kernel for RHEL5
1987 Details : Modify the kernel config file more closer RHEL5.
1991 Description: Build failure against Centos5 (RHEL5)
1992 Details : Define PAGE_SIZE when it isn't present.
1996 Description: test_7 on sanity-quota.sh can't be used on separate machines
1997 Details : add facet to handle it
2001 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2002 Details : with NFS, the anon dentry's parent was set to itself in
2003 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2004 corresponding dentry other than use rec->ur_name.
2006 Severity : enhancement
2008 Description: lfs setstripe enhancement
2009 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2013 Description: enable data checksumming by default
2014 Details : enable checksum by default, allow --disable-checksum
2015 configure option and "-o nochecksum" mount option. Checksums
2016 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2020 Description: lustre support for RHEL5
2021 Details : Add support for RHEL5.
2025 Description: Client eviction due to failover config
2026 Details : after a connection loss, the lustre client should attempt to
2027 reconnect to the last active server first before trying the
2028 other potential connections.
2032 Description: when mds and osts use different quota unit(32bit and 64bit),
2033 quota will be released repeatly.
2034 Details : void sending multiple quota reqs to mds, which will keep the
2035 status between the reqs.
2039 Description: osts not allocated evenly to files
2040 Details : change the condition to increase offset_idx
2043 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2045 Description: Data corruption for OSTs that are formatted larger than 2TB
2047 Details : When generating the bio request for lustre file writes the
2048 sector number would overflow a temporary variable before being
2049 used for the IO. The data reads correctly from Lustre (which
2050 will overflow in a similar manner) but other file data or
2051 filesystem metadata may be corrupted in some cases.
2055 Description: Tune the kernel for good SCSI performance.
2056 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2057 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2062 Description: rewrite ext2-derived code in obdclass/uuid.c
2063 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2064 from scratch preserving functionality.
2068 Description: "lfs find -obd UUID" prints directories
2069 Details : "lfs find -obd UUID" will return all directory names instead
2070 of just file names. It is incorrect because the directories
2071 do not reside on the OSTs.
2075 Description: use i_size_read and i_size_write in 2.6 port
2076 Details : replace inode->i_size access with i_size_read/write()
2079 Frequency : when removing large files
2081 Description: scheduling issue during removal of large Lustre files
2082 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2083 It causes scheduling issues when removing large files (17TB in the
2087 Frequency : only with liblustre clients on XT3
2089 Description: evictions taking too long
2090 Details : allow llrd to evict clients directly on OSTs
2093 Frequency : only on ppc
2095 Description: /proc/fs/lustre/devices broken on ppc
2096 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2097 We should make sure the type of 'index' is loff_t and then cast
2098 explicitly as needed below. Do not assign an explicitly cast
2102 Frequency : only for rhel5
2104 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2105 Details : Modify the target file & which_kernel.
2107 Severity : enhancement
2109 Description: omit set fsid for export NFS
2110 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2111 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2115 Frequency : Always, for kernels after 2.6.16
2116 Description: Fix warning idr_remove called for id=.. which is not allocated.
2117 Details : Last kernels save old s_dev before kill super and not allow
2118 to restore from callback - restore it before call kill_anon_super.
2122 Description: buffer overruns could theoretically occur
2123 Details : llapi_semantic_traverse() modifies the "path" argument by
2124 appending values to the end of the origin string, and a
2125 overrun may occur. Adding buffer overrun check in liblustreapi.
2129 Description: Fix error on 'ls .' at the top of the Lustre mount.
2130 Details : Don't revalidate dentry if it is a root dentry.
2134 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2135 Details : Modify vm-tunables-rhel4.patch.
2139 Description: Kernel config for 2.6.18-vanilla.
2140 Details : Modify targets/2.6-vanilla.target.in.
2141 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2142 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2143 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2144 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2148 Description: improve handling recoverable errors
2149 Details : if request processig with error which can be recoverable on server
2150 request should be resend, otherwise page released from cache and
2155 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2156 Details : Modify target file & which_patch.
2157 A flaw was found in the IA32 system call emulation provided
2158 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2159 value could be stored in the %RAX register, which could trigger an
2160 out-of-bounds system call table access. An untrusted local user
2161 could exploit this flaw to run code in the kernel
2162 (ie a root privilege escalation). (CVE-2007-4573).
2166 Description: O_DIRECT bypasses client statistics.
2167 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2168 read_ahead_stats were updated. copied the stats section from
2169 osc_send_oap_rpc() into async_internal().
2173 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2177 Description: change order of libsysio includes
2178 Details : '#include sysio.h' should always come before '#include xtio.h'
2182 Description: handle "serious error: objid * already exists" more gracefully
2183 Details : If LAST_ID value on disk is smaller than the objects existing in
2184 the O/0/d* directories, it indicates disk corruption and causes an
2185 LBUG(). If the object is 0-length, then we should use the existing
2186 object. This will help to avoid a full fsck in most cases.
2188 Severity : enhancement
2190 Description: adapt the lustre_config script to support the upgrade case
2191 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2192 server targets to 1.6 in parallel.
2196 Description: To avoid grant space > avaible space when the disk is almost
2197 full. Without this patch you might see the error "grant XXXX >
2198 available" or some LBUG about grant, when the disk is almost
2200 Details : In filter_check_grant, for non_grant cache write, we should
2201 check the left space by if (*left > ungranted + bytes), instead
2202 of (*left > ungranted), because only we are sure the left space
2203 is enough for another "bytes", then the ungrant space should be
2204 increase. In client, we should update cl_avail_grant only there
2205 is OBD_MD_FLGRANT in the reply.
2209 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2210 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2211 described in CVE-2007-4573. This problem could allow untrusted
2212 local users to gain root access.
2215 Frequency : when using O_DIRECT and quotas
2217 Description: Incorrect file ownership on O_DIRECT output files
2218 Details : block usage reported by 'lfs quota' does not take into account
2219 files that have been written with O_DIRECT.
2224 Description: touch file failed when fs is not full
2225 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2226 otherwise we can get ENOSP while fs is not full.
2230 Description: parallel lock callbacks
2231 Details : Instead of sending blocking and completion callbacks as separated
2232 requests, adding them to a set and sending in parallel.
2235 Frequency : only for Cray XT3
2236 Bugzilla : 12829/13455
2237 Description: Changing primary group doesn't change the group lustre assigns to
2239 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2240 overridden with the primary group provided by the group upcall,
2241 whereas the supplied fsgid can be trusted if it is in the list of
2242 supplementary groups returned by the group upcall.
2244 Severity : enhancement
2246 Description: Allow masking D_WARNING, D_ERROR messages from console
2247 Details : Console messages can now be disabled via lnet.printk.
2251 Description: User code with malformed file open parameter crashes client node
2252 Details : Before packing join_file req, all the related reference should be
2253 checked carefully in case some malformed flags cause fake
2254 join_file req on client.
2258 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2259 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2260 mode is equal to the granted mode. However, it should also check
2261 that LDLM_FL_CP_REQD is not set.
2265 Description: Heavy nfs access might result in deadlocks
2266 Details : After ELC code landed, it is now improper to enqueue any mds
2267 locks under och_sem, because enqueue might want to decide to
2268 cancel open locks for same inode we are holding och_sem for.
2272 Description: Client eviction while running blogbench
2273 Details : A lot of unlink operations with concurrent I/O can lead to a
2274 deadlock causing evictions. To address the problem, the number of
2275 oustanding OST_DESTROY requests is now throttled to
2276 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2277 callbacks are processed in priority.
2281 Description: enable ACLs on MDS by default
2282 Details : ACLs must be enabled on MDS by default.
2285 Frequency : PPC/PPC64 only
2287 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2288 Details : fix duplicated definitions between asm-ppc64/types.h and
2289 lustre_types.h on PPC.
2292 Frequency : PPC/PPC64 only
2294 Description: asm-ppc/segment.h does not exist
2295 Details : fix compile issue on PPC.
2299 Description: better handle error messages in extents code
2302 Frequency : RHEL4 only
2304 Description: mkfs is very slow on IA64/RHEL4
2305 Details : A performance regression has been discovered in the MPT Fusion
2306 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2307 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2308 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2310 Severity : enhancement
2312 Description: SNMP support enhancement
2313 Details : Adding total number of sampled request for an MDS node in snmp
2316 Severity : enhancement
2318 Description: Optimize ldlm waiting list processing for PR extent locks
2319 Details : When processing waiting list for read extent lock and meeting read
2320 lock that is same or wider to it that is not contended, skip
2321 processing rest of the list and immediatelly return current
2322 status of conflictness, since we are guaranteed there are no
2323 conflicting locks in the rest of the list.
2327 Description: Time out and refuse to reconnect
2328 Details : When the failover node is the primary node, it is possible
2329 to have two identical connections in imp_conn_list. We must
2330 compare not conn's pointers but NIDs, otherwise we can defeat
2331 connection throttling.
2335 Description: port llog fixes from b1_6 into HEAD
2336 Details : Port llog reference couting and some llog cleanups from b1_6
2337 (bug 10800) into HEAD, for protect from panic and access to already
2338 free llog structures.
2342 Description: Detect stride IO mode in read-ahead
2343 Details : When a client does stride read, read-ahead should detect that and
2344 read-ahead pages according to the detected stride pattern.
2348 Description: data checksumming impacts single node performance
2349 Details : add support for several checksum algorithm. Currently, only CRC32
2350 and Adler-32 are supported. The checksum type can be changed on
2351 the fly via /proc/fs/lustre/osc/*/checksum_type.
2355 Description: use adler32 for page checksums
2356 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2361 Description: build for x2 fails
2362 Details : fix compile issue on Cray systems.
2366 Description: Properly match for duplicate locks
2367 Details : Due to different lock order from skiplists code, we need to
2368 traverse entire list for now
2371 Frequency : only on PPC/SLES10
2373 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2374 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2375 be defined. Add a hack in mkfs_lustre.c to work around this compile
2380 Description: LASSERT on MDS when client holding flock lock dies
2381 Details : ldlm pool logic depends on number of granted locks equal to
2382 number of released locks which is not true for flock locks, so
2383 just exclude such locks from consideration.
2387 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2388 Details : Use fsfilt_sync() for both the whole filesystem sync and
2389 individual file sync to eliminate dangerous inode locking
2390 with I_LOCK that can lead to a deadlock.
2394 Description: performance in 1.6.3
2395 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2396 sure that 1MB requests are not fragmented by the block layer.
2398 Severity : enhancement
2400 Description: organize the server-side client stats on per-nid basis
2401 Details : Change the structure of stats under obdfilter and mds to
2409 The "uuid"s file would list the uuids of _active_ exports.
2410 And the clear entry is to clear all stats and stale nids.
2412 Severity : enhancement
2414 Description: eliminate client locks in face of contention
2415 Details : file contention detection and lockless i/o implementation
2416 for contended files.
2420 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2423 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2425 Description: MDS may lose file striping (and hence file data) in some cases
2426 Details : If there are additional extended attributes stored on the MDS,
2427 in particular ACLs, SELinux, or user attributes (if user_xattr
2428 is specified for the client mount options) then there is a risk
2429 of attribute loss. Additionally, the Lustre file striping
2430 needs to be larger than default (e.g. striped over all OSTs),
2431 and an additional attribute must be stored initially in the
2432 inode and then increase in size enough to be moved to the
2433 external attribute block (e.g. ACL growing in size) for file
2438 Description: skiplist implementation simplification
2439 Details : skiplists are used to group compatible locks on granted list
2440 that was implemented as tracking first and last lock of each
2441 lock group the patch changes that to using doubly linked lists
2445 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2446 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2447 because LASSERT on both the data supplied by a client, and the
2448 data on disk is dangerous and incorrect.
2450 Severity : enhancement
2452 Description: Slow truncate/writes to huge files at high offsets.
2453 Details : Directly associate cached pages to lock that protect those pages,
2454 this allows us to quickly find what pages to write and remove
2455 once lock callback is received.
2459 Description: more ldlm soft lockups
2460 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2461 starve other threads from the resource lock for a long time in
2462 case of long waiting queue, so change the debug level from
2463 D_OTHER to the less frequently used D_INFO.
2465 Severity : enhancement
2467 Description: add -gid, -group, -uid, -user options to lfs find
2471 Description: Hung threads in invalidate_inode_pages2_range
2472 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2473 once one is completed. As a result, some RPCs are stuck in the
2474 queue and are never sent.
2478 Description: filter threads hungs on waiting journal commit
2479 Details : Cleanup filter group llog code, then only filter group llog will
2480 be only created in the MDS/OST syncing process.
2484 Description: Procfs and llog threads access destoryed import sometimes.
2485 Details : Sync the import destoryed process with procfs and llog threads by
2486 the import refcount and semaphore.
2488 Severity : enhancement
2490 Description: openlock cache of b1_6 port to HEAD
2495 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2496 Details : If insertion of an extent fails, then discard the inode
2497 preallocation and free data blocks else it can lead to duplicate
2502 Description: don't always update ctime in ext3_xattr_set_handle()
2503 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2504 In some cases the ctime should not be updated, for example for
2505 2.0->1.8 compatibility it is necessary to delete an xattr and it
2506 should not update the ctime.
2510 Bugzilla : 15713/16362
2511 Description: Assertion in iopen_connect_dentry in 1.6.3
2512 Details : looking up an inode via iopen with the wrong generation number can
2513 populate the dcache with a disconneced dentry while the inode
2514 number is in the process of being reallocated. This causes an
2515 assertion failure in iopen since the inode's dentry list contains
2516 both a connected and disconnected dentry.
2520 Description: assertion failure in ldlm_handle2lock()
2521 Details : fix a race between class_handle_unhash() and class_handle2object()
2522 introduced in lustre 1.6.5 by bug 13622.
2527 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2528 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2533 Description: oops in sd_iostats_seq_show()
2534 Details : unloading/reloading the scsi low level driver triggers a kernel
2535 bug when trying to access the sd iostat file.
2540 Description: Kernel panics during QLogic driver reload
2541 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2542 patch, causing memory corruption.
2547 Description: journal_dev option does not work in b1_6
2548 Details : pass mount option during pre-mount.
2550 Severity : enhancement
2552 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2553 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2554 extent information of a file. It can be used to map logical blocks
2555 in a file to physical blocks in the block device.
2559 Description: LDLM soft lockups - improvement
2560 Details : It is be possible to send the lock handle along with each read
2561 or write request because the client is already doing a lock match
2562 itself so there isn't any reason the OST should have to re-do that
2566 Frequency : only on Cray X2
2568 Description: X2 build failures
2569 Details : fix build failures on Cray X2.
2573 Description: xid & resent requests
2574 Details : Initialize RPC XID from clock at startup (randomly if clock is
2577 Severity : enhancement
2579 Description: Add lustre_start utility to start or stop multiple Lustre servers
2584 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2585 Details : In case of memory pressure, list_del() can be called twice on
2586 req->rq_history_list, causing a kernel oops.
2590 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2591 Details : fix stack overflow in the distributed lock manager by defering
2592 export eviction after a failed AST to the elt thread instead of
2593 handling it in the dlm interpret routine.
2597 Description: Convert some comments to new format.
2598 Details : Update documenting comments to match doxygen conventions.
2602 Description: Grammar fixes.
2603 Details : A couple of trivial spelling fixes.
2607 Description: OSD_COUNTERS-mandatory
2608 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2609 to simplify the code.
2613 Description: simplify cmm_device freeing logic.
2614 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2618 Description: Add lockdep support to dt_object_operations locking interface.
2619 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2620 parameter indicating lock ordering. Update mdd code to use new
2625 Description: Introduce failloc constants for lockless IO tests.
2626 Details : Add two new failloc constants to test lockless IO. Only one of
2627 them in implemented---another is checked in yet to be landed
2632 Description: Add lockdep support for inode mutex.
2633 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2634 in the situations where Lustre has to lock more than one inode
2639 Description: Add optional invariants checking support.
2640 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2641 configure switch. This macro is to be used for consistency and
2642 sanity checks that are too expensive to be left in `production'
2647 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2648 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2653 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2654 Details : Use spin_lock_nested() in (the only) situation where more than
2655 one ldlm_lock is locked simultaneously. Also, fix possible
2656 dead-lock in ldlm_lock_change_resource() by enforcing particular
2661 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2662 Details : Instead of specifying each ldlm_lock call-back through separate
2663 parameter, wrap them into struct ldlm_callback_suite.
2667 Description: Kill join_lru obd method and its callers.
2668 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2669 in memory---a requirement implemented through ->o_join_lru() obd
2670 method in HEAD. Remove this method and its users.
2674 Description: Add asynchronous ldlm ENQUEUE completion handler.
2675 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2676 case that stock ldlm_completion_ast() cannot handle as it waits
2677 until lock is granted. Introduce new ldlm_completion_ast_async()
2678 for this. Also comment ldlm_completion_ast().
2682 Description: ldlm_error <-> errno conversion.
2683 Details : Add functions to map (rather arbitrary) between LDLM error codes
2684 and standard errno values. CLIO needs this to prevent LDLM specific
2685 constants from escaping ldlm and osc.
2689 Description: Kill unused ldlm_handle2lock_ns() function.
2690 Details : Kill unused ldlm_handle2lock_ns() function.
2694 Description: Add lu_ref support to ldlm_lock
2695 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2696 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2697 and ldlm_resource. LDLM interface has to be changed, because code that
2698 releases a reference on a lock, has to "know" what reference this is.
2699 In the most frequent case
2701 lock = ldlm_handle2lock(handle);
2703 LDLM_LOCK_PUT(lock);
2705 no changes are required. When any other reference (received _not_ from
2706 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2707 instead of LDLM_LOCK_PUT().
2709 Arguably, changes are pervasive, and interface requires some discipline
2710 for proper use. On the other hand, it was very instrumental in finding
2711 a few leaked lock references.
2715 Description: Add ldlm_lock_addref_try().
2716 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2717 attempts to addref a lock that might be being canceled
2722 Description: Add ldlm_weigh_callback().
2723 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2724 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2725 hard-coded `number of pages' logic.
2729 Description: Add lockdep annotations to llog code.
2730 Details : Use appropriately tagged _nested() locking calls in the places
2731 where llog takes more than one ->lgh_lock lock.
2735 Description: Add loi_kms_set().
2736 Details : Wrap kms updates into a helper function.
2740 Description: Constify instances of struct lsm_operations.
2741 Details : Constify instances of struct lsm_operations.
2745 Description: lu_conf support.
2746 Details : On a server, a file system object is uniquely identified
2747 by a fid, which is sufficient to locate and load all object
2748 state (inode). On a client, on the other hand, more data are
2749 necessary instantiate an object. Change lu_object_find() and
2750 friends to take additional `lu_conf' argument describing object.
2751 Typically this includes layout information.
2755 Description: lu_context fixes.
2756 Details : Introduce new lu_context functions that are needed on the client
2757 side, where some system threads (ptlrpcd) are shared by multiple
2758 modules, and so cannot be stopped during module shutdown.
2762 Description: Add start and stop methods to lu_device_type_operations.
2763 Details : Introduce two new methods in lu_device_type_operations, that are
2764 invoked when first instance of a given type is created and last one
2765 is destroyed respectively. This is need by CLIO.
2769 Description: Add lu_ref support to struct lu_device.
2770 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2771 track leaked references.
2775 Description: Introduce lu_kmem_descr.
2776 Details : lu_kmem_descr and its companion interface allow to create
2777 and destroy a number of kmem caches at once.
2781 Description: Fix lu_object finalization race.
2782 Details : Fix a race between lu_object_find() finding an object and its
2783 concurrent finalization. This race is (most likely) not possible
2784 on the server, but might happen on the client.
2788 Description: Introduce lu_ref interface.
2789 Details : lu_ref is a debugging module allowing to track references to
2790 a given object. It is quite cpu expensive, and has to be
2791 explicitly enabled with --enable-lu_ref. See usage description
2796 Description: Factor lu_site procfs stats into a separate function.
2797 Details : Separate lu_site stats printing code into a separate function
2798 that can be reused on a client.
2802 Description: Constify instances of struct {lu,dt,md}_device_operations.
2803 Details : Constify instances of struct {lu,dt,md}_device_operations.
2807 Description: Introduce struct md_site and move meta-data specific parts of
2808 struct lu_site here.
2809 Details : Move md-specific fields out of struct lu_site into special struct
2810 md_site, so that lu_site can be used on a client.
2814 Description: Kill mdd_lov_destroy().
2815 Details : Remove unused mdd code.
2819 Description: Add st_block checking to multistat.c.
2820 Details : Add st_block checking to multistat.c.
2824 Description: Add lu_ref support to struct obd_device.
2825 Details : Add lu_ref tracking to obd_device.
2829 Description: Kill obd_set_fail_loc().
2830 Details : Remove unused code.
2834 Description: Add special type for ptlrpc_request interpret functions.
2835 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2836 there. Actual usage will be in CLIO.
2840 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2841 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2842 doesn't work with lockdep.
2846 Description: Add rwv.c test program.
2847 Details : New testing program exercising readv(2) and writev(2) (Qian).
2851 Description: Add sendfile.c test program.
2852 Details : New testing program exercising sendfile(2) (Jay).
2856 Description: Ratelimit a message that can be very frequent.
2857 Details : Ratelimit a memory allocation failure message that can
2862 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2863 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2867 Description: Liblustre build fixes.
2868 Details : Liblustre build fixes.
2872 Description: libcfs: add cfs_{need,cond}_resched() interface.
2873 Details : libcfs: add cfs_{need,cond}_resched() definition and
2874 implementations for Linux, NT, and liblustre.
2876 Severity : enhancement
2878 Description: More exported tunables for mballoc
2879 Details : Add support for tunable preallocation window and new tunables for
2880 large/small requests
2884 Description: Detect corruption of block bitmap and checking for preallocations
2885 Details : Checks validity of on-disk block bitmap. Also it does better
2886 checking of number of applied preallocations. When corruption is
2887 found, it turns filesystem readonly to prevent further corruptions.
2891 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2892 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2893 window to 0. If it does hit cache miss, and read-pattern is still
2894 stride-io mode, does not reset the stride window, but also does not
2895 increase the stride window length in this case.
2899 Frequency : only for big-endian servers
2900 Description: Check if system is big-endian while mounting fs with extents feature
2901 Details : Mounting a filesystem with extents feature will fail on big-endian
2902 systems since ext3-based ldiskfs is not supported on big-endian
2903 systems. This can be over-riden with "bigendian_extents" mount option.
2905 Severity : enhancement
2907 Description: The root squash functionality
2908 Details : A security feature, which is to prevent users from being able
2909 to mount lustre on their desktop, run as root, and delete
2910 all of the files in the filesystem. The goal is accomplished by
2911 remapping user id (UID) and group id (GID) of the root user to
2912 a UID and GID specified by the system administartor via Lustre
2913 configuration management server (MGS). The functionality also
2914 allows to specify sets of clients for which the remapping does
2919 Description: Excessive recovery window
2920 Details : With AT enabled, the recovery window can be excessively long (6000+
2921 seconds). To address this problem, we no longer use
2922 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2923 timeout no longer depends on the service time, it is set to
2924 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2925 time via pb_service_time.
2929 Description: Watchdog triggered on MDS failover
2930 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2931 the OSTs know that the MDS "UUID" can be reused for the same export
2932 from a different NID, so we do not need to wait for the export to be
2936 Frequency : rare, only if using MMP with Linux RAID
2938 Description: MMP doesn't work with Linux RAID
2939 Details : While using HA for Lustre servers with Linux RAID, it is possible
2940 that MMP will not detect multiple mounts. To make this work we
2941 need to unplug the device queue in RAID when the MMP block is being
2942 written. Also while reading the MMP block, we should read it from
2943 disk and not the cached one.
2945 Severity : enhancement
2947 Description: open file using fid
2948 Details : A file can be opened using just its fid, like
2949 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2952 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2955 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2956 Details : add proper locking for creation of jbd_4k slab cache
2960 Description: MMP check in ext3_remount() fails without displaying any error
2961 Details : When multiple mount protection fails during remount, proper error
2964 Severity : enhancement
2966 Description: Allow stripe size to be up to 4G-64k
2967 Details : Fix math logic to allow large stripe sizes.
2971 Description: add check for >8TB ldiskfs filesystems
2972 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2973 Don't allow >8TB ldiskfs filesystems to be mounted without
2974 force_over_8tb mount option
2979 Descriptoin: OSS complains "too many blocks" and crashes
2980 Details : In filter_commitrw_write():
2981 1. lquota_chkquota();
2982 2. obdo_from_inode();
2983 3. lquota_pending_commit();
2984 "2." has changed uid and gid in this case, then caused
2985 lquota_chkquota() and lquota_pending_commit() to use
2986 different uid and gid. That is the root of the bug.
2990 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2991 Details : Detect and print out the 'stripe count = -1' case properly.
2995 Description: Out or order replies might be lost on replay
2996 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2997 request with tid smaller than one currently being added, add it
2998 to the start, not end of the list.
3000 --------------------------------------------------------------------------------
3002 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3004 * Support for kernels:
3005 2.6.5-7.283 (SLES 9),
3006 2.6.9-55.EL (RHEL 4),
3007 2.6.16.46-0.14 (SLES 10),
3008 2.6.18.8 vanilla (kernel.org)
3009 * Client support for unpatched kernels:
3010 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3011 2.6.16 - 2.6.22 vanilla (kernel.org)
3012 * Due to recently discovered recovery problems, we do not recommend
3013 using patchless RHEL 4 clients with this or any earlier release.
3014 * Recommended e2fsprogs version: 1.39.cfs8
3015 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3016 * Starting with this release, the ldiskfs backing filesystem required
3017 by Lustre is now in its own package, lustre-ldiskfs. This package
3018 should be installed. It is versioned separately from Lustre and
3019 may be released separately in future.
3023 Description: block reactivating mgc import until all deactivates complete
3024 Details : Fix race when failing back MDT/MGS to itself (testing)
3027 Frequency : at statup only
3029 Description: mds_lov_synchronize race leads to various problems
3030 Details : simultaneous MDT->OST connections at startup can cause the
3031 sync to abort, leaving the OSC in a bad state.
3033 Severity : enhancement
3035 Description: add optional extra BUILD_VERSION info
3036 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3037 to override the lustre version.
3040 Frequency : 2.6.18 servers only
3042 Description: ll_kern_mount() doesn't release the module reference
3043 Details : The ldiskfs module reference count never drops down to 0
3044 because ll_kern_mount() doesn't release the module reference.
3049 Description: server LBUG when using old ost_num_threads parameter
3050 Details : Accept the old ost_num_threads parameter but warn that it
3051 is deprecated, and fix an off-by-one error that caused an LBUG.
3056 Description: Transient SCSI error results in persistent IO issue
3057 Details : iobuf->dr_error is not reinitialized to 0 between two
3061 Frequency : sometimes when underlying device returns I/O errors
3063 Description: OSTs not going read-only during write failures
3064 Details : OSTs are not remounted read-only when the journal commit threads
3065 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3066 instead of the ext3 wrappers.
3070 Description: poor connect scaling with increasing client count
3071 Details : Don't run filter_grant_sanity_check for more than 100 exports
3072 to improve scaling for large numbers of clients.
3075 Frequency : SLES10 only
3077 Description: sanity-quota.sh quotacheck failed: rc = -22
3078 Details : Quotas cannot be enabled on SLES10.
3081 Frequency : liblustre clients only
3083 Description: getdirentries does not give error when run on compute nodes
3084 Details : getdirentries does not fail when the size specified as an argument
3085 is too small to contain at least one entry
3087 Severity : enhancement
3089 Description: Add LNET router traceability for debug purposes
3090 Details : If a checksum failure occurs with a router as part of the
3091 IO path, the NID of the last router that forwarded the bulk data
3092 is printed so it can be identified.
3097 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3098 Details : Due to a race condition, liblustre clients were occasionally
3099 evicted incorrectly.
3101 Severity : enhancement
3103 Description: lfs setstripe use optional parameters instead of postional
3106 Severity : enhancement
3108 Description: Nanosecond timestamp support for ldiskfs
3109 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3110 resolution timestamps. There is not yet support for this at
3111 the Lustre filesystem level.
3114 Frequency : during server recovery
3116 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3117 Details : request with rq_no_resend flag not awake l_wait_event if they get
3121 Frequency : nfs export on patchless client
3123 Description: connectathon hang when test nfs export over patchless client
3124 Details : Disconnected dentry cannot be found with lookup, so we do not need
3125 to unhash it or make it invalid
3128 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3129 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3130 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3135 Description: open req refcounting wrong on reconnect
3136 Details : If reconnect happened between getting open reply from server and
3137 call to mdc_set_replay_data in ll_file_open, we will schedule
3138 replay for unreferenced request that we are about to free.
3139 Subsequent close will crash in variety of ways.
3140 Check that request is still eligible for replay in
3141 mdc_set_replay_data().
3146 Description: disable writes to filesystem when reading health_check file
3147 Details : the default for reading the health_check proc file has changed
3148 to NOT do a journal transaction and write to disk, because this
3149 can cause reads of the /proc file to hang and block HA state
3150 checking on a healthy but otherwise heavily loaded system. It
3151 is possible to return to the previous behaviour during configure
3152 with --enable-health-write.
3154 Severity : enhancement
3156 Description: 64-bit inode version
3157 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3158 to the inode. This will be required for version-based recovery.
3163 Description: MDS fails to start if a duplicate client export is detected
3164 Details : in some rare cases it was possible for a client to connect to
3165 an MDS multiple times. Upon recovery the MDS would detect this
3166 and fail during startup. Handle this more gracefully.
3168 Severity : enhancement
3170 Description: Add -o localflock option to simulate old noflock
3172 Details : This will achieve local-only flock/fcntl locks
3178 Description: log_commit_thread vs filter_destroy race leads to crash
3179 Details : Take import reference before releasing llog record semaphore
3184 Description: Wrong request locking in request set processing
3185 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3186 imp_delayed_list, in this place should be used imp_lock.
3189 Frequency : when reconnection
3191 Description: Grant Leak when osc reconnect to OST
3192 Details : When osc reconnect ost, OST(filter) should check whether it
3193 should grant more space to client by comparing fed_grant and
3194 cl_avail_grant, and return the granted space to client instead
3195 of "new granted" space, because client will call osc_init_grant
3196 to update the client grant space info.
3199 Frequency : when client reconnect to OST
3201 Description: Grant Leak when osc do resend and replay bulk write
3202 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3203 bulk write request, because the grant info while be sync between
3204 OSC and OST when reconnect, and we should ignore the grant info
3205 these of resend/replay write req.
3210 Description: Grant space more than available left space sometimes.
3211 Details : When then OST is about to be full, if two bulk writing from
3212 different clients came to OST. Accord the avaliable space of the
3213 OST, the first req should be permitted, and the second one
3214 should be denied by ENOSPC. But if the seconde arrived before
3215 the first one is committed. The OST might wrongly permit second
3216 writing, which will cause grant space > available space.
3219 Frequency : when client is evicted
3221 Description: Grant might be wrongly erased when osc is evicted by OST
3222 Details : when the import is evicted by server, it will fork another
3223 thread ptlrpc_invalidate_import_thread to invalidate the
3224 import, where the grant will be set to 0. While the original
3225 thread will update the grant it got when connecting. So if
3226 the former happened latter, the grant will be wrongly errased
3227 because of this race.
3232 Description: Checking Stale with correct fid
3233 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3234 to check whether it is stale, because sometimes, we want the
3235 enqueue happened anyway, and op_data.fid2 will not be initialized.
3237 Severity : enhancement
3239 Description: update patchless client
3240 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3243 Frequency : only with 2.4 kernel
3245 Description: random memory corruption
3246 Details : size of struct ll_inode_info is to big for union inode.u and this
3247 can be cause of random memory corruption.
3252 Description: Memory leak in recovery
3253 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3254 It should also check obd_fail before fsfilt_start, otherwise if
3255 fsfilt_start return -EROFS,(failover mds during mds recovery).
3256 then the req will return with repmsg->transno = 0 and rc = EROFS.
3257 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3258 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3259 be freed no matter whether fsfilt_commit success or not.
3262 Frequency : only with huge count clients
3264 Description: Prevents from taking the superblock lock in llap_from_page for
3266 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3267 for llap_from_page call in ll_removepage prevents from taking the
3268 superblock lock for a soon died page.
3273 Description: Not check open intent error before release open handle
3274 Details : in some rare cases, the open intent error is not checked before
3275 release open handle, which may cause
3276 ASSERTION(open_req->rq_transno != 0), because it tries to release
3277 the failed open handle.
3282 Description: Set cat log bitmap only after create log success.
3283 Details : in some rare cases, the cat log bitmap is set too early. and it
3284 should be set only after create log success.
3288 Description: Accessing a block bevice can re-enable I/O when Lustre is
3289 tearing down a device.
3290 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3295 Description: service threads may hog cpus when there are a lot of requests
3296 Details : Insert cond_resched to give other threads a chance to use some CPU
3301 Description: the cat log was not initialized in recovery
3302 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3303 unlink log on mds will not be initialized until mds post
3304 recovery. And also in mds post recovery, the unlink log will
3305 initialization will be done asynchronausly, so there will be race
3306 between add unlink log and unlink log initialization.
3310 Description: brw_stats were being printed incorrectly
3311 Details : brw_stats were being printed as log2 but all of them were not
3312 recorded as log2. Also remove some code duplication arising from
3313 filter_tally_{read,write}.
3317 Frequency : rare, only in recovery.
3318 Description: ASSERTION(req->rq_type != LI_POISON) failed
3319 Details : imp_lock should be held while iterating over imp_sending_list for
3320 prevent destroy request after get timeout in ptlrpc_queue_wait.
3324 Description: replay-single.sh test 52 fails
3325 Details : A lock's skiplist need to be cleanup when it being unlinked
3326 from its resource list.
3330 Description: Short directio read returns full requested size rather than
3332 Details : Direct I/O operations should return actual amount of bytes
3333 transferred rather than requested size.
3335 Severity : enhancement
3337 Description: metadata RPC reduction (e.g. for rm performance)
3338 Details : decrease the amount of synchronous RPC between clients and servers
3339 by canceling conflicing lock before the operation on the client
3340 and packing thier handles into the main operation RPC to server.
3342 Severity : enhancement
3344 Description: Async OSC create to avoid the blocking unnecessarily.
3345 Details : If an OST has no remaining object, system will block on the
3346 creation when it needs to create a new object on this OST. Now,
3347 ways use pre-created objects when available, instead of blocking on
3348 an empty osc while others are not empty. If we must block, we
3349 block for the shortest possible period of time.
3353 Description: improve handling recoverable errors
3354 Details : if request processig with error which can be recoverable on server
3355 request should be resend, otherwise page released from cache and
3358 Severity : enhancement
3360 Description: refine locking for avoid write wrong info into lov_objid file
3361 Details : fix possible races with add new target and write/update data in
3364 --------------------------------------------------------------------------------
3366 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3371 Frequency : on some architectures
3373 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3374 Details : Uninitialized flags sometimes cause configuration commands to
3378 Frequency : patchless clients only
3380 Description: missing __iget() symbol export
3381 Details : The __iget() symbol export is missing. To avoid the need for
3382 this on patchless clients the deathrow inode reaper is turned
3383 off, and we depend on the VM to clean up old inodes. This
3384 dependency was during via the fix for bug 12181.
3386 --------------------------------------------------------------------------------
3388 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3390 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3391 INTEROPERATE with older versions automatically. In many cases a
3392 special upgrade step is needed. Please read the
3393 user documentation before upgrading any part of a live system.
3394 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3395 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3396 * WARNING: Lustre configuration and startup changes are required with
3397 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3401 Severity : enhancement
3403 Description: MountConf
3404 Details : Lustre configuration is now managed via mkfs and mount
3405 commands instead of lmc and lconf. New obd types (MGS, MGC)
3406 are added for dynamic configuration management. See
3407 https://mail.clusterfs.com/wikis/lustre/MountConf for
3410 Severity : enhancement
3412 Description: dynamic OST addition
3413 Details : OSTs can now be added to a live filesystem
3415 Severity : enhancement
3417 Description: startup order invariance
3418 Details : MDTs and OSTs can be started in any order. Clients only
3419 require the MDT to complete startup.
3421 Severity : enhancement
3423 Description: parallel, asynchronous orphan cleanup
3424 Details : orphan cleanup is now performed in separate threads for each
3425 OST, allowing parallel non-blocking operation.
3427 Severity : enhancement
3429 Description: optimized stripe assignment
3430 Details : stripe assignments are now made based on ost space available,
3431 ost previous usage, and OSS previous usage, in order to try
3432 to optimize storage space and networking resources.
3434 Severity : enhancement
3436 Description: Permanently set tunables
3437 Details : All writable /proc/fs/lustre tunables can now be permanently
3438 set on a per-server basis, at mkfs time or on a live system.
3440 Severity : enhancement
3442 Description: Lustre message v2
3443 Details : Add lustre message format v2.
3445 Severity : enhancement
3447 Description: client OST exclusion list
3448 Details : Clients can be started with a list of OSTs that should be
3449 declared "inactive" for known non-responsive OSTs.
3453 Description: SPEC SFS validation failure on NFS v2 over lustre.
3454 Details : Changes the blocksize for regular files to be 2x RPC size,
3455 and not depend on stripe size.
3457 Severity : enhancement
3459 Description: Multiple MD RPCs in flight.
3460 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3461 To avoid overly-overloading MDS, introduce a limit on number of
3462 MDS RPCs in flight for a single client and add /proc controls
3463 to adjust this limit.
3465 Severity : enhancement
3467 Description: client read/write statistics
3468 Details : Add client read/write call usage stats for performance
3469 analysis of user processes.
3470 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3471 file access. extents_stats shows chunk size distribution.
3472 extents_stats_per_process show chunk size distribution per
3475 Severity : enhancement
3477 Description: mds statistics
3478 Details : Add detailed mds operations statistics in
3479 /proc/fs/lustre/mds/*/stats.
3483 Description: Failure of copying files with lustre special EAs.
3484 Details : Client side always return success for setxattr call for lustre
3485 special xattr (currently only "trusted.lov").
3490 Description: Refcount LNET uuids
3491 Details : The global LNET uuid list grew linearly with every startup;
3492 refcount repeated list entries instead of always adding to
3495 Severity : enhancement
3497 Description: Dynamic service threads
3498 Details : Within a small range, start extra service threads
3499 automatically when the request queue builds up.
3502 Frequency : mixed-endian client/server environments
3504 Description: mixed-endian crashes
3505 Details : The new msg_v2 system had some failures in mixed-endian
3508 Severity : enhancement
3510 Description: Easy OST removal
3511 Details : OSTs can be permanently deactivated with e.g. 'lctl
3512 conf_param lustre-OST0001.osc.active=0'
3514 Severity : enhancement
3516 Description: MGS proc entries
3517 Details : Added basic proc entries for the MGS showing what filesystems
3520 Severity : enhancement
3522 Description: provide MGS failover
3523 Details : Added config lock reacquisition after MGS server failover.
3525 Severity : enhancement
3527 Description: add Linux 2.4 support
3528 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3532 Description: a large application tries to do I/O to the same resource and dies
3533 in the middle of it.
3534 Details : Check the req->rq_arrival time after the call to
3535 ost_brw_lock_get(), but before we do anything about
3536 processing it & sending the BULK transfer request. This
3537 should help move old stale pending locks off the queue as
3538 quickly as obd_timeout.
3541 Frequency : when an incorrect nid is specified during startup
3543 Description: ptlrpc connect to non-existant node causes kernel crash
3544 Details : LNET can't be re-entered from an event callback, which
3545 happened when we expire a message after the export has been
3546 cleaned up. Instead, hand the zombie cleanup off to another
3549 Severity : enhancement
3551 Description: plain/inodebits lock performance improvement
3552 Details : Grouping plain/inodebits in granted list by their request modes
3553 and bits policy, thus improving the performance of search through
3557 Frequency : only if OST filesystem is corrupted
3559 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3560 Details : for a short time RPCs with bulk IO are in the replay list,
3561 but replay of bulk IOs is unimplemented. If the OST filesystem
3562 is corrupted due to disk cache incoherency and then replay is
3563 started it is possible to trip an assertion. Avoid putting
3564 committed RPCs into the replay list at all to avoid this issue.
3567 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3569 Description: System hang on startup
3570 Details : This bug allowed the liblustre (e.g. catamount) client to
3571 return to the app before handling all startup RPCs. This
3572 could leave the node unresponsive to lustre network traffic
3573 and manifested as a server ptllnd timeout.
3575 Severity : enhancement
3577 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3578 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3579 boolean to control whether to print peer debug info when a
3580 client's RPC times out.
3583 Frequency : only for kernels with patches from Lustre below 1.4.3
3585 Description: Remove old rdonly API
3586 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3589 Frequency : only for devices with external journals
3591 Description: Set external device read-only also
3592 Details : During a commanded failover stop, we set the disk device
3593 read-only while the server shuts down. We now also set any
3594 external journal device read-only at the same time.
3597 Frequency : when upgrading from 1.4 while trying to change parameters
3599 Description: The wrong (new) MDC name was used when setting parameters for
3600 upgraded MDT's. Also allows changing of OSC (and MDC)
3601 parameters if --writeconf is specified at tunefs upgrade time.
3604 Frequency : when setting specific ost indicies
3606 Description: QOS code breaks on skipped indicies
3607 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3608 created with --index need not be sequential.
3610 Severity : enhancement
3612 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3613 Details : The uninit_groups feature works in conjunction with the kernel
3614 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3615 up the pass1 processing of e2fsck. This is a read-only feature
3616 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3617 mount filesystems that have had this feature enabled.
3619 Severity : enhancement
3621 Description: Improve multi-block allocation algorithm to avoid fragmentation
3622 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3623 allocation locality and avoid filesystem fragmentation.
3625 ------------------------------------------------------------------------------
3627 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3629 * Support for kernels:
3630 2.4.21-47.0.1.EL (RHEL 3)
3631 2.6.5-7.283 (SLES 9)
3632 2.6.9-42.0.10.EL (RHEL 4)
3633 2.6.12.6 vanilla (kernel.org)
3634 2.6.16.27-0.9 (SLES 10)
3635 * Recommended e2fsprogs version: 1.39.cfs5
3637 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3641 Frequency : occasional, depends on client load and configuration
3642 Bugzilla : 12181, 12203
3643 Description: data loss for recently-modified files
3645 Details : In some cases it is possible that recently written or created
3646 files may not be written to disk in a timely manner (this should
3647 normally be within 30s unless client IO load is very high).
3648 The problem appears as zero-length files or files that are a
3649 multiple of 1MB in size after a client crash or client eviction
3650 that are missing data at the end of the file.
3652 This problem is more likely to be hit on clients where files are
3653 repeatedly created and unlinked in the same directory, clients
3654 have a large amount of RAM, have many CPUs, the filesystem has
3655 many OSTs, the clients are rebooted frequently, and/or the files
3656 are not accessed by other nodes after being written.
3658 The presence of the problem can be detected by looking at
3659 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3660 smaller than the second (nr_unused) then dirty files will not
3661 be flushed automatically to disk. "sync; sleep 10" should be
3662 run several times on the node before unmounting it to update
3663 Lustre (this is also safe to run on nodes without this problem).
3665 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3666 that can cause this same problem, so customers using that kernel
3667 also need to update the kernel in addition to Lustre. In order
3668 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3670 It is normal that files written just before a client crash (less
3671 than 30s) may not yet have been flushed to disk, even for local
3675 Frequency : frequent on thin XT3 nodes
3677 Description: UUID collision on thin XT3 Linux nodes
3678 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3679 randomly, since we relied on an insufficiently-seeded PRNG.
3684 Description: OSS hangs after "All ost request buffers busy"
3685 Details : A deadlock between quota and journal operations caused OSS
3686 hangs after printing "All ost request buffers busy."
3689 Frequency : always on liblustre builds
3691 Description: Cleanup compiler warnings on liblustre
3694 Frequency : always on liblustre builds on XT3
3696 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3700 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3701 > 32000 subdirectories
3702 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3703 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3704 subdirectory count crosses 32000. This will aid e2fsck to
3705 correctly handle more than 32000 subdirectories.
3708 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3710 Description: System hang on startup
3711 Details : This bug allowed the liblustre (e.g. catamount) client to
3712 return to the app before handling all startup RPCs. This
3713 could leave the node unresponsive to lustre network traffic
3714 and manifested as a server ptllnd timeout.
3716 Severity : enhancement
3718 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3719 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3720 boolean to control whether to print peer debug info when a
3721 client's RPC times out.
3726 Description: make O_SYNC working on 2.6 kernels
3727 Details : 2.6 kernels use different method for mark pages for write,
3728 so need add a code to lustre for O_SYNC work.
3733 Description: Failure to close file and release space on NFS
3734 Details : Put inode details into lock acquired in ll_intent_file_open.
3735 Use mdc_intent_lock in ll_intent_open to properly
3736 detect all kind of errors unhandled by mdc_enqueue.
3741 Description: proc file read during shutdown sometimes raced obd removal,
3743 Details : Add lock to prevent obd access after proc file removal.
3746 Frequency : Only for files larger than 4GB on 32-bit clients.
3748 Description: improperly doing page alignment of locks
3749 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3750 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3755 Description: RPCs being resent when they shouldn't be
3756 Details : Some RPCs that should not be resent are being resent. This
3757 can cause inconsistencies in the RPC state machine. Do not
3758 resend such requests.
3761 Frequency : rare, only with NFS export
3763 Description: Crash on NFS re-export node
3764 Details : under very unusual load conditions an assertion is hit in
3765 ll_intent_file_open()
3768 Frequency : only if OST filesystem is corrupted
3770 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3771 Details : for a short time RPCs with bulk IO are in the replay list,
3772 but replay of bulk IOs is unimplemented. If the OST filesystem
3773 is corrupted due to disk cache incoherency and then replay is
3774 started it is possible to trip an assertion. Avoid putting
3775 committed RPCs into the replay list at all to avoid this issue.
3780 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3781 Details : Large single O_DIRECT read and write calls can fail to allocate
3782 a sufficiently large buffer to process the request. In case of
3783 allocation failure the allocation is retried with a smaller
3784 buffer and broken into smaller requests.
3786 Severity : enhancement
3788 Description: Add -o localflock option to simulate old noflock behaviour.
3789 Details : This will achieve local-only flock/fcntl locks coherentness.
3794 Description: versioning check is incomplete
3795 Details : Checking the version difference of client vs. server, report
3796 error if the gap is too big.
3801 Description: add support PG_writeback bit
3802 Details : add support for PG_writeback bit for Lustre, for more carefull
3803 work with page cache in 2.6 kernel. This also fix some deadlocks
3804 and remove hack for work O_SYNC with 2.6 kernel.
3806 Severity : enhancement
3808 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3809 Details : The uninit_groups feature works in conjunction with the kernel
3810 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3811 up the pass1 processing of e2fsck. This is a read-only feature
3812 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3813 mount filesystems that have had this feature enabled.
3815 Severity : enhancement
3817 Description: Improve multi-block allocation algorithm to avoid fragmentation
3818 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3819 allocation locality and avoid filesystem fragmentation.
3821 ------------------------------------------------------------------------------
3823 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3825 * Support for kernels:
3826 2.6.9-42.0.3.EL (RHEL 4)
3827 2.6.5-7.276 (SLES 9)
3828 2.4.21-47.0.1.EL (RHEL 3)
3829 2.6.12.6 vanilla (kernel.org)
3830 2.6.16.21-0.8 (SLES10)
3831 * Recommended e2fsprogs version: 1.39.cfs2-0
3833 * The backwards-compatible /proc/sys/portals symlink has been removed
3834 in this release. Before upgrading, please ensure that you change
3835 any configuration scripts or /etc/sysctl.conf files that access
3836 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3837 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3838 in advance of the upgrade on any system running Lustre 1.4.6 or
3839 newer, since /proc/sys/lnet was added in that version.
3840 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3844 Frequency : only when quota is used
3846 Description: avoid scanning export list for quota master
3847 Details : Change the algorithms to avoid scanning export list in order
3848 to improve the efficiency.
3851 Frequency : MDS failover only, very rarely
3853 Description: "went back in time" messages on mds failover
3854 Details : The greatest transno may be lost when the current operation
3855 finishes with an error (transno==0) and the client's last_rcvd
3856 record is over-written. Save the greatest transno in the
3857 mds_last_transno for this case.
3860 Frequency : always for specific kernels and striping counts
3862 Description: client may get "Matching packet too big" without ACL support
3863 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3864 when trying to access files in certain configurations. The
3865 clients should in fact be denied when mounting because they do
3866 not understand ACLs.
3869 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3871 Description: many clients connecting with IO in progress causes connect timeouts
3872 Details : Avoid synchronous journal commits to avoid delays caused by many
3873 clients connecting/disconnecting when bulk IO is in progress.
3874 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3875 OST_IO_PORTAL to avoid delays behind potentially many pending
3879 Frequency : occasionally with multiple writers to a single file
3881 Description: shared writes to file may result in wrong size reported by stat()
3882 Details : Allow growing of kms when extent lock is cancelled
3885 Frequency : always with random mmap IO to multi-striped file
3887 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3888 Details : Only if the hole is at the end of OST object so that kms is too
3889 small. Fix is to increase kms accordingly in ll_nopage.
3892 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3894 Description: writes to a missing object would leak memory on the OST
3895 Details : If there is an inconsistency between the MDS and OST filesystems,
3896 such that the MDS references an object that doesn't exist, writes
3897 to that object will leak memory due to incorrect cleanup in the
3898 error handling path, eventually running out of memory on the OST.
3903 Description: Creating too long symlink causes lustre errors
3904 Details : Check symlink and name lengths before sending requests to MDS.
3907 Frequency : only if flock is enabled (not on by default)
3909 Description: posix locks not released on fd closure on 2.6.9+
3910 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3911 kernels, this caused such locks not to be released on fd close and
3912 then assertions on fs unmount about still used locks.
3915 Frequency : MDS failover only, very rarely
3917 Description: clients may get ASSERTION(granted_lock != NULL)
3918 Details : When request was taking a long time, and a client was resending
3919 a getattr by name lock request. The were multiple lock requests
3920 with the same client lock handle and
3921 mds_getattr_name->fixup_handle_for_resent_request found one of the
3922 lock handles but later failed with ASSERTION(granted_lock != NULL).
3927 Description: handle->h_buffer_credits > 0, assertion failure
3928 Details : h_buffer_credits is zero after truncate, causing assertion
3929 failure. This patch extends the transaction or creates a new
3933 Frequency : NFS re-export or patchless client
3934 Bugzilla : 11179, 10796
3935 Description: Crash on NFS re-export node (__d_move)
3936 Details : We do not want to hash the dentry if we don't have a lock.
3937 But if this dentry is later used in d_move, we'd hit uninitialised
3938 list head d_hash, so we just do this to init d_hash field but
3939 leave dentry unhashed.
3942 Frequency : NFS re-export or patchless client
3944 Description: NFS exports has problem with symbolic link
3945 Details : lustre client didn't properly install dentry when re-exported
3946 to NFS or running patchless client.
3949 Frequency : NFS re-export or patchless client
3951 Description: Various nfs/patchless fixes.
3952 Details : fixes reuse disconected alias for lookup process - this fixes
3953 warning "find_exported_dentry: npd != pd",
3954 fix permission error with open files at nfs.
3955 fix apply umask when do revalidate.
3958 Frequency : occasional
3960 Description: Crash on NFS re-export node
3961 Details : calling clear_page() on the wrong pointer triggered oops in
3962 generic_mapping_read().
3965 Frequency : rarely, using O_DIRECT IO
3967 Description: unaligned directio crashes client with LASSERT
3968 Details : check for unaligned buffers before trying any requests.
3971 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3973 Description: stale data returned from RAID cache
3974 Details : If only a small amount of IO is done to the RAID device before
3975 reading it again it is possible to get stale data from the RAID
3976 cache instead of reading it from disk.
3979 Frequency : always for sles10 kernel
3981 Description: sles10 support
3982 Details : ll_follow_link: compile fixes and using of nd_set_link
3983 under newer kernels.
3986 Frequency : depends on arch, kernel and compiler version, always on sles10
3989 Description: recursive or deep enough symlinks cause stack overflow
3990 Details : getting rid of large stack-allocated variable in
3994 Frequency : depends on hardware
3996 Description: lustre write performance loss in the SLES10 kernel
3997 Details : the performance loss is caused by using of write barriers in the
3998 ext3 code. The SLES10 kernel turns barrier support on by
3999 default. The fix is to undo that change for ldiskfs.
4001 ------------------------------------------------------------------------------
4003 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4005 * Support for kernels:
4006 2.6.9-42.0.3EL (RHEL 4)
4007 2.6.5-7.276 (SLES 9)
4008 2.4.21-47.0.1.EL (RHEL 3)
4009 2.6.12.6 vanilla (kernel.org)
4013 Frequency : quota enabled and large files being deleted
4015 Description: releasing more than 4GB of quota at once hangs OST
4016 Details : If a user deletes more than 4GB of files on a single OST it
4017 will cause the OST to spin in an infinite loop. Release
4018 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4023 Description: statfs data retrieved from /proc may be stale or zero
4024 Details : When reading per-device statfs data from /proc, in the
4025 {kbytes,files}_{total,free,avail} files, it may appear
4026 as zero or be out of date.
4029 Frequency : systems with MD RAID1 external journal devices
4031 Description: lconf's call to blkid is confused by RAID1 journal devices
4032 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4033 of returning all block devices that match the journal UUID.
4036 Frequency : always, for aggregate stripe size over 4GB
4038 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4039 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4040 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4043 Frequency : always if "lfs find" used on a local file/directory
4045 Description: "lfs find" segfaults if used on a local file/directory
4046 Details : The case where a directory component was not specified wasn't
4047 handled correctly. Handle this properly.
4050 Frequency : always on ppc64
4052 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4053 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4054 perform endianness conversion.
4057 Frequency : rarely (truncate to non-zero file size after write under load)
4058 Bugzilla : 10730, 10687
4059 Description: Files padded with zeros to next 4K multiple
4060 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4061 it is possible that files that are truncated to a non-zero size
4062 immediately after being written are filled with zero bytes beyond
4063 the truncated size. No file data is lost.
4065 Severity : enhancement
4067 Description: Allow recovery/failover for liblustre clients.
4068 Details : liblustre clients were unaware of failover configurations until
4071 Severity : enhancement
4073 Description: user file locks should fail when not mounting with flock option
4074 Details : Set up an error-returning stub in ll_file_operations.lock field
4075 to prevent incorrect behaviour when client is mounted without
4076 flock option. Also, set up properly f_op->flock field for
4080 Frequency : always on ia64
4082 Description: "lfs df" loops on printing out MDS statfs information
4083 Details : The obd_ioctl_data was not initialized and in some systems
4084 this caused a failure during the ioctl that did not return
4085 an error. Initialize the struct and return an error on failure.
4088 Frequency : SLES 9 only
4090 Description: Error of copying files with lustre special EAs as root
4091 Details : Client side always return success for setxattr call for lustre
4092 special xattr (currently only "trusted.lov").
4095 Frequency : rarely on clusters with both ia64+i386 clients
4097 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4098 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4099 clients are doing concurrent writes to the same file it is
4100 possible that the ia64 clients may LASSERT because the OST
4101 extent locks are not PAGE_SIZE aligned. Ensure that grown
4102 locks are always aligned on the request boundary.
4105 Frequency : specific use, occasional
4107 Description: Overwriting in use executable truncates on-disk binary image
4108 Details : If one node attempts to overwrite an executable in use by
4109 another node, we now correctly return ETXTBSY instead of
4110 truncating the file.
4112 Severity : enhancement
4114 Description: Async OSC create to avoid the blocking unnecessarily.
4115 Details : If an OST has no remaining object, system will block on the creating
4116 when need to create a new object on this OST. Now, ways use
4117 pre-created objects when available, instead of blocking on an
4118 empty osc while others are not empty. If we must block, we block
4119 for the shortest possible period of time.
4124 Description: chmod on Lustre root is propagated to other clients
4125 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4126 invalid by the follow_mount time.
4129 Frequency : liblustre clients only
4131 Description: Race in 'instant cancel' lock handling could lead to such locks
4132 never to be granted in case of SMP MDS
4133 Details : Do not destroy not yet granted but cbpending locks in
4137 Frequency : replay/resend of open
4139 Description: non null lock assetion failure in mds_intent_policy
4140 Details : Trying to replay/resend lockless open requests resulted in
4141 mds_open() returning 0 with no lock. Now it sets a flag if
4142 a lock is going to be returned.
4144 Severity : enhancement
4146 Description: Checksum enhancements
4147 Details : New checksum enhancements allow for resending RPCs that failed
4150 Severity : enhancement
4152 Description: Tunables on number of dirty pages in cacche
4153 Details : Allow to set limit on number of dirty pages cached.
4158 Description: client crash on unmount - lock still has references
4159 Details : In some error handling cases it was possible to leak a lock
4160 reference on a client while accessing a file. This was not
4161 harmful to the client during operation, but would cause the
4162 client to crash when the filesystem is unmounted.
4165 Frequency : specific case, rare
4167 Description: ETXTBSY on mds though file not in use by client
4168 Details : ETXTBSY is no longer incorrectly returned when attempting to
4169 chmod or chown a directory that the user previously tried to
4170 execute or a currently-executing binary.
4173 Frequency : extremely rare except on liblustre-based clients
4175 Description: Lustre space not freed when files are deleted
4176 Details : Clean up open-unlinked files after client eviction. Previously
4177 the unlink was skipped and the files remained as orphans.
4182 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4183 Details : In some cases it was possible to send a blocking callback to
4184 a client doing a glimpse, even though that client didn't get
4185 a lock granted. When the glimpse lock is cancelled on the OST
4186 the freed lock is left on the waiting list and corrupted the list.
4189 Frequency : all core dumps
4191 Description: Broke core dumps to lustre
4192 Details : Negative dentry may be unhashed if parent does not have UPDATE
4193 lock, but some callers, e.g. do_coredump, expect dentry to be
4194 hashed after successful create, hash it in ll_create_it.
4196 ------------------------------------------------------------------------------
4198 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4200 * Support for kernels:
4201 2.6.9-42.0.2.EL (RHEL 4)
4202 2.6.5-7.276 (SLES 9)
4203 2.4.21-40.EL (RHEL 3)
4204 2.6.12.6 vanilla (kernel.org)
4208 Frequency : always on RHEL 3
4210 Description: Number of open files grows over time
4211 Details : The number of open files grows over time, whether or not
4212 Lustre is started. This was due to a filp leak introduced
4213 by one of our kernel patches.
4215 ------------------------------------------------------------------------------
4217 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4219 * Support for kernels:
4220 2.6.9-42.EL (RHEL 4)
4221 2.6.5-7.276 (SLES 9)
4222 2.4.21-40.EL (RHEL 3)
4223 2.6.12.6 vanilla (kernel.org)
4228 Bugzilla : 5719, 9635, 9792, 9684
4229 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4230 Details : If a server is under heavy load and cannot reply to new
4231 connection requests before the client resends the (re)connect,
4232 the connection handling code can behave badly if two service
4233 threads are concurrently handing separate (re)connections from
4234 the same client. Add better locking to the connection handling
4235 code, and ensure that only a single connection will be processed
4236 for a given client UUID, even if the lock is dropped.
4238 Severity : enhancement
4240 Description: add TCP zero-copy support to kernel
4241 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4242 sends if the hardware supports scatter-gather and checksumming.
4243 This allows socklnd to do client-write and server-read more
4244 efficiently and reduce CPU utilization from skbuf copying.
4247 Frequency : only if NFS exporting from client
4249 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4250 Details : If setting attributes on a file created under NFS that had
4251 never been opened it would be possible to oops the client
4252 if the file had no objects.
4255 Frequency : always for liblustre
4257 Description: liblustre client does MDS+OSTs setattr RPC for each write
4258 Details : When doing a write from a liblustre client, the client
4259 incorrectly issued an RPC to the MDS and each OST the file was
4260 striped over in order to update the timestamps. When writing
4261 with small chunks and many clients this could overwhelm the MDS
4262 with RPCs. In all cases it would slow down the write because
4263 these RPCs are unnecessary.
4265 Severity : enhancement
4267 Description: allow number of MDS service threads to be changed at module load
4268 Details : It is now possible to change the number of MDS service threads
4269 running. Adding "options mds mds_num_threads={N}" to the MDS's
4270 /etc/modprobe.conf will set the number of threads for the next
4271 time Lustre is restarted (assuming the "mds" module is also
4272 reloaded at that time). The default number of threads will
4273 stay the same, 32 for most systems.
4278 Description: OST crash if filesystem is unformatted or corrupt
4279 Details : If an OST is started on a device that has never been formatted
4280 or if the filesystem is corrupt and cannot even mount then the
4281 error handling cleanup routines would dereference a NULL pointer.
4286 Description: NULL pointer deref in llap_from_page.
4287 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4288 incorrect) mapping if the page was truncated/reclaimed while it was
4289 searched for. Check for this condition and skip such pages when
4290 doing readahead. Introduce extra check to llap_from_page() to
4291 verify page->mapping->host is non-NULL (so page is not anonymous).
4294 Frequency : Sometimes when using sys_sendfile
4296 Description: "page not covered by a lock" warnings from ll_readpage
4297 Details : sendfile called ll_readpage without right page locks present.
4298 Now we introduced ll_file_sendfile that does necessary locking
4299 around call to generic_file_sendfile() much like we do in
4303 Frequency : with certain MDS communication failures at client mount time
4305 Description: NULL pointer deref after failed client mount
4306 Details : a client connection request may delayed by the network layer
4307 and not be sent until after the PTLRPC layer has timed out the
4308 request. If the client fails the mount immediately it will try
4309 to clean up before the network times out the request. Add a
4310 reference from the request import to the obd device and delay
4311 the cleanup until the network drops the request.
4314 Frequency : occasionally during client (re)connect
4316 Description: assertion failure during client (re)connect
4317 Details : processing a client connection request may be delayed by the
4318 client or server longer than the client connect timeout. This
4319 causes the client to resend the connection request. If the
4320 original connection request is replied in this interval, the
4321 client may trip an assertion failure in ptlrpc_connect_interpret()
4322 which thought it would be the only running connect process.
4325 Frequency : only with obd_echo servers and clients that are rebooted
4327 Description: kernel BUG accessing uninitialized data structure
4328 Details : When running an obd_echo server it did not start the ping_evictor
4329 thread, and when a client was evicted an uninitialized data
4330 structure was accessed. Start the ping_evictor in the RPC
4331 service startup instead of the OBD startup.
4333 Severity : enhancement
4334 Bugzilla : 10193 (patchless)
4335 Description: Remove dependency on various unexported kernel interfaces.
4336 Details : No longer need reparent_to_init, exit_mm, exit_files,
4337 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4340 Frequency : rare (only users of deprecated and unsupported LDAP config)
4342 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4343 Details : LDAP apparently contains 'lustreName' attributes instead of
4344 'name'. A simple remapping of the name is sufficient.
4347 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4349 Description: waiting_locks_callback trips kernel BUG if client is evicted
4350 Details : Running with the dump_on_timeout debug flag turned on makes
4351 it possible that the waiting_locks_callback() can try to dump
4352 the Lustre kernel debug logs from an interrupt handler. Defer
4353 this log dumping to the expired_lock_main() thread.
4355 Severity : enhancement
4357 Description: Support NFS exporting on 2.6 kernels.
4358 Details : Implement non-rawops metadata methods for NFS server to use without
4359 changing NFS server code.
4362 Frequency : very rare (synthetic metadata workload only)
4364 Description: two racing renames might cause an MDS thread to deadlock
4365 Details : Running the "racer" program may cause one MDS thread to rename
4366 a file from being the source of a rename to being the target of
4367 a rename at exactly the same time that another thread is doing
4368 so, and the second thread has already enqueued these locks after
4369 doing a lookup of the target and is trying to relock them in
4370 order. Ensure that we don't try to re-lock the same resource.
4373 Frequency : only very large systems with liblustre clients
4375 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4376 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4377 all OSTs in parallel. This allows the eviction of stale liblustre
4378 clients to proceed much faster than if they were done in series,
4379 and also offers similar improvements for other set_info RPCs.
4384 Description: excessive CPU usage during initial read phase on client
4385 Details : During the initial read phase on a client, it would agressively
4386 retry readahead on the file, consuming too much CPU and impacting
4387 performance (since 1.4.5.8). Improve the readahead algorithm
4388 to avoid this, and also improve some other common cases (read
4389 of small files in particular, where "small" is files smaller than
4390 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4395 Description: MDS crash when receiving packet with unknown intent.
4396 Details : Do not LBUG in unknown intent case, just return -EFAULT
4398 Severity : enhancement
4399 Bugzilla : 9293, 9385
4400 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4401 Details : Do not serialize getattr (non-intent version) and statfs.
4404 Frequency : occasional, when OST network is overloaded/intermittent
4406 Description: client evicted by OST after bulk IO timeout
4407 Details : If a client sends a bulk IO request (read or write) the OST
4408 may evict the client if it is unresposive to its data GET/PUT
4409 request. This is incorrect if the network is overloaded (takes
4410 too long to transfer the RPC data) or dropped the OST GET/PUT
4411 request. There is no need to evict the client at all, since
4412 the pinger and/or lock callbacks will handle this, and the
4413 client can restart the bulk request.
4416 Frequency : Always when mmapping file with no objects
4418 Description: client crashes when mmapping file with no objects
4419 Details : Check that we actually have objects in a file before doing any
4420 operations on objects in ll_vm_open, ll_vm_close and
4426 Description: Request leak when working with deleted CWD
4427 Details : Introduce advanced request refcount tracking for requests
4428 referenced from lustre intent.
4430 Severity : Enhancement
4432 Description: Cache open file handles on client.
4433 Details : MDS now will return special lock along with openhandle, if
4434 requested and client is allowed to hold openhandle, even if unused,
4435 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4436 closing files for every read/write openration.
4438 Severity : Enhancement
4440 Description: Cache open negative dentries on client when possible.
4441 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4442 negative dentries on lock revocation.
4447 Description: Remounting a client read-only wasn't possible with a zconf mount
4448 Details : It wasn't possible to remount a client read-only with llmount.
4450 Severity : enhancement
4451 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4452 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4453 MPICH is included to add Lustre-specific ADIO interfaces.
4454 This is based closely on the UFS ADIO layer and only differs
4455 in file creation, in order to allow the OST striping to be set.
4456 This is user-contributed code and not supported by CFS.
4461 Description: extended inode attributes (immutable, append-only) work improperly
4462 when 2.4 and 2.6 kernels are used on client/server or vice versa
4463 Details : Introduce kernel-independent values for these flags.
4465 Severity : enhancement
4468 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4469 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4470 granularity. Now, Fractional MB granularity is be supported,
4471 this is very useful for low memory system.
4473 Severity : enhancement
4475 Description: Getattr by fid
4476 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4477 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4478 have locks and avoids getattr rpc requests for it on every stat.
4481 Frequency : Always, for filesystems larger than 2TB
4483 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4484 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4485 filesystems larger than 2TB to be mounted with mballoc enabled.
4488 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4490 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4491 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4492 filesystems larger than 2TB is mounted then the kernel will
4493 silently corrupt the start of the filesystem. CONFIG_LBD is
4494 enabled for all CFS-supported kernels, but the possibility of
4495 this happening with a modified kernel config exists.
4497 Severity : enhancement
4499 Description: add client O_DIRECT support for 2.6 kernels
4500 Details : It is now possible to do O_DIRECT reads and writes to files
4501 in the Lustre client mountpoint on 2.6 kernel clients.
4503 Severity : enhancement
4505 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4506 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4507 parallel, not waiting for response from every OST before sending
4508 a rpc to the next OST.
4513 Description: setattr vs write race when updating file timestamps
4514 Details : Client processes that update a file timestamp into the past
4515 right after writing to the file (e.g. tar) it is possible that
4516 the updated file modification time can be reset to the current
4517 time due to a race between processing the setattr and write RPC.
4519 Severity : enhancement
4521 Description: Bring 'lfs find' closer in line with regular Linux find.
4522 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4523 -print0 options and obtains all the needed info through the lustre
4526 Severity : enhancement
4528 Description: support up to 1024 configured devices on one node
4529 Details : change obd_dev array from statically allocated to dynamically
4530 allocated structs as they are first used to reduce memory usage
4535 Description: Flush dirty partially truncated pages during truncate
4536 Details : Immediatelly flush partially truncated pages in filter_setattr,
4537 this way we completely avoid having any pages in page cache on OST
4538 and can retire ugly workarounds during writes to flush such pages.
4543 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4544 Details : protect inode from truncation within vfs_unlink() context
4545 just take a reference before calling vfs_unlink() and release it
4546 when parent's i_sem is free.
4551 Description: last_id value checked outside lock on OST caused LASSERT failure
4552 Details : If there were multiple MDS->OST object precreate requests in
4553 flight, it was possible that the OST's last object id was checked
4554 outside a lock and incorrectly tripped an assertion. Move checks
4555 inside locks, and discard old precreate requests.
4558 Frequency : always, if extents are used on OSTs
4560 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4561 Details : OSTs using the extents format would not zero the high 16 bits of
4562 the index physical block number. This is not a problem for any
4563 OST filesystems smaller than 16TB, and no kernels support ext3
4564 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4565 new/modified files) and can be fixed for existing filesystems
4566 with e2fsprogs-1.39-cfs1.
4571 Description: import connection selection may be incorrect if timer wraps
4572 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4573 import connections to be ignored if the 32-bit jiffies counter
4574 wraps. Use a 64-bit jiffies counter.
4577 Frequency : very large clusters immediately after boot
4579 Description: LNET request buffers exhausted under heavy short-term load
4580 Details : If a large number of client requests are generated on a service
4581 that has previously never seen so many requests it is possible
4582 that the request buffer growth cannot keep up with the spike in
4583 demand. Instead of dropping incoming requests, they are held in
4584 the LND until the RPC service can accept more requests.
4587 Frequency : Sometimes during replay
4589 Description: Assertion failure in ll_local_open after replay.
4590 Details : If replay happened on an open request reply before we were able
4591 to set replay handler, reply will become not swabbed tripping the
4592 assertion in ll_local_open. Now we set the handler right after
4593 recognising of open request
4596 Frequency : very rare
4598 Description: kernel reports "badness in vsnprintf"
4599 Details : Reading from the "recovery_status" /proc file in small chunks
4600 may cause a negative length in lprocfs_obd_rd_recovery_status()
4601 call to vsnprintf() (which is otherwise harmless). Exit early
4602 if there is no more space in the output buffer.
4604 Severity : enhancement
4606 Description: clear OBD RPC statistics by writing to them
4607 Details : It is now possible to clear the OBD RPC statistics by writing
4608 to the "stats" file.
4613 Description: Client mtime is not the same on different clients after utimes
4614 Details : In some cases, the client was using the utimes() syscall on
4615 a file cached on another node. The clients now validate the
4616 ctime from the MDS + OSTs to determine which one is right.
4621 Description: Inability to activate failout mode
4622 Details : lconf script incorrectly assumed that in pythong string's numeric
4623 value is used in comparisons.
4626 Frequency : always with multiple stripes per file
4628 Description: Inefficient object allocation for mutli-stripe files
4629 Details : When selecting which OSTs to stripe files over, for files with
4630 a stripe count that divides evenly into the number of OSTs,
4631 the MDS is always picking the same starting OST for each file.
4632 Return the OST selection heuristic to the original design.
4637 Description: mount failures may take full timeout to return an error
4638 Details : Under some heavy load conditions it is possible that a
4639 failed mount can wait for the full obd_timeout interval,
4640 possibly several minutes, before reporting an error.
4641 Instead return an error as soon as the status is known.
4643 Frequency : quota enabled and large files being deleted
4645 Description: releasing more than 4GB of quota at once hangs OST
4646 Details : If a user deletes more than 4GB of files on a single OST it
4647 will cause the OST to spin in an infinite loop. Release
4648 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4653 Description: statfs data retrieved from /proc may be stale or zero
4654 Details : When reading per-device statfs data from /proc, in the
4655 {kbytes,files}_{total,free,avail} files, it may appear
4656 as zero or be out of date.
4659 Frequency : systems with MD RAID1 external journal devices
4661 Description: lconf's call to blkid is confused by RAID1 journal devices
4662 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4663 of returning all block devices that match the journal UUID.
4666 Frequency : always, for aggregate stripe size over 4GB
4668 Description: assertion fails when trying to use 4GB stripe size
4669 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4670 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4673 Frequency : always on ppc64
4675 Description: the first write on an ext3 filesystem with mballoc got stuck
4676 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4677 perform endianness conversion.
4679 ------------------------------------------------------------------------------
4681 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4683 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4684 INTEROPERATE with older versions automatically. Please read the
4685 user documentation before upgrading any part of a live system.
4686 * WARNING: Lustre networking configuration changes are required with
4687 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4690 * Support for kernels:
4691 2.6.9-22.0.2.EL (RHEL 4)
4692 2.6.5-7.244 (SLES 9)
4693 2.6.12.6 vanilla (kernel.org)
4696 Severity : enhancement
4697 Bugzilla : 7981/8208
4698 Description: Introduced Lustre Networking (LNET)
4699 Details : LNET is new networking infrastructure for Lustre, it includes
4700 a reorganized network configuration mode (see the user
4701 documentation for full details) as well as support for routing
4702 between different network fabrics. Lustre Networking Devices
4703 (LNDs) for the supported network fabrics have also been
4704 created for this new infrastructure.
4706 Severity : enhancement
4707 Description: Introduced Access control lists
4708 Details : clients can set ACLs on files and directories in order to have
4709 more fine-grained permissions than the standard Unix UGO+RWX.
4710 The MDS must be started with the "-o acl" mount option.
4712 Severity : enhancement
4713 Description: Introduced filesystem quotas
4714 Details : Administrators may now establish per-user quotas on the
4717 Severity : enhancement
4719 Description: Configuration change for the XT3
4720 The PTLLND is now used to run Lustre over Portals on the XT3
4721 The configure option(s) --with-cray-portals are no longer used.
4722 Rather --with-portals=<path-to-portals-includes> is used to
4723 enable building on the XT3. In addition to enable XT3 specific
4724 features the option --enable-cray-xt3 must be used.
4729 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4730 Details : A race between allocating a new llcd and re-getting the llcd_lock
4731 allowed another thread to grab newly-allocated llcd.
4733 Severity : enhancement
4735 Description: 2.6 OST async journal commit and locking fix to improve performance
4736 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4737 2.6 kernels are now async as they already were in 2.4 kernels so
4738 that they can commit concurrently with the network bulk transfer.
4739 For block-allocated files the filter allocation semaphore is held
4740 to avoid filesystem fragmentation during allocation. BKL lock
4741 removed for 2.6 xattr operations where it is no longer needed.
4746 Description: lconf incorrectly determined whether two IP networks could talk
4747 Details : In some more complicated routing and multiple-network
4748 configurations, lconf will avoid trying to make a network
4749 connection to a disjoint part of the IP space. It was doing the
4750 math incorrectly for one set of cases.
4755 Description: Fix for potential infinite loop processing records in an llog.
4756 Details : If an llog record is corrupted/zeroed, it is possible to loop
4757 forever in llog_process(). Validate the llog record length
4758 and skip the remainder of the block on error.
4761 Frequency : occasional (liblustre only)
4763 Description: liblustre could not open files whose last component is a symlink
4764 Details : sysio_path_walk() would incorrectly pass the open intent to
4765 intermediate path components.
4768 Frequency : rare (liblustre only with non-standard tuning)
4769 Bugzilla : 7201 (7350)
4770 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4771 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4772 after it has been released, only do this for referenced locks
4774 Severity : enhancement
4776 Description: specifying an (invalid) directory default stripe_size of -1
4777 would reset the directory default striping
4778 Details : stripe_size -1 was used internally to signal directory stripe
4779 removal, now use "all default" to signal dir stripe removal
4780 as a directory striping of "all default" is not useful
4783 Frequency : common for large clusters running liblustre clients
4785 Description: doing an ls when liblustre clients are running is slow
4786 Details : sending a glimpse AST to a liblustre client waits for every AST
4787 to time out, as liblustre clients will not respond. Since they
4788 cannot cache data we refresh the OST lock LVB from disk instead.
4790 Severity : enhancement
4792 Description: doing an ls at the same time as file IO can be slow
4793 Details : enqueue and other "small" requests can be blocked behind many
4794 large IO requests. Create a new OST IO portal for non-IO
4795 requests so they can be processed faster.
4798 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4800 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4801 created file with mode 000
4802 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4803 allow an ftruncate() as a non-root user to a file with mode 000.
4804 Lustre now respects this flag to disable mode checks when
4805 truncating a file owned by the user
4808 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4811 Description: Revoking locks from clients that went dead or catatonic might take
4813 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4814 cancellation of such locks instant on servers without waiting for
4815 any reply from clients. Clients drops these locks when cancel
4816 notification from server is received without replying.
4819 Frequency : liblustre-only, when liblustre client dies or becomes busy
4821 Description: Doing ls on Linux clients can take a long time with active
4823 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4824 granting such locks to it in the first place if possible. Locks
4825 are taken by proxy on the OST during the read or write and
4826 dropped immediately afterward. Add connect flags handling, do
4827 not grant locks to liblustre clients for glimpse ASTs.
4829 Severity : enhancement
4831 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4832 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4833 sequential reads and behaves badly with applications doing random
4834 reads. Improve it by reading ahead at least read region, and
4835 avoiding excessive large RPC for small reads.
4837 Severity : enhancement
4839 Description: Creating more than 1000 files for a single job may cause a load
4840 imbalance on the OSTs if there are also a large number of OSTs.
4841 Details : qos_prep_create() uses an OST index reseed value that is an
4842 even multiple of the number of available OSTs so that if the
4843 reseed happens in the middle of the object allocation it will
4844 still utilize the OSTs as uniformly as possible.
4849 Description: OST or MDS may oops in ping_evictor_main()
4850 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4851 but doesn't restart at beginning of obd_exports_timed list
4854 Severity : enhancement
4856 Description: improve by-nid export eviction on the MDS and OST
4857 Details : allow multiple exports with the same NID to be evicted at one
4858 time without re-searching the exports list.
4861 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4863 Description: MDS may oops in groups_free()
4864 Details : in rare race conditions a newly allocated group_info struct is
4865 freed again, and this can be NULL. The 2.4 compatibility code
4866 for groups_free() checked for a NULL pointer, but 2.6 did not.
4869 Frequency : common for liblustre clients doing little filesystem IO
4870 Bugzilla : 9352, 7313
4871 Description: server may evict liblustre clients accessing contended locks
4872 Details : if a client is granted a lock or receives a completion AST
4873 with a blocking AST already set it would not reply to the AST
4874 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4877 Frequency : lfs setstripe, only systems with more than 160 OSTs
4879 Description: unable to set striping with a starting offset beyond OST 160
4880 Details : llapi_create_file() incorrectly limited the starting stripe
4881 index to the maximum single-file stripe count.
4884 Frequency : LDAP users only
4886 Description: lconf did not handle in-kernel recovery with LDAP properly
4887 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4889 Severity : enhancement
4891 Description: bind OST threads to NUMA nodes to improve performance
4892 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4893 node and do their allocations there to localize memory access
4895 Severity : enhancement
4897 Description: llmount can determine client NID directly from Myrinet (GM)
4898 Details : the client NID code from gmnalnid was moved directly into
4899 llmount, removing the need to use this or specifying the
4900 client NID explicitly when mounting GM clients with zeroconf
4903 Frequency : if client is started with down MDS
4905 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4906 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4907 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4908 ptlrpc_import_delayed_req() because we didn't check if the
4909 request was interrupted, and we also didn't break out of the
4910 event loop if there was no timeout
4915 Description: data loss during non-page-aligned writes to a single file from
4916 both multiple nodes and multiple threads on one node at same time
4917 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4918 inconsistency led to false short-reads, that were cached and later
4919 used by ->prepare_write() to fill in partially written page,
4920 leading to data loss.
4923 Frequency : always, if lconf --abort_recovery used
4925 Description: lconf --abort_recovery fails with 'Operation not supported'
4926 Details : lconf was attempting to abort recovery on the MDT device and not
4929 Severity : enhancement
4931 Description: remove cleanup logs
4932 Details : replace lconf-generated cleanup logs with lustre internal
4933 cleanup routines. Eliminates the need for client-cleanup and
4936 Severity : enhancement
4938 Description: add support for EAs (user and system) on lustre filesystems
4939 Details : it is now possible to store extended attributes in the Lustre
4940 client filesystem, and with the user_xattr mount option it
4941 is possible to allow users to store EAs on their files also
4943 Severity : enhancement
4945 Description: Add possibility (config option) to show minimal available OST free
4947 Details : When compiled with --enable-mindf configure option, statfs(2)
4948 (and so, df) will return least minimal free space available from
4949 all OSTs as amount of free space on FS, instead of summary of
4950 free spaces of all OSTs.
4952 Severity : enhancement
4954 Description: do not expand extent locks acquired on OST-side
4955 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4956 by server: they are not cached anyway.
4959 Frequency : when mmap is used/binaries executed from Lustre
4961 Description: Unmmap pages before throwing them away from read cache.
4962 Details : llap_shrink cache now attempts to unmap pages before discarding
4963 them (if unmapping failed - do not discard). SLES9 kernel has
4964 extra checks that trigger if this unmapping is not done first.
4969 Description: lconf didn't resolve symlinks before checking to see whether a
4970 given mountpoint was already in use
4973 Frequency : when migrating failover services
4974 Bugzilla : 6395, 9514
4975 Description: When migrating a subset of services from a node (e.g. failback
4976 from a failover service node) the remaining services would
4977 time out and evict clients.
4978 Details : lconf --force (implied by --failover) sets the global obd_timeout
4979 to 5 seconds in order to quickly disconnect, but this caused
4980 other RPCs to time out too quickly. Do not change the global
4981 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4983 Severity : enhancement
4984 Frequency : if MDS is started with down OST
4985 Bugzilla : 9439,5706
4986 Description: Allow startup/shutdown of an MDS without depending on the
4987 availability of the OSTs.
4988 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4989 Add appropriate locking and lov-osc refcounts for safe
4990 cleaning. Add osc abort_inflight calls in case the
4991 synchronize never started.
4994 Frequency : occasional (Cray XT3 only)
4996 Description: root not authorized to access files in CRAY_PORTALS environment
4997 Details : The client process capabilities were not honoured on the MDS in
4998 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4999 been accessed by an authorized user then root was able to access
5000 the file on the local client also. The root user capabilities
5001 are now allowed on the MDS, as this environment has secure UID.
5004 Frequency : occasional
5006 Description: ldiskfs "too long searching" message happens too often
5007 Details : A debugging message (otherwise harmless) prints too often on
5008 the OST console. This has been reduced to only happen when
5009 there are fragmentation problems on the filesystem.
5014 Description: Division by zero in statfs when all OSCs are inactive
5015 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5016 lov_statfs divides by value returned by lov_get_stripecnt().
5020 Bugzilla : 9489, 3273
5021 Description: First write from each client to each OST was only 4kB in size,
5022 to initialize client writeback cache, which caused sub-optimal
5023 RPCs and poor layout on disk for the first written file.
5024 Details : Clients now request an initial cache grant at (re)connect time
5025 and so that they can start streaming writes to the cache right
5026 away and always do full-sized RPCs if there is enough data.
5027 If the OST is rebooted the client also re-establishes its grant
5028 so that client cached writes will be honoured under the grant.
5033 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5034 Details : Now I/O RPCs go to different portal number and (presumably) fast
5035 lock requests (and glimses) and other RPCs get their own service
5036 threads pool that should be able to service those RPCs
5039 Severity : enhancement
5041 Description: Ability to exchange lustre version between client and servers and
5042 issue warnings at client side if client is too old. Also for
5043 liblustre clients there is ability to refuse connection of too old
5045 Details : New 'version' field is added to connect data structure that is
5046 filled with version info. That info is later checked by server and
5050 Frequency : rare, liblustre only.
5051 Bugzilla : 9296, 9581
5052 Description: Two simultaneous writes from liblustre at offset within same page
5053 might proceed at the same time overwriting eachother with stale
5055 Details : I/O lock withing llu_file_prwv was released too early, before data
5056 actually was hitting the wire. Extended lock-holding time until
5057 server acknowledges receiving data.
5060 Frequency : extremely rare. Never observed in practice.
5062 Description: avoid generating lustre_handle cookie of 0.
5063 Details : class_handle_hash() generates handle cookies by incrementing
5064 global counter, and can hit 0 occasionaly (this is unlikely, but
5065 not impossible, because initial value of cookie counter is
5066 selected randonly). Value of 0 is used as a sentinel meaning
5067 "unassigned handle" --- avoid it. Also coalesce two critical
5068 sections in this function into one.
5070 Severity : enhancement
5072 Description: allow liblustre clients to delegate truncate locking to OST
5073 Details : To avoid overhead of locking, liblustre client instructs OST to
5074 take extent lock in ost_punch() on client's behalf. New connection
5075 flag is added to handle backward compatibility.
5077 Severity : enhancement
5078 Bugzilla : 4928, 7341, 9758
5079 Description: allow number of OST service threads to be specified
5080 Details : a module parameter allows the number of OST service threads
5081 to be specified via "options ost ost_num_threads={N}" in the
5082 OSS's /etc/modules.conf or /etc/modprobe.conf.
5086 Bugzilla : 6146, 9635, 9895
5087 Description: servers crash with bad pointer in target_handle_connect()
5088 Details : In rare cases when a client is reconnecting it was possible that
5089 the connection request was the last reference for that export.
5090 We would temporarily drop the export reference and get a new
5091 one, but this may have been the last reference and the export
5092 was just destroyed. Get new reference before dropping old one.
5094 Severity : enhancement
5095 Frequency : if client is started with failover MDS
5097 Description: Allow multiple MDS hostnames in the mount command
5098 Details : Try to read the configuration from all specified MDS
5099 hostnames during a client mount in case the "primary"
5102 Severity : enhancement
5104 Description: Stop sending data to evicted clients as soon as possible.
5105 Details : Check if the client we are about to send or are sending data to
5106 was evicted already. (Check is done every second of waiting,
5107 for which l_wait_event interface was extended to allow checking
5108 of exit condition at specified intervals).
5111 Frequency : rare, normally only when NFS exporting is done from client
5113 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5115 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5118 Frequency : always, if srand() is called before liblustre initialization
5120 Description: Liblustre uses system PRNG disturbing its usage by user application
5121 Details : Introduce internal to lustre fast and high-quality PRNG for
5122 lustre usage and make liblustre and some other places in generic
5123 lustre code to use it.
5125 Severity : enhancement
5126 Bugzilla : 9477, 9557, 9870
5127 Description: Verify that the MDS configuration logs are updated when xml is
5128 Details : Check if the .xml configuration logs are newer than the config
5129 logs stored on the MDS and report an error if this is the case.
5130 Request --write-conf, or allow starting with --old_conf.
5132 Severity : enhancement
5134 Description: Handle symlinks in the path when checking if Lustre is mounted.
5135 Details : Resolve intermediate symlinks when checking if a client has
5136 mounted a filesystem to avoid duplicate client mounts.
5141 Description: lconf can hit an error exception but still return success.
5142 Details : The lconf command catches the Command error exception at the top
5143 level script context and will exit with the associated exit
5144 status, but doesn't ensure that this exit status is non-zero.
5149 Description: failure of ptlrpc thread startup can cause oops
5150 Details : Starting a ptlrpc service thread can fail if there are a large
5151 number of threads or the server memory is very fragmented.
5152 Handle this without oopsing.
5155 Frequency : always, only if liblustre and non-default acceptor port was used
5157 Description: liblustre cannot connect to servers with non-default acceptor port
5158 Details : tcpnal_set_default_params() was not called and was therefore
5159 ignoring the environment varaible TCPNAL_PORT, as well as other
5160 TCPNAL_ environment variables
5165 Description: two objects could be created on the same OST for a single file
5166 Details : If an OST is down, in some cases it was possible to create two
5167 objects on a single OST for a single file. No problems other
5168 than potential performance impact and spurious error messages.
5172 Bugzilla : 5681, 9562
5173 Description: Client may oops in ll_unhash_aliases
5174 Details : Client dcache may become inconsistent in race condition.
5175 In some cases "getcwd" can fail if the current directory is
5181 Description: Inode refcounting problems in NFS export code
5182 Details : link_raw functions used to call d_instantiate without obtaining
5183 extra inode reference first.
5187 Bugzilla : 9942, 9903
5188 Description: Referencing freed requests leading to crash, memleaks with NFS.
5189 Details : We used to require that call to ll_revalidate_it was always
5190 followed by ll_lookup_it. Also with revalidate_special() it is
5191 possible to call ll_revalidate_it() twice for the same dentry
5192 even if first occurence returned success. This fix changes semantic
5193 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5194 reference on a request referred from the intent.
5195 ll_intent_release() then releases such a request.
5198 Frequency : rare, normally benchmark loads only
5200 Description: unlinked inodes were kept in memory on the client
5201 Details : If a client is repeatedly creating and unlinking files it
5202 can accumulate a lot of stale inodes in the inode slab cache.
5203 If there is no other client load running this can cause the
5204 client node to run out of memory. Instead flush old inodes
5205 from client cache that have the same inode number as a new inode.
5208 Frequency : SLES9 2.6.5 kernel and long filenames only
5209 Bugzilla : 9969, 10379
5210 Description: utime reports stale NFS file handle
5211 Details : SLES9 uses out-of-dentry names in some cases, which confused
5212 the lustre dentry revalidation. Change it to always use the
5216 Frequency : rare, unless heavy write-truncate concurrency is continuous
5217 Bugzilla : 4180, 6984, 7171, 9963, 9331
5218 Description: OST becomes very slow and/or deadlocked during object unlink
5219 Details : filter_destroy() was holding onto the parent directory lock
5220 while truncating+unlinking objects. For very large objects this
5221 may block other threads for a long time and slow overall OST
5222 responsiveness. It may also be possible to get a lock ordering
5223 deadlock in this case, or run out of journal credits because of
5224 the combined truncate+unlink. Solution is to do object truncate
5225 first in one transaction without parent lock, and then do the
5226 final unlink in a new transaction with the parent lock. This
5227 reduces the lock hold time dramatically.
5230 Frequency : rare, 2.4 kernels only
5232 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5233 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5234 before unmounting their filesystem in order to avoid deadlock.
5235 The kernel_locked() function in 2.4 kernels only checks whether
5236 the kernel lock is held, not whether it is this process that is
5237 holding it as 2.6 kernels do.
5242 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5243 Details : The client ptlrpc code may be trying to reconnect to a down
5244 server before a previous connection attempt has timed out.
5245 Increase the reconnect interval to be longer than the connection
5246 timeout interval to avoid sending duplicate connections to
5250 Frequency : echo_client brw_test command
5252 Description: fix echo_client to work with OST preallocated code
5253 Details : OST preallocation code (5137) didn't take echo_client IO path
5254 into account: echo_client calls filter methods outside of any
5255 OST thread and, hence, there is no per-thread preallocated
5256 pages and buffers to use. Solution: hijack pga pages for IO. As
5257 a byproduct, this avoids unnecessary data copying.
5261 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5262 Description: Client can oops in mdc_commit_close() after open replay
5263 Details : It was possible for the MDS to return an open request with no
5264 transaction number in mds_finish_transno() if the client was
5265 evicted, but without actually returning an error. Clients
5266 would later try to replay that open and may trip an assertion
5267 Simplify the client close codepath, and always return an error
5268 from the MDS in case the open is not successful.
5271 Frequency : rare, 2.6 OSTs only
5273 Description: OST may deadlock under high load on fragmented files
5274 Details : If there was a heavy load and highly-fragmented OST filesystems
5275 it was possible to have all the OST threads deadlock waiting on
5276 allocation of biovecs, because the biovecs were not released
5277 until the entire RPC IO was completed. Instead, release biovecs
5278 as soon as they are complete to ensure forward IO progress.
5280 Severity : enhancement
5282 Description: Support for specifying external journal device at mount
5283 Details : If an OST or MDS device is formatted with an external journal
5284 device, this device major/minor is stored in the ext3 superblock
5285 and may not be valid for failover. Allow detecting and
5286 specifying the external journal at mount time.
5291 Description: Mounting an MDS with pending unlinked files may cause oops
5292 Details : target_finish_recovery() calls mds_postrecov() which returned
5293 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5294 considers this an error and immediately begins cleaning up the
5295 lov, just after starting the mds_lov process
5297 Severity : enhancement
5299 Description: Implement 'lfs df' to report actual free space on per-OST basis
5300 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5301 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5302 usage of objects; '-h' to report in human readable format.
5304 ------------------------------------------------------------------------------
5306 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5313 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5314 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5315 at startup during orphan unlinking. Instead, skip buddy update
5316 before setup, as it will be regenerated after recovery is complete.
5321 Description: If an OST is inactive, its locks might reference stale inodes.
5322 Details : lov_change_cbdata() must iterate over all namespaces, even if
5323 they are inactive to clear inode references from the lock.
5325 Severity : enhancement
5326 Frequency : occasional, if non-standard max_dirty_mb used
5328 Description: Client will block write RPCs if not enough grant
5329 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5330 then the client will block writes while waiting for another RPC
5331 to complete instead of consuming its dirty limit. With change
5332 we get improved performance when max_dirty_mb is small.
5334 Severity : enhancement
5335 Bugzilla : 3389, 6253
5336 Description: Add support for supplementary groups on the MDS.
5337 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5338 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5339 lookups for user supplementary groups into a cache.
5343 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5344 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5345 for liblustre O_EXCL is enforced by MDS.
5349 Description: readdir never returns NULL in liblustre.
5350 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5351 dentry in struct dirent.
5355 Frequency : liblustre only, depends on application IO pattern
5356 Description: liblustre clients evicted if not contacting servers
5357 Details : Don't put liblustre clients into the ping_evictor list, so
5358 they will not be evicted by the pinger ever.
5360 Severity : enhancement
5362 Description: Add ability to evict clients by NID from MDS.
5363 Details : By echoing "nid:$NID" string into
5364 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5365 $NID would be instantly evicted from this MDS and from all active
5366 OSTs connected to it.
5370 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5371 Details : lookup_it_finish() used to query file size from OSTs that was not
5376 Description: service threads change working directory to that of init
5377 Details : Starting lustre service threads may pin the working directory
5378 of the parent thread, making that filesystem busy. Threads
5379 now change to the working directory of init to avoid this.
5383 Frequency : during shutdown only
5384 Description: shutdown with a failed MDS or OST can cause unmount to hang
5385 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5390 Frequency : 2.6 only
5391 Description: chmod/chown may include an extra supplementary group
5392 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5393 supplementary group and if none is specified this is used.
5396 Bugzilla : 5479 (6816)
5398 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5399 Details : If lookup is in progress on a file that is unlinked we might try
5400 to revalidate the inode and fail in revalidate after lookup is
5401 complete and ll_file_open() enqueues the open again but
5402 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5405 Frequency : always, if lconf --abort_recovery used
5407 Description: lconf --abort_recovery fails with 'Operation not supported'
5408 Details : lconf was attempting to abort recovery on the MDT device and not
5411 ------------------------------------------------------------------------------
5413 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5418 Frequency : rare (only unsupported configurations with a node running as an
5420 Bugzilla : 6514, 5137
5421 Description: Mounting a Lustre file system on a node running as an OST could
5423 Details : OSTs now preallocates memory needed to write out data at
5424 startup, instead of when needed, to avoid having to
5425 allocate memory in possibly low memory situations.
5426 Specifically, if the file system is mounted on on OST,
5427 memory pressure could force it to try to write out data,
5428 which it needed to allocate memory to do. Due to the low
5429 memory, it would be unable to do so and the node would
5430 become unresponsive.
5432 Severity : enhancement
5434 Description: Addition of lconf --service command line option
5435 Details : lconf now accepts a '--service <arg>' option, which is
5436 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5438 Severity : enhancement
5440 Description: Failover mode is now the default for OSTs.
5441 Details : By default, OSTs will now run in failover mode. To return to
5442 the old behaviour, add '--failout' to the lmc line for OSTs.
5444 Severity : enhancement
5446 Description: Health checks are now provided for MDS and OSTs
5447 Details : Additional detailed health check information on MSD and OSTs
5448 is now provided through the procfs health_check value.
5451 Frequency : occasional, depends on IO load
5453 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5454 after numerous create/delete cycles
5455 Details : The ext3 inode allocation policy would not allocate new inodes
5456 very well on the OSTs because there are no new directories
5457 being created. Instead we look for groups with free space if
5458 the parent directories are nearly full.
5463 Description: Network or server problems during mount may cause partially
5464 mounted clients instead of returning an error.
5465 Details : The config llog parsing code may overwrite the error return
5466 code during mount error handling, returning success instead
5472 Description: MDS can fail to allocate large reply buffers
5473 Details : After long uptimes the MDS can fail to allocate large reply
5474 buffers (e.g. zconf client mount config records) due to memory
5475 fragmentation or consumption by the buffer cache. Preallocate
5476 some large reply buffers so that these replies can be sent even
5477 under memory pressure.
5481 Frequency : rare (liblustre)
5482 Description: fsx running with liblustre complained that using truncate() to
5483 extend the file doesn't work. This patch corrects that issue.
5484 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5485 ATTR_SIZE and lsm use in llu_setattr_raw.
5489 Frequency : rare, only 2.6 kernels
5490 Description: Unusual file access patterns on the MDS may result in inode
5491 data being lost in very rare circumstances.
5492 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5493 code in the 2.6 kernel caused the inode and/or EA data not to
5494 be read from disk, causing single-file corruption.
5498 Frequency : rare, only 2.6 filesystems using extents
5499 Description: Heavy concurrent write and delete load may cause data corruption.
5500 Details : It was possible under high-load situations to have an extent
5501 metadata block in the block device cache from a just-unlinked
5502 file overwrite a newly-allocated data block. We now unmap any
5503 metadata buffers that alias just-allocated data blocks.
5507 Frequency : filesystems with default stripe_count larger than 77
5508 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5509 Details : lconf specifies an inode size of 4096 bytes when the default
5510 stripe_count is larger than 77. This conflicts with the default
5511 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5512 this case to avoid pinning too much memory for large EAs.
5514 ------------------------------------------------------------------------------
5516 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5521 Frequency : rare (extremely heavy IO load with hundreds of clients)
5523 Description: Client is evicted, gets IO error writing to file
5524 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5525 caused clients to be evicted for AST timeouts. The fixes in
5526 bug 5192 mean we no longer need to have such short AST timeouts
5527 so ldlm_timeout has been increased.
5530 Frequency : occasional during --force or --failover shutdown under load
5531 Bugzilla : 5949, 4834
5532 Description: Server oops/LBUG if stopped with --force or --failover under load
5533 Details : a collection of import/export refcount and cleanup ordering
5534 issues fixed for safer force cleanup
5537 Frequency : only filesystems larger than 120 OSTs
5538 Bugzilla : 5990, 6223
5539 Description: lfs getstripe would oops on a very large filesystem
5540 Details : lov_getconfig used kfree on vmalloc'd memory
5543 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5544 Bugzilla : 6242, 6243
5545 Description: reading from files that had been truncated to a non-zero size
5546 but never opened returned no data
5547 Details : ll_file_read() reads zeros from no-object files to EOF
5552 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5553 Details : There's little harm in aborting MDS/OSS recovery and letting it
5554 try again, so I removed the LASSERT and return an error instead.
5556 Severity : enhancement
5558 Description: New debugging infrastructure for tracking down data corruption
5559 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5560 (b) cover more of the client-side code path, and (c) try to narrow
5561 down where problems occurred
5565 Bugzilla : 3819, 4364, 4397, 6313
5566 Description: Racing close and eviction MDS could cause assertion in mds_close
5567 Details : It was possible to get multiple mfd references during close and
5568 client eviction, leading to one thread referencing a freed mfd.
5570 Severity: : enhancement
5571 Bugzilla : 3262, 6359
5572 Description: Attempts to reconnect to servers are now more aggressive.
5573 Details : This builds on the enhanced upcall-less recovery that was added
5574 in 1.4.2. When trying to reconnect to servers, clients will
5575 now try each server in the failover group every 10 seconds. By
5576 default, clients would previously try one server every 25 seconds.
5581 Description: After recovery, certain operations trigger a failed
5582 assertion on a client.
5583 Details : Failing over an mds, using lconf -d --failover, while a
5584 client was doing a readdir() call would cause the client to
5585 LBUG after recovery completed and the readdir() was resent.
5587 Severity : enhancement
5589 Description: Default groups are now added by lconf
5590 Details : You can now run lconf --group <servicename> without having to
5591 manually add groups with lmc.
5594 Frequency : occasional
5596 Description: Nodes with an elan id of 0 trigger a failed assertion
5599 Frequency : always when accessing e.g. tty/console device nodes
5601 Description: tty and some other devices nodes cannot be used on lustre
5602 Details : file's private_data field is used by device data and lustre
5603 values in there got lost. New field was added to struct file to
5604 store fs-specific private data.
5607 Frequency : when exporting Lustre via NFS
5609 Description: NFSD failed occasionally when looking up a path component
5610 Details : NFSD is looking up ".." which was broken in ext3 directories
5611 that had grown large enough to become hashed.
5614 Frequency : Clusters with multiple interfaces not on the same subnet
5616 Description: Nodes will repeatedly try to reconnect to an interface which it
5617 cannot reach and report an error to the log.
5618 Details : Extra peer list entries will be created by lconf with some peers
5619 unreachable. lconf now validates the peer before adding it.
5622 Frequency : Only if a default stripe is set on the filesystem root.
5624 Description: Setting a default stripe on the filesystem root prevented the
5625 filesystem from being remounted.
5626 Details : The client was sending extra request flags in the root getattr
5627 request and did not allocate a reply buffer for the dir EA.
5630 Frequency : occasional, higher if lots of files are accessed by one client
5631 Bugzilla : 6159, 6097
5632 Description: Client trips assertion regarding lsm mismatch/magic
5633 Details : While revalidating inodes the VFS looks up inodes with ifind()
5634 and in rare cases can find an inode that is being freed.
5635 The ll_test_inode() code will free the lsm during ifind()
5636 when it finds an existing inode and then the VFS later attaches
5637 this free lsm to a new inode.
5641 Bugzilla : 6422, 7030
5642 Description: MDS deadlock between mkdir and client eviction
5643 Details : Creating a new file via mkdir or mknod (starting a transaction
5644 and getting the ns lock) can deadlock with client eviction
5645 (gets ns lock and trying to finish a synchronous transaction).
5648 Frequency : occasional
5649 Description: While starting a server, the fsfilt_ext3 module could not be
5651 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5652 kernels. Previously, lconf would still use the ext3 name
5653 when trying to load modules. Now, it will correctly use
5654 ext3 on 2.4 and ldiskfs on 2.6.
5656 Severity : enhancement
5657 Description: The default stripe count has been changed to 1
5658 Details : The interpretation of the default stripe count (0, to lfs
5659 or lmc) has been changed to mean striping across a single
5660 OST, rather than all available. For general usage we have
5661 found a stripe count of 1 or 2 works best.
5663 Severity : enhancement
5664 Description: Add support for compiling against Cray portals.
5665 Details : Conditional compiling for some areas that are different
5669 Frequency : occasional
5670 Bugzilla : 6409, 6834
5671 Description: Creating files with an explicit stripe count may lead to
5672 a failed assertion on the MDS
5673 Details : If some OSTs are full or unavailable, creating files may
5674 trigger a failed assertion on the MDS. Now, Lustre will
5675 try to use other servers or return an error to the
5679 Frequency : occasional
5681 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5682 caused lock timeout during readahead (since 1.4.2).
5683 Details : Processes doing readahead might match a lock that hasn't been
5684 granted yet if there are overlapping and conflicting lock
5685 requests. The readahead process waits on ungranted lock
5686 (original lock is CBPENDING), while OST waits for that process
5687 to cancel CBPENDING read lock and eventually evicts client.
5689 Severity : enhancement
5691 Description: Initial enabling of flock support for clients
5692 Details : Implements fcntl advisory locking and file status functions.
5693 This feature is provided as an optional mount flag (default
5694 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5695 locking are implemented yet, and those that are are not guaranteed
5696 to be completely correct in production environments.
5697 mount -t lustre -o [flock|noflock] ...
5700 Frequency : occasional
5702 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5703 assertion in the ext3 JBD (journaling) layer.
5704 Details : The b_committed_data struct is protected by the big kernel lock
5705 in 2.4 kernels, serializing journal_commit_transaction() and
5706 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5707 access to this struct. In 2.6 kernels there is finer grained
5708 locking to improve SMP performance of the JBD layer.
5712 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5717 Description: merge and cleanup kernel patches.
5720 -----------------------------------------------------------------------------
5722 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5724 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5725 versions of Lustre. Please update all servers and clients to
5726 version 1.4.2 or later at the same time. You must also run
5727 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5728 to update the configuration logs.
5730 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5731 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5732 - swab llog records of type '0' so we get proper header size/idx (5861)
5733 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5734 - fix rename of one directory over another leaking an inode (5953)
5735 - avoid SetPageDirty on 2.6 (5981)
5736 - don't re-add just-being-destroyed locks to the waiting list (5653)
5737 - when creating new directories, inherit the parent's custom
5738 striping settings if present parent (3048)
5739 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5740 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5741 - don't hold client locks on temporary worklist from l_lru (5666)
5742 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5743 - automatically evict dead clients (5921)
5744 - Update file size properly in create+truncate+fstat case (6196)
5745 - Do not unhash mountpoint dentries, do not allow removal of
5747 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5748 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5749 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5750 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5751 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5752 - Correctly handle reads of files with no objects (6243)
5753 - lctl recover will also mark a device active if deactivate used (5933)
5755 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5756 - allow --write-conf on an MDS with different nettype than client (5619)
5757 - don't write config llogs to MDS for mounts not from that MDS (5617)
5758 - lconf should create multiple TCP connections from a client (5201)
5759 - init scripts are now turned off by default; run chkconfig --on
5760 lustre and chkconfig --on lustrefs to use them
5761 - upcalls are no longer needed for clients to recover to failover
5763 - add --abort-recovery option to lconf to abort recovery on device
5765 - add support for an arbitrary number of OSTs (3026)
5766 - Quota support protocol changes.
5767 - forward compatibility changes to wire structs (6007)
5768 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5769 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5770 - improved "lustre status" script
5771 - initialize blocksize for non-regular files (6062)
5772 - added --disable-server and --disable-client configure options (5782)
5773 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5774 - Vanilla 2.4.29 support
5775 - increase maximum number of obd devices to 520 (6242)
5776 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5777 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5778 SLES 9 2.6.5 kernels
5779 - limit stripes per file to 160 (the maximum EA size) (6093)
5781 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5784 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5785 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5786 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5787 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5788 - small ext3 extents cleanups and fixes (5733)
5789 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5790 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5791 - avoid needless client->OST connect, fix handle mismatch (5317)
5792 - fix DLM error path that led to out-of-sync client, long delays (5779)
5793 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5794 - fix several locking issues related to i_size (5492,5624,5654,5672)
5795 - don't move pending lock onto export if it is already evicted (5683)
5796 - fix kernel oops when creating .foo in unlinked directory (5548)
5797 - fix deadlock in obdfilter statistics vs. object create (5811)
5798 - use time_{before,after} to avoid timer jiffies wrap (5882)
5799 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5800 - Do not leak request if server was not able to process it (5154)
5801 - If mds_open unable to find parent dir, make that negative lookup(5154)
5802 - don't create new directories with extent-mapping (5909, 5936)
5804 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5805 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5806 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5807 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5808 - support for RHEL4 kernel on i686 (5773)
5809 - provide error messages when incompatible logs are encountered (5898)
5811 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5812 * version 1.4.0.10 (1.4.1 release candidate 1)
5814 - don't keep a lock reference when lock is not granted (4238)
5815 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5816 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5817 - drop import inflight refcount on signal_completed_replay error (5255)
5818 - unlock page after async write error during send (3677)
5819 - handle missing objects in filter_preprw_read properly (5265)
5820 - no transno return for symlink open, don't save no-trasno open (3440)
5821 - don't try to complete elan receive that already failed (4012)
5822 - free RPC server reply state on error (5406)
5823 - clean up thread from ptlrpc_start_thread() on error (5160)
5824 - readahead could read extra page into cache that wasn't ejected (5388)
5825 - prevent races in class_attach/setup/cleanup/detach (5260)
5826 - don't dereference de->d_inode after l_dput of de (5458)
5827 - use "int" for stripe value returned from lock_to_stripe (5544)
5828 - mballoc allocation and error-checking fixes in 2.6 (5504)
5829 - block device patches to fix I/O request sizes in 2.6 (5482)
5830 - look up hostnames for IB nals (5602)
5831 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5832 - don't start multiple acceptors for the same port (5277)
5833 - fix incorrect LASSERT in mds_getattr_name (5635)
5834 - export a proc file for general "ping" checking (5628)
5835 - fix "lfs check" to not block when the MDS is down (5628)
5837 - service request history (4965)
5838 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5839 - create an "evict_client" /proc entry on OSTs, like the MDS has
5840 - fix mount usage message, return errors per mount(8) (5168)
5841 - change grep [] to grep "[]" in tests so they work in more UMLs
5842 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5843 - remove ancient LOV_MAGIC_V0 EA support (5047)
5844 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5845 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5846 - allow readahead allocations to fail when low on memory (5383)
5847 - mmap locking landed again, after considerable improvement (2828)
5848 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5850 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5853 - send OST transaction number in read/write reply to free req (4966)
5854 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5855 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5856 - the watchdog thread now runs as interruptible (5246)
5857 - flock/lockf fixes (but it's still disabled, pending 5135)
5858 - don't use EXT3 constants in llite code (5094)
5859 - memory shortage at startup could cause assertion (5176)
5861 - reorganization of lov code
5862 - single portals codebase
5864 - add extents/mballoc support (5025)
5865 - direct I/O reads in the obdfilter (4048)
5866 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5868 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5871 - send OST transaction number in read/write reply to free req (4966)
5872 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5873 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5874 - the watchdog thread now runs as interruptible (5246)
5875 - handle missing objects in filter_preprw_read properly (5265)
5876 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5877 - drop import inflight refcount on signal_completed_replay error (5255)
5878 - unlock page after async write error during send (3677)
5879 - return original error code on reconstructed replies (3761)
5880 - no transno return for symlink open, don't save no-trasno open (3440)
5882 - add pid to ldlm debugging output (4922)
5883 - bump the watchdog timeouts -- we can't handle 30sec yet
5884 - extra debugging for orphan dentry/inode bug (5259)
5886 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5889 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5890 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5891 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5892 - fix NULL dereference in /proc/sys/portals/routes (4827)
5893 - allow failed mdc_close() operations to be interrupted (4561)
5894 - stop precreate on OST before MDS would time out on it (4778)
5895 - don't send partial-page writes before EOF from client (4410)
5896 - discard client grant for sub-page writes on large-page clients (4520)
5897 - don't free dentries not owned by NFS code, check generation (4806)
5898 - fix lsm leak if mds_create_objects() fails (4801)
5899 - limit debug_daemon file size, always print CERROR messages (4789)
5900 - use transno after validating reply (3892)
5901 - process timed out requests if import state changes (3754)
5902 - update mtime on OST during writes, return in glimpse (4829)
5903 - add mkfsoptions to LDAP (4679)
5904 - use ->max_readahead method instead of zapping global ra (5039)
5905 - don't interrupt __l_wait_event() during strace
5907 - add software watchdogs to catch hung threads quickly (4941)
5908 - make lustrefs init script start after nfs is mounted
5909 - fix CWARN/ERROR duplication (4930)
5910 - return async write errors to application if possible (2248)
5911 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5912 - print NAL number in %x format (4645)
5913 - update barely-supported suse-2.4.21-171 series (4842)
5914 - support for sles 9 %post scripts
5915 - support for building 2.6 kernel-source packages
5916 - support for sles km_* packages
5918 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5921 - ignore -ENOENT errors in osc_destroy (3639)
5922 - notify osc create thread that OSC is being cleaned up (4600)
5923 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5924 - reconstruct ost_handle() like mds_handle() (4657)
5925 - create a new thread to do import eviction to avoid deadlock (3969)
5926 - let lconf resolve symlinked-to devices (4629)
5927 - don't unlink "objects" from directory with default EA (4554)
5928 - hold socknal file ref over connect in case target is down (4394)
5929 - allow more than 32000 subdirectories in a single directory (3244)
5930 - fix blocks count for O_DIRECT writes (3751)
5931 - OST returns ENOSPC from object create when no space left (4539)
5932 - don't send truncate RPC if file size isn't changing (4410)
5933 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5934 - bind to privileged port in socknal and tcpnal (3689)
5936 - rate limit CERROR/CWARN console message to avoid overload (4519)
5937 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5938 - basic mmap support (3918)
5939 - kernel patch series update from b1_4 (4711)
5941 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5944 - avoid crash during MDS cleanup with OST shut down (2775)
5945 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5946 - don't use bad inodes on the MDS (3744)
5947 - dynamic object preallocation to improve recovery speed (4236)
5948 - don't hold spinlock over lock dumping or change debug flags (4401)
5949 - don't zero obd_dev when it is force cleaned (3651)
5950 - print grants to console if they go negative (4431)
5951 - "lctl deactivate" will stop automatic recovery attempts (3406)
5952 - look for existing locks in ldlm_handle_enqueue() (3764)
5953 - don't resolve lock handle twice in recovery avoiding race (4401)
5954 - revalidate should check working dir is a directory (4134)
5956 - don't always mark "slow" obdfilter messages as errors (4418)
5958 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5961 - don't close LustreDB during write_conf until it is done (3860)
5962 - fix typo in lconf for_each_profile (3821)
5963 - allow dumping logs from multiple threads at one time (3820)
5964 - don't allow multiple threads in OSC recovery (3812)
5965 - fix debug_size parameters (3864)
5966 - fix mds_postrecov to initialize import for llog ctxt (3121)
5967 - replace config semaphore with spinlock (3306)
5968 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5969 - don't allow enqueue to complete on a destroyed export (3822)
5970 - down write_lock before checking llog header bitmap (3825)
5971 - recover from lock replay timeout (3764)
5972 - up llog sem before sending rpc (3652)
5973 - reduce ns lock hold times when setting kms (3267)
5974 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5975 - fix NULL deref and obd_dev leak on setup error (3312)
5976 - replace some LBUG about llog ops with error handling (3841)
5977 - don't match INVALID dentries from d_lookup and spin (3784)
5978 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5979 - fix invalid assertion in ptlrpc_set_wait (3880)
5981 - add libwrap support for the TCP acceptor (3996)
5982 - add /proc/sys/portals/routes for non-root route listing (3994)
5983 - allow setting MDS UUID in .xml (2580)
5984 - print the stack of a process that LBUGs (4228)
5986 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5989 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5990 - make sure to unset replay flag from failed open requests (3440)
5991 - if default stripe count is 0, use OST count for inode size (3636)
5992 - update parent mtime/ctime on client for create/unlink (2611)
5993 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5994 - free recovery state on server during a forced cleanup (3571)
5995 - unregister_reply for resent reqs (3063)
5996 - loop back devices mounting and status check on 2.6 (3563)
5997 - fix resource-creation race that can provoke i_size == 0 (3513)
5998 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5999 - more debugging for page-accounting assertion (3746)
6000 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6001 - avoid deadlock after precreation failure (3758)
6002 - fix race and lock order deadlock in orphan handling (3450, 3750)
6003 - add validity checks when grabbing inodes from l_ast_data (3599)
6005 - add /proc/.../recovery_status to obdfilter (3428)
6006 - lightweight CDEBUG infrastructure, debug daemon (3668)
6007 - change default OSC RPC parameters to be better on small clusters
6008 - turn off OST read cache for files smaller than 32MB
6009 - install man pages and include them in rpms (3100)
6010 - add new init script for (un)mounting lustre filesystems (2593)
6011 - run chkconfig in %post for init scripts (3701)
6012 - drop scimac NAL (unmaintained)
6014 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6017 - clean kiobufs before and after use (3485)
6018 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6019 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6020 - change init script to fail more clearly if not run as root (1528)
6021 - allow clients to reconnect during replay (1742)
6022 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6023 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6024 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6025 - fix cond_resched (3554)
6026 - only evict unfinished clients after recovery (3515)
6027 - allow bulk resend, prevent data loss (3570)
6028 - dynamic ptlrpc request buffer allocation (2102)
6029 - don't allow unlinking open directory if it isn't empty (2904)
6030 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6031 - remove extraneous obd dereference causing LASSERT failure (3334)
6032 - don't use get_cycles() when creating temp. files on the mds (3156)
6033 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6034 - handle EEXIST for set-stripe, set proper directory name (3336)
6036 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6037 - fix ksocknal_fmb_callback() error messages (2918)
6039 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6042 - don't copy lvb into (possibly NULL) reply on error (2983)
6043 - don't deref dentry after dput, don't free lvb on error (2922)
6044 - use the kms to determine writeback rpc length (2947)
6045 - increment oti_logcookies when osc is inactive (2948)
6046 - update client's i_blocks count via lvb messages (2543)
6047 - handle intent open/close of special files properly (1557)
6048 - mount MDS with errors=remount-ro, like obdfilter (2009)
6049 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6050 - don't use cancelling-locks' kms values (2947)
6051 - use highest lock extent for kms, not last one (2925)
6052 - don't dereference ERR_PTR() dentry in error handling path (3107)
6053 - fix thread race in portals_debug_dumplog() (3122)
6054 - create lprocfs device entries at setup instead of at attach (1519)
6055 - common AST error handler, don't evict client on completion race (3145)
6056 - zero nameidata in detach_mnt in 2.6 (3118)
6057 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6058 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6059 - initialize RPC timeout timer earlier for 2.6 (3219)
6060 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6061 - print nal/nid for unknown nid (3258)
6062 - additional checks for oscc recovery before doing precreate (3284)
6063 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6064 - don't crash in mdc_close for bad permissions on open (3285)
6065 - zero i_rdev for non-device files (3147)
6066 - clear page->private before handing to FS, better assertion (3119)
6067 - tune the read pipeline (3236)
6068 - fix incorrect decref of invalidated dentry (2350)
6069 - provide read-ahead stats and refine rpc in flight stats (3328)
6070 - don't hold journal transaction open across create RPC (3313)
6071 - update atime on MDS at close time (3265)
6072 - close LDAP connection when recovering to avoid server load (3315)
6073 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6074 - don't leak open file on MDS after open resend (3325)
6075 - serialize filter_precreate and filter_destroy_precreated (3329)
6076 - loop device shouldn't call sync_dev() for nul device (3092)
6077 - clear page cache after eviction (2766)
6078 - resynchronize MDS->OST in background (2824)
6079 - refuse to mount the same filesystem twice on same mountpoint (3394)
6080 - allow llmount to create routes for mounting behind routers (3320)
6081 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6082 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6083 - fix rare problem with rename on htree directories (3417)
6085 - allow default OST striping configuration per directory (1414)
6086 - fix compilation for qswnal for 2.6 kernels (3125)
6087 - increase maximum number of MDS request buffers for large systems
6088 - change liblustreapi to be useful for external progs like lfsck (3098)
6089 - increase local configuration timeout for slow disks (3353)
6090 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6092 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6095 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6096 - don't overwrite extent policy data in reply if lock was blocked (2901)
6097 - drop filter export grants atomically with removal from device (2663)
6098 - del obd_self_export from work_list in class_disconnect_exports (2908)
6099 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6100 - swab reply message in mdc_close, other PPC fixes (2464)
6101 - fix destroying of named logs (2325)
6102 - overwrite old logs when running lconf --write_conf (2264)
6103 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6104 - fix race in target_handle_connect (2898)
6105 - mds_reint_create() should take same inode create lock (2926)
6106 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6107 - don't close files for self_export to avoid uninitialized obd (2936)
6108 - allow MDS with the same name as client node (2939)
6109 - hold dentry reference for closed log files for unlink (2325)
6110 - reserve space for all logs during transactions (2059)
6111 - don't evict page beyond end of stripe extent (2925)
6112 - don't oops on a deleted current working directory (2399)
6113 - handle hard links to targets without a parent properly (2517)
6114 - don't dereference NULL lock when racing during eviction (2867)
6115 - don't grow lock extents when lots of conflicting locks (2919)
6117 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6120 - account for cache space usage on clients to avoid data loss (974)
6121 - lfsck support in lustre kernel code (2349)
6122 - reduce journal credits needed for BRW writes (2370)
6123 - orphan handling to avoid losing space on client/server crashes
6124 - ptlrpcd can be blocked, stopping ALL progress (2477)
6125 - use lock value blocks to assist in proper KMS, faster stat (1021)
6126 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6127 - recovery for initial connections (2355)
6128 - fixes for mds_cleanup_orphans (1934)
6129 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6130 - block all file creations until orphan recovery completes (1901)
6131 - client remove rq_connection from request struct (2423)
6132 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6133 - recovery timer race (2670)
6134 - mdc_close recovey bug (2532)
6135 - ptlrpc cleanup bug (2710)
6136 - mds timeout on local locks (2588)
6137 - namespace lock held during RPCs (2431)
6138 - handle interrupted sync write properly (2503)
6139 - don't try to handle a message that hasn't been replied to (2699)
6140 - client assert failure during cleanup after abort recovery (2701)
6141 - leak mdc device after failed mount (2712)
6142 - ptlrpc_check_set allows timedout requests to complete (2714)
6143 - wait for inflight reqs when ptlrpcd finishes (2710)
6144 - make sure unregistered services are removed from the srv_list
6145 - reset bulk XID's when resending them (caught by 1138 test)
6146 - unregister_bulk after timeout
6147 - fix lconf error (2694)
6148 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6149 - readahead locks pages, leaves pending causing memory pressure (2673)
6150 - increase OST request buffers to 4096 on large machines (2729)
6151 - fix up permission of existing directories in simple_mkdir (2661)
6152 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6153 - don't assign transno to errored transactions (2742)
6154 - don't delete objects on OST if given a bogus objid from MDS (2751)
6155 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6156 - if rq_no_resend, then timeout request after recovery (2432)
6157 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6158 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6159 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6160 - MDS recovery shouldn't delete objects that it has given out (2730)
6161 - if enqueue arrives after completion, don't clobber LVB (2819)
6162 - don't unlock pages twice when trigger_group_io returns error (2814)
6163 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6164 - don't write pages to disk if there was an error (1450)
6165 - don't ping imports that have recovery disabled (2676)
6166 - take buffered bytes into account when balancing socknal conn (2817)
6167 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6168 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6169 - remove little-endian swabbing of llog records (1987)
6170 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6171 - retry reposting request buffers if they fail (1191)
6172 - grow extent at grant time to avoid granting a revoked lock (2809)
6173 - lock revoke doesn't evict page if covered by a second lock (2765)
6174 - disable VM readahead to avoid reading outside lock extents (2805)
6176 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6177 - updated kernel patches for hp-2.4.20 kernel (2681)
6179 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6182 - fix truncated write corruption (2366)
6183 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6185 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6186 - ensure len(uuid) < 37 in lmc (1171)
6187 - fix ia64 OOPS in llog_test (2255)
6188 - zero end of page at obdfilter for partial page writes (2648)
6189 - don't leave stale dentries around after renames (bug 2428)
6190 - fix timeouts when evicting a client with a single lock held (2642)
6191 - set deadline for the initial HELLO message to drain (2634)
6192 - print out dotted-quad IP addresses in the socknal (2302)
6194 - additional debugging for MDS client eviction problem (2443)
6195 - fix mkfsoptions support for osts (2603, 2604)
6197 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6200 - add series for the vanilla 2.6.0 kernel
6201 - add series for the vanilla 2.4.24 kernel
6202 - add series for a cray x86/64 UL kernel drop
6203 - fix xattr patches for the vanilla 2.4.19 series
6205 - generate true UUIDs in lmc (1171)
6206 - have portals stack dumping break in UML (2466)
6207 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6208 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6209 - wake callback waiting threads on client eviction (2460)
6210 - Add --ptldebug and --subsystem to lmc (1719)
6211 - update assertion to allow safe interrupt allocation
6212 - set rq_no_resend for cancel requests (2432)
6213 - recalculate ptlrpcd timeout after resend (2494)
6214 - call vfs_rmdir when removing pending directories (2368)
6215 - fix renaming a file to itself (2429)
6216 - lmc creates a default one-stripe lov (2454)
6217 - expand procfs space to handle large clusters (2326)
6218 - increase UML stack to avoid overflow
6219 - update lconf's list of debug and subsystem masks
6220 - fix lfs find --obd (2510)
6221 - /proc tunable for disabling filter read caching (2591)
6222 - stop rpm packages from altering slapd.conf (2301)
6223 - disable nagle in the socknal under 0conf (2578)
6224 - choose mds inode size based on stripe count (2572)
6225 - fix kernel-source rpm problems (2516)
6227 - add --disable-doc to avoid pdf generation (2421)
6228 - update documentation, tests, type-os, comments
6229 - avoid format warnings on ia64
6230 - remove the TOE NAL
6231 - tiny code cleanups by removing unused fields
6233 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6236 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6237 - remove the most blatant lies from BUILDING file (2371)
6238 - change default debug level to reasonable production setting
6239 - reduce client side cache size to reduce cache flush time
6240 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6241 - make TCP zerocopy and pinger support enabled by default (2476)
6242 - sync writes completed after process exits caused crashes (2319)
6243 - maintain correct mount count on the MDS (2356)
6244 - backout 1557, because 2316 wasn't really fixed
6245 - better file I/O statistics gathering in /proc
6246 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6247 - another kernel patch to fix zero-copy TCP function export
6248 - don't take duplicate lock when processing re-sent getattr (2420)
6249 - lctl uses obd_self_export instead of creating new conn (2353)
6250 - MDS/OST recovery case which requires object creation asserted (2425)
6251 - move lfs from /usr/sbin to /usr/bin in packages
6252 - fix race between mds_client_add and mds_client_free (2417)
6253 - use kmalloc instead of slabs in portals (2430)
6254 - don't create duplicate records when a failover MDS is present (2442)
6255 - remove unnecessary mount age check (2332)
6256 - don't remove directory inodes from locks prematurely (2451)
6257 - don't break if MDS service name is the same as hostname (2103)
6258 - fix races in client write RPC generation when cache full (2482)
6260 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6263 - remove now-unused request->rq_obd (278)
6264 - if an allocation fails, print out how much memory we've used (1933)
6265 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6266 - add the "configurable stack size" patch to most series files (1256)
6267 - ability to write large log records, for 100+ OST configs (2306)
6268 - fix NULL deref when filter_prep fails (2314)
6269 - fix operator precedence error in filter_sync
6270 - dynamic allocation of socknal TX descriptors (2315)
6271 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6272 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6273 - ensure that CWARN messages in Portals always get to the syslog
6274 - __init/__exit are not for prototype decls (ldlm_init/exit)
6275 - x86-64 compile warning fixes
6276 - fix gateway LMC keyword conflict (2318)
6277 - fix MDS lock inversions in getattr/reint paths (1844)
6278 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6279 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6280 - DLM race condition prevented some lock evictions (2328)
6281 - ENOMEM detection and retry on socknal sends (2230)
6282 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6283 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6284 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6285 - better error messages when a client is rejected during recovery (1505)
6286 - avoid cancelling locks which were never granted, after failure (2330)
6287 - fix i_sem/journal inversion in mds_client_add (2333)
6288 - fix truncate/getattr lock cycle deadlock (2334)
6289 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6290 - fix two rare exit paths which could leak an l_lock() ref (2321)
6291 - fencepost error in MDS/OST orphan recovery (2226)
6292 - make log record alignment 8 bytes (1988)
6293 - lstripe now fails when requested offset > ost_count (2237)
6294 - ensure that all kernel series have a complete list.h (1607)
6295 - fix crashes in special-file operations (2316)
6296 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6298 - allow configurable automake binary, for testing new versions
6299 - small update to the lfs documentation
6301 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6303 * fix negative export reference count in fsfilt_sync (2312)
6305 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6306 * release candidate 0.9.1
6308 - orphans are moved into the PENDING directory for possible recovery
6309 - replayed opens now open by fid for orphan/rename safety (1042)
6310 - last close of an orphan inode generates a transno (683)
6311 - chdir() and mount() now pin the directory entry (1020)
6312 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6313 - discard very old requests without processing them (1502)
6314 - remove some common, well-understood CERRORs (1505)
6315 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6316 - clear "grant" flags in OST replies until OST grant code lands (1644)
6317 - fix read performance by not clobbering i_blksize on client (1598)
6318 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6319 - make LRU size a /proc tunable, clears locks when reduced (707)
6320 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6321 - server locks take references on exports now (1558)
6322 - build fixes for 2.4.20-rh trees (1663)
6323 - return an error from lov_create if all OSCs are inactive (1751)
6324 - fix import levels when a reconnect happens without a timeout (1597)
6325 - exit early from mds_open if we get a lookup error (1749)
6326 - partial page read at EOF wouldn't wait for disk before sending (1642)
6327 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6328 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6329 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6330 - write/truncate lock inversion (1639)
6331 - Don't auto-load obdclass, portals modules during cleanup (1495)
6332 - fix timestamps from jumping to "now" (1763)
6333 - extra journal assertions (1648)
6334 - add an extra multiunlink test (1771)
6335 - fix read_record/write_record API (1776)
6336 - fix leak of offset_extent, possible incorrect i_size later (1772)
6337 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6338 - Debugging for the kqswnal_get_idle_tx problems (1820)
6339 - Allow recovery to be attempted multiple times (1536)
6340 - Write out MDS last_rcvd file after it is first created (1600)
6341 - Fix tx_descriptor leak in failed transmit situations (1827)
6342 - ext3 journaling fixes for assertion failure after IO error (1871)
6343 - class_export_put() on freed export after completion AST error (1896)
6344 - Fix revalidate looping in VFS (1322)
6345 - Don't access a freed export during MDS_REINT timeout (1521)
6346 - Add open-unlink recovery support on the MDS (1673,1764)
6347 - Return an error if no MDS data was read from last_rcvd (1946)
6348 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6349 - Don't setup a disk device that doesn't match exported UUID (317)
6350 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6351 - avoid committing NULL handle in force close
6352 - local.sh is now a one-stripe LOV configuration
6353 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6354 - fix ext3 htree duplicate directory entry corruption (1516)
6355 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6356 - update to new LOV EA format (2097)
6357 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6359 2003-06-15 Phil Schwan <phil@clusterfs.com>
6362 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6363 - per-import/export recovery handling (958, 931, 959)
6364 - multiple last-rcvd slots, for serving multiple FSes (949)
6365 - connections are again shared between multiple imp/exports (963, 964)
6366 - "umount -f" would hang if any requests needed to be sent (393, 978)
6367 - avoid pinning large req buffer by copying for queued messages (989)
6368 - add "uuid" to "lctl device" command to help upcalls (991)
6369 - "open" RPCs with transnos would confuse recovery counters (1037)
6370 - do proper endian conversion of all wire messages (288, 340, 891)
6371 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6372 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6373 - fix ptlrpc_connection leak in target_handle_connect (1174)
6374 - fix import refcounting bug in OST and MDS cleanup (1134)
6375 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6376 - fix very unlikely obd_types race condition (501)
6377 - remove osc_open hack for echo_client (1187)
6378 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6379 - a failure in read_inode2 leads to deadlock (1139)
6380 - cancel ack-locks as soon as transaction is committed (1072)
6381 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6382 - make sure to commitrw after any preprw to avoid deadlock (1162)
6383 - failing to execute a file in a lustre FS would lock inode (1203)
6384 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6385 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6386 - obd_uuid2tgt would walk off the end of the list (1255)
6387 - on IA64 the osc would give portals incorrect bulk size (1258)
6388 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6389 - fix lock inversion caused by new llite matching code (1282)
6390 - limit the number of dirty pages on a client to 10MB (1286)
6391 - timed out locks were not being corrected cancelled (1289)
6392 - fix O_DIRECT above 4GB on IA-32 (1292)
6393 * major user-visible changes
6394 - fail out/fail over policy now controlled by the upcall (993)
6396 - add OBD_PING to check server availability and failure (954)
6397 - lustre messages are now sent in sending host order (288, 340, 891)
6398 - add eadatalen to MDS getattr reply (340)
6399 - OST read replies may contain second buffer, with per-page status (593)
6401 2003-03-11 Phil Schwan <phil@clusterfs.com>
6404 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6405 - failed open()s could cause deadlock; fixed (867, 869)
6406 - stop cancelling OST locks when files are closed (481)
6407 - overlapping XID spaces caused network corruption (851, 853)
6408 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6409 - setattr_raw added, to do single-RPC, server-side setattrs
6410 - lmc/lconf syntax change for OST UUIDs
6411 - fix crashy race condition between ptlrpc_free_req and osc_close
6412 - don't use request in mdc_enqueue if we hit a timeout (889)
6413 - don't set the inode i_size for regular files from the MDS (896)
6414 - handle out of order completion AST (842)
6415 - don't LBUG if a lock request times out after receiving AST (913)
6416 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6417 - if a bad lock AST arrives, send an error instead of dropping entirely
6418 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6419 - fix leak in bulk IO when only partially completed (899, 900, 926)
6420 - fix O_DIRECT for ia64 (55)
6421 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6422 - C-z after timeout could hang a process forever; fixed (977)
6424 - client-side I/O cache (678, 924, 929, 941, 970)
6426 - READPAGE and SETATTRs which don't take server-side locks get
6429 2003-02-11 Phil Schwan <phil@clusterfs.com>
6432 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6433 - Fix fsfilt_extN_readpage() to read a full page of directory
6434 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6435 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6436 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6437 - Most intent operations were removed, in favour of a new RPC mode
6438 that does a single RPC to the server and bypasses most of the VFS
6439 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6440 - Aggressively cancel local locks on DLM servers
6441 - mds_reint_unlink sends EA to the client if it's the last nlink.
6442 client uses that EA to unlink OST objects.
6443 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6444 - recursive symlinks were fixed (439)
6445 - fixed NULL deref in DEBUG_REQ
6446 - filter_update_lastobjid no longer calls sync, which annoyed extN
6447 - fixed multi-client small-writes to a single file problem (445)
6448 - fixed mtime updates during file writes (607)
6449 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6450 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6451 - fixed Linux OST size reporting problem (444, 656)
6452 - OST now updates object mtime with writes or setattr (607, 619)
6453 - client verifies file size before zeroing page past EOF (445)
6454 - OST now writes last allocated objid to disk with allocation (108)
6455 - LOV on echo now works (409)
6457 - mds_reint_unlink sends a new buffer, with the EA included. this
6458 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6459 set if a regular file was being unlinked, and it was the last link
6460 - use PtlGet from the target for bulk writes (315)
6461 - OST now updates object mtime with writes or setattr (607, 619)
6462 - LDLM now has a grant-time callback to revalidate locked items, if
6464 - Many MDS operations were reorganized to combat race conditions
6466 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6467 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6469 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6472 - Fully reactivate OST imports after reconnection (512, others)
6473 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6474 - More graceful error handling for truncating on dead OST (515)
6475 - Don't error out unless we're actually accessing dead stripes (474)
6476 - Fix garbage sizes when stripes are missing (410)
6477 - LRU counters were broken, causing constant lock purge (433, 432)
6478 - garbage on read from stripes with failed OSTs (441)
6479 - mark OSCs as active before reconnecting during recovery (438)
6480 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6481 - lfind did not preserve OST order in output (443)
6482 - symlinks cause hung clients, incorrect data (439)
6483 - stop dereferencing request after dropping refcount (457)
6484 - don't LASSERT(spin_is_locked) on non-SMP (455)
6485 - fixes for many rename() bugs
6486 - fstat didn't correctly synchronize attributes (399)
6487 - server must handle lock cancellation during blocking AST prep (487)
6488 - bulk descriptors were free()d too soon (511)
6489 - fix paths in lconf, which would load incorrect modules (451, 507)
6490 - fix confusing lconf 'host not found' error message (386)
6491 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6492 - fix race condition in mdc_blocking_ast() for inode access (526)
6493 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6494 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6495 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6496 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6497 - fix lstripe in several ways (kernel side) (527)
6498 - fix request leak in ldlm_cli_enqueue (262)
6499 - incorrect OSC was marked inactive after OST failure
6500 - call mds_fs_cleanup before unmounting filesystem (524)
6501 - fix races between taking ns_lock and ldlm_lock_change_resource
6502 - fix races updating LOV export open file list
6503 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6504 - fix recovery NULL deref in ldlm_cli_cancel_unused
6505 - fix some DLM races by using new hash table for lock handles (419)
6506 - permit the client to specify desired inodes, at replay
6507 - duplicate requests when we queue them for replay reintegration
6508 - fix last_rcvd offset calculation
6509 - sync after each recovered transaction, so we always make progress
6510 - never, not always, ERESTART requests without transnos
6511 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6513 - skip replay if the MDS says that the client is already connected
6514 - don't check for a recovery-enabled export to match lctl's UUID
6515 - don't INC_USE_COUNT for phantom exports
6516 - don't crash when cleaning up phantom exports (567)
6517 - don't double-finish or set replay data for errored mdc_open requests
6518 - abort requests when they time out, so we don't get old replies
6519 - send/receive replies for AST messages again
6520 - if the client says that it doesn't have the lock, cancel it on the
6522 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6523 mark it as destroyed, it will all work out when decref is called
6524 - fix module use counts (22, 581)
6526 - ASTs now expect a reply (server cancels lock on error reply)
6528 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6531 - fix many simultaneous client startup (392)
6532 - fix dentry->d_it clobbering
6533 - credentials weren't being shipped for readdir/getattr operations
6534 - remove invalid assertions triggered during some concurrent MD
6536 - proper Lustre versions added (336, 389)
6537 - fix memory leak for create error case (398)
6538 - fix LOV locking bug that would get cli/srv out of sync
6539 - fix echo client over LOV (409)
6540 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6541 - fix double-O_EXCL intent crash (424)
6542 - avoid sending multiple lock CANCELs (352)
6544 - MDS can do multi-client recovery (modulo bugs in new code)
6546 - many updates, edits, cleanups
6548 2002-11-18 Phil Schwan <phil@clusterfs.com>
6551 - fix null d_it dereference (346)
6552 - fix full OST/dbench hang (333)
6553 - fix permission problem with file removal (286)
6554 - fix removal of OSCs from LOV when they fail
6555 - fix NULL deref during bulk timeout (214)
6556 - fix problems related to multiple filesystems on one MDS (241)
6557 - fixed serious subtle metadata locking bugs
6558 - free locks on clients when inodes are removed due to memory
6560 - fix inode pointer in lock data (285)
6561 - partial support for multiple MDS on a single host (241)
6562 - data locks weren't cancelled at clear_inode time (290, 311)
6563 - intent locks could lead to unbounded lock growth (205)
6564 - added a maximum lock count, an LRU list, and a flusher
6565 - fix multiple rename (365)
6566 - properly abstracted the echo client
6567 - OSC locked 1 byte too many; fixed
6568 - rewrote brw callback code:
6569 - fixed recovery bugs related to LOVs (306)
6570 - fixed too-many-pages-in-one-write crash (191)
6571 - fixed (again) crash in sync_io_timeout (214)
6572 - probably fixed callback-related race (385)
6574 - Add capability to MDS protocol
6575 - LDLM cancellations and callbacks on different portals
6577 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6580 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6581 - defer open object destruction to close time (601981, 138)
6582 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6583 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6584 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6585 - don't call obd_setattr in ll_file_release for destroyed objects
6586 * protocol change to lustre_msg: move |version| and add |flags|
6587 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6588 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6589 * added replay of create, unlink, link and rename operations during
6590 MDS failover; recovery should be much more robust now
6591 * remove failed OSCs from LOVs (only lov_create uses this so far)
6592 * the lustre-HOWTO was brought (more) up to date (582544)
6594 2002-10-23 Phil Schwan <phil@clusterfs.com>
6597 - in-use dentries weren't being reused properly (617851)
6598 - prevent multiple LDLM setup (599178)
6599 - fix LOV size calculations for truncate (617853)
6600 - fix client handling of MDS intent errors (POSIX)
6601 - fix permission bug in lovstripe.c test (624321)
6602 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6603 - truncate past end of file could corrupt data
6604 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6605 - a race in recovery could return ETIMEDOUT to apps (623947)
6606 - building outside the source directory was fixed
6607 * the lustre-HOWTO was brought (more) up to date (582544)
6608 * major progress was made on recovery functionality
6610 2002-10-10 Phil Schwan <phil@clusterfs.com>
6613 - recovery deadlock fix
6614 - rm -rf causes LBUG fix (617817)
6615 - file open by multiple tasks fix (618962)
6616 - directory permissions bugs (602707 and 620007)
6617 - journal_stop fixed with locking (611313)
6618 - O_APPEND failures resolved (618273, perhaps 614459)
6619 - lconf PATH fix (619770)
6620 - IA64 build fix (621450)
6621 - RPC buffer sizes scale with amount of memory
6623 2002-10-01 Phil Schwan <phil@clusterfs.com>
6626 - locks would be cancelled without throwing away data pages,
6627 resulting in inconsistent data (605627)
6628 - inode attributes were not always being refreshed (605627, 612449)
6629 - lconf now continues to cleanup after lctl reports an error
6630 - MDS now enforces user permissions (602707)
6631 - lprocfs cleanup fixed, but not yet enabled (614157)
6632 - fixed infinite server hang, should a client not respond to an AST
6633 - avoid going into recovery if user calls readlink() with a buffer
6634 that's too small (613941)
6635 - AST RPCs no longer require replies (614867) -- this may be changed
6636 - don't crash server if client sends an IOV that's too big (611336)
6637 - fixed lock conversion deadlock (611892)
6638 - fixed the following of symlinks (614622)
6639 * recovery: the server can remove locks from a client that dies, other
6640 clients can make progress
6641 * more extN patch fixes
6642 * compile-time configurable ptlrpc buffer allocations
6644 - collaborative read cache document
6645 - Lustre Lite Performance CDR document-in-progress
6647 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6650 - fix typo in patch-2.4.18
6652 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6655 - clear ptlrpc request each time in handle_incoming_request()
6656 - unlink of files now destroys the object on the OST
6658 2002-09-19 Peter Braam <braam@clusterfs.com>
6660 * add hard link support
6661 * change obdfile creation method
6662 * kernel patch changed
6664 2002-09-19 Peter Braam <braam@clusterfs.com>
6667 - stack overflow bug in extN fixed
6669 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6671 * documentation updates
6672 - add man pages for config tools
6673 - update tests/README to describe testing with new config tools
6674 - finish metadata API descriptions
6675 * bug fixes and cleanups
6676 - statfs workaround for 16TB limit
6677 - LOV stripe allocation improved, can stripe on subset of OSTs
6678 - LOV file size/IO offset was wrong for files > 4GB in size
6679 - object EA data was being dropped, caused files to be unreadable
6680 - memory overflow with non-LOV OST caused memory corruption
6681 - fixed regression tests to work with new config tools, obdfilter
6682 - fixed bug when directory size became larger than 1 block
6683 - fixed bug (for single client case) when PWD was deleted
6684 - invalidate local directory pages when doing intent-based ops
6685 - avoid LDLM oops when lock callback contained bad data
6687 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6689 * documentation updates
6690 * bug fixes and cleanups
6691 - configuration tools
6694 - 64-bit compile warnings
6695 - 64-bit internal statfs data
6697 * test_brw on persistent OST devices
6699 * lprocfs (disabled)
6701 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6703 * documentation updates
6704 * bug fixes and cleanups
6705 * configuration tools
6707 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6710 * many small fixes to 0_5_4
6711 * io/network handling
6712 * thinkos in MDS operations
6714 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6717 * crucial basic fixes to 0.5.3
6718 * IOR, Iozone work over Elan
6721 2002-08-07 Phil Schwan <phil@clusterfs.com>
6722 * version 0_5_3, our first alpha
6723 * we use the new Portals iovs
6724 * documentation updates
6725 * bug fixes and cleanups
6726 * small changes in the DLM wire protocol
6728 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6729 * version 0_5_1 with some initial stability,
6730 * locking on MD and file I/O.
6731 * documentation updates
6732 * several bug fixes since 0.5.0
6733 * small changes in wire protocol
6735 2002-07-18 Phil Schwan <phil@clusterfs.com>
6737 * delivered as Lustre Light Alpha
6738 * fixed a crash after handling invalid MDS requests
6739 * fixed directory pages for architectures with non-4k pages sizes
6741 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6742 * release version v0_4_4
6743 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6745 * Fixes a number of interruption problems with OST operations.
6746 * Update documentation for portals header changes
6747 * Move all wire protocol structs/defines to lustre_idl.h
6748 * Fixes symlink length bug.
6749 * Add tcpdump to repository.
6751 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6752 * release version v0_4_3
6753 * Fixes statfs for inodes on extN.
6754 * Fixes bug in runtests which would delete /etc/hosts.
6755 * Use 64-bit object IDs wherever possible (not into VFS though)
6756 Remove ost_get_info, which is unused by lustre, and out of date.
6758 2002-07-03 Peter Braam <braam@clusterfs.com>
6759 * release version v0_4_2 Fixes a lookup error (type not passed)
6760 * move forward to head of Portals
6761 * move forward to latest Lustre kernel
6763 2002-06-25 Peter Braam <braam@clusterfs.com>
6764 * release version v0_4_1. Hopefully stable on single node use.