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 3.10.0-1160.6.1.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 4.18.0-193.19.1.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 3.10.0-1160.6.1.el7 (RHEL7.9)
26 4.12.14-95.48 (SLES12 SP4)
27 4.12.14-122.20 (SLES12 SP5)
28 4.12.14-197.64 (SLES15 SP1)
29 5.3.18-24.24 (SLES15 SP2)
30 4.15.0-48 (Ubuntu 18.04)
31 * Other clients known to build on these kernels at some point (others may also work):
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 3.10.0-957.27.2.el7 (RHEL7.6)
34 4.14.0-49.13.1.el7a (RHEL7.5)
35 4.14.0-115.2.2.el7a (RHEL7.6)
36 4.18.0-80.11.2.el8 (RHEL8.0)
37 4.18.0-147.8.1.el8 (RHEL8.1)
38 4.18.0-193.19.1.el8 (RHEL8.2)
39 4.18.0-240.1.1.el8 (RHEL8.3)
40 4.4.120-92.70 (SLES12 SP2)
41 4.4.180-94.100 (SLES12 SP3)
42 4.4.0-131 (Ubuntu 16.04)
43 5.4.0-37 (Ubuntu 20.04)
44 5.4.0 (vanilla kernel.org)
45 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
46 * Recommended ZFS version: 2.0.0
47 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
48 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
49 stack overflow. For more information, please refer to bugzilla 17630.
50 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
51 the re-exporting nodes to be mounted with "32bitapi" mount option
53 --------------------------------------------------------------------------------
57 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
58 for currently supported client and server kernel versions.
59 * Server primary kernels built and tested during release cycle:
60 3.10.0-1062.1.1.el7 (RHEL7.7)
61 * Other server kernels known to build and work at some point (others may also work):
62 3.10.0-862.14.4.el7 (RHEL7.5)
63 3.10.0-957.27.2.el7 (RHEL7.6)
64 4.18.0-80.7.1.el8 (RHEL8.0)
65 4.4.120-92.70 (SLES12 SP2)
66 4.4.180-94.100 (SLES12 SP3)
67 4.4.0-131 (Ubuntu 16.04)
68 4.15.0-32 (Ubuntu 18.04)
69 vanilla linux 4.6.7 (ZFS only)
70 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
71 * Client primary unpatched kernels built and tested during release cycle:
72 3.10.0-1062.1.1.el7 (RHEL7.7)
73 4.12.14-95.29 (SLES12 SP4)
74 4.15.0-48 (Ubuntu 18.04)
75 * Other clients known to build on unpatched kernels at some point (others may also work):
76 3.10.0-862.14.4.el7 (RHEL7.5)
77 3.10.0-957.27.2.el7 (RHEL7.6)
78 4.14.0-49.13.1.el7a (RHEL7.5)
79 4.14.0-115.2.2.el7a (RHEL7.6)
80 4.18.0-80.7.1.el8 (RHEL8.0)
81 4.4.120-92.70 (SLES12 SP2)
82 4.4.180-94.100 (SLES12 SP3)
83 4.4.0-131 (Ubuntu 16.04)
84 vanilla linux 4.15.0 (ZFS only)
85 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
86 * Recommended ZFS / SPL version: 0.7.13
87 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
88 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
89 stack overflow. For more information, please refer to bugzilla 17630.
90 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
91 the re-exporting nodes to be mounted with "32bitapi" mount option
92 * Note for Ubuntu 14.04.5:
93 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
94 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
96 --------------------------------------------------------------------------------
100 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
101 for currently supported client and server kernel versions.
102 * Server known to build on the following kernels (others may also work):
103 2.6.32-696.18.7.el6 (RHEL6.9)
104 3.10.0-862.14.4.el7 (RHEL7.5)
105 3.10.0-957.el7 (RHEL7.6)
106 4.4.120-92.70 (SLES12 SP2)
107 4.4.162-94.69 (SLES12 SP3)
108 4.4.0-131 (Ubuntu 16.04)
109 4.15.0-32 (Ubuntu 18.04)
110 vanilla linux 4.6.7 (ZFS only)
111 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
112 * Client known to build on unpatched kernels (others may also work):
113 2.6.32-696.18.7.el6 (RHEL6.9)
114 2.6.32-754.6.3.el6 (RHEL6.10)
115 3.10.0-862.14.4.el7 (RHEL7.5)
116 3.10.0-957.el7 (RHEL7.6)
117 4.14.0-49.13.1.el7a (RHEL7.5)
118 4.14.0-115.2.2.el7a (RHEL7.6)
119 4.4.120-92.70 (SLES12 SP2)
120 4.4.162-94.69 (SLES12 SP3)
121 4.4.0-131 (Ubuntu 16.04)
122 4.15.0-32 (Ubuntu 18.04)
123 vanilla linux 4.15.0 (ZFS only)
124 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
125 * Recommended ZFS / SPL version: 0.7.9
126 * Tested with ZFS / SPL version: 0.7.9
127 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
128 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
129 stack overflow. For more information, please refer to bugzilla 17630.
130 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
131 the re-exporting nodes to be mounted with "32bitapi" mount option
132 * Note for Ubuntu 14.04.5:
133 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
134 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
136 --------------------------------------------------------------------------------
138 2018-04-03 Intel Corporation
140 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
141 for currently supported client and server kernel versions.
142 * Server known to build on patched kernels:
143 2.6.32-431.29.2.el6 (RHEL6.5)
144 2.6.32-504.30.3.el6 (RHEL6.6)
145 2.6.32-573.26.1.el6 (RHEL6.7)
146 2.6.32-642.15.1.el6 (RHEL6.8)
147 2.6.32-696.18.7.el6 (RHEL6.9)
148 3.10.0-693.21.1.el7 (RHEL7.4)
149 3.0.101-0.47.71 (SLES11 SP3)
150 3.0.101-107 (SLES11 SP4)
151 3.12.74-60.64.40 (SLES12 SP1)
152 4.4.114-92.64 (SLES12 SP2)
153 4.4.114-94.11 (SLES12 SP3)
154 3.13.0-101 (Ubuntu 14.04, ZFS only)
155 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
156 4.4.0-116 (Ubuntu 16.04)
157 vanilla linux 4.6.7 (ZFS only)
158 * Client known to build on unpatched kernels:
159 2.6.32-431.29.2.el6 (RHEL6.5)
160 2.6.32-504.30.3.el6 (RHEL6.6)
161 2.6.32-573.26.1.el6 (RHEL6.7)
162 2.6.32-642.15.1.el6 (RHEL6.8)
163 2.6.32-696.18.7.el6 (RHEL6.9)
164 3.10.0-693.21.1.el7 (RHEL7.4)
165 3.0.101-0.47.71 (SLES11 SP3)
166 3.0.101-107 (SLES11 SP4)
167 3.12.74-60.64.40 (SLES12 SP1)
168 4.4.114-92.64 (SLES12 SP2)
169 4.4.114-94.11 (SLES12 SP3)
170 3.13.0-101 (Ubuntu 14.04)
171 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
172 4.4.0-116 (Ubuntu 16.04)
174 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
175 * Recommended ZFS / SPL version: 0.7.6
176 * Tested with ZFS / SPL version: 0.7.6
177 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
178 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
179 stack overflow. For more information, please refer to bugzilla 17630.
180 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
181 the re-exporting nodes to be mounted with "32bitapi" mount option
182 * Note for Ubuntu 14.04.5:
183 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
184 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
186 --------------------------------------------------------------------------------
188 2017-07-05 Intel Corporation
190 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
191 for currently supported client and server kernel versions.
192 * Server known to build on patched kernels:
193 2.6.32-431.29.2.el6 (RHEL6.5)
194 2.6.32-504.30.3.el6 (RHEL6.6)
195 2.6.32-573.26.1.el6 (RHEL6.7)
196 2.6.32-642.15.1.el6 (RHEL6.8)
197 2.6.32-696.3.1.el6 (RHEL6.9)
198 3.10.0-514.21.1.el7 (RHEL7.3)
199 3.0.101-0.47.71 (SLES11 SP3)
200 3.0.101-100 (SLES11 SP4)
201 3.12.74-60.64.40 (SLES12 SP1)
202 4.4.59-92.17 (SLES12 SP2)
203 vanilla linux 4.6.7 (ZFS only)
204 * Client known to build on unpatched kernels:
205 2.6.32-431.29.2.el6 (RHEL6.5)
206 2.6.32-504.30.3.el6 (RHEL6.6)
207 2.6.32-573.26.1.el6 (RHEL6.7)
208 2.6.32-642.15.1.el6 (RHEL6.8)
209 2.6.32-696.3.1.el6 (RHEL6.9)
210 3.10.0-514.21.1.el7 (RHEL7.3)
211 3.0.101-0.47.71 (SLES11 SP3)
212 3.0.101-100 (SLES11 SP4)
213 3.12.74-60.64.40 (SLES12 SP1)
214 4.4.59-92.17 (SLES12 SP2)
216 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
217 * Recommended ZFS / SPL version: 0.6.5.9
218 * Tested with ZFS / SPL version: 0.6.5.9
219 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
220 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
221 stack overflow. For more information, please refer to bugzilla 17630.
222 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
223 the re-exporting nodes to be mounted with "32bitapi" mount option
225 --------------------------------------------------------------------------------
227 12-07-2016 Intel Corporation
229 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
230 for currently supported client and server kernel versions.
231 * Server known to build on patched kernels:
232 2.6.32-431.29.2.el6 (RHEL6.5)
233 2.6.32-504.30.3.el6 (RHEL6.6)
234 2.6.32-573.26.1.el6 (RHEL6.7)
235 2.6.32-642.11.1.el6 (RHEL6.8)
236 3.10.0-514.el7 (RHEL7.3)
237 3.0.101-0.47.71 (SLES11 SP3)
238 3.0.101-80 (SLES11 SP4)
239 3.12.62-60.64.8 (SLES12 SP1)
240 vanilla linux 4.5.7 (ZFS only)
241 * Client known to build on unpatched kernels:
242 2.6.32-431.29.2.el6 (RHEL6.5)
243 2.6.32-504.30.3.el6 (RHEL6.6)
244 2.6.32-573.26.1.el6 (RHEL6.7)
245 2.6.32-642.11.1.el6 (RHEL6.8)
246 3.10.0-514.el7 (RHEL7.3)
247 3.0.101-0.47.71 (SLES11 SP3)
248 3.0.101-80 (SLES11 SP4)
249 3.12.62-60.64.8 (SLES12 SP1)
251 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
252 * Recommended ZFS / SPL version: 0.6.5.8
253 * Tested with ZFS / SPL version: 0.6.5.8
254 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
255 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
256 stack overflow. For more information, please refer to bugzilla 17630.
257 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
258 the re-exporting nodes to be mounted with "32bitapi" mount option
260 --------------------------------------------------------------------------------
262 02-29-2016 Intel Corporation
264 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
265 for currently supported client and server kernel versions.
266 * Server known to build on patched kernels:
267 2.6.32-431.29.2.el6 (RHEL6.5)
268 2.6.32-504.30.3.el6 (RHEL6.6)
269 2.6.32-573.12.1.el6 (RHEL6.7)
270 3.10.0-327.3.1.el7 (RHEL7.2)
271 3.0.101-0.47.71 (SLES11 SP3)
272 3.0.101-68 (SLES11 SP4)
273 vanilla linux 4.2.1 (ZFS only)
274 * Client known to build on unpatched kernels:
275 2.6.32-431.29.2.el6 (RHEL6.5)
276 2.6.32-504.30.3.el6 (RHEL6.6)
277 2.6.32-573.12.1.el6 (RHEL6.7)
278 3.10.0-327.3.1.el7 (RHEL7.2)
279 3.0.101-0.47.71 (SLES11 SP3)
280 3.0.101-68 (SLES11 SP4)
283 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
284 * Recommended ZFS / SPL version: 0.6.4.2
285 * Tested with ZFS / SPL version: 0.6.4.2
286 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
287 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
288 stack overflow. For more information, please refer to bugzilla 17630.
289 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
290 the re-exporting nodes to be mounted with "32bitapi" mount option
292 --------------------------------------------------------------------------------
294 03-10-2015 Intel Corporation
296 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
297 for currently supported client and server kernel versions.
298 * Server known to build on patched kernels:
299 2.6.32-431.29.2.el6 (RHEL6.5)
300 2.6.32-504.8.1.el6 (RHEL6.6)
301 3.0.101-0.46 (SLES11 SP3)
302 * Client known to build on unpatched kernels:
303 2.6.32-431.29.2.el6 (RHEL6.5)
304 2.6.32-504.8.1.el6 (RHEL6.6)
305 3.10.0-123.20.1.el7 (RHEL7)
306 3.0.101-0.46 (SLES11 SP3)
307 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
308 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
309 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
310 stack overflow. For more information, please refer to bugzilla 17630.
311 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
312 the re-exporting nodes to be mounted with "32bitapi" mount option
314 Severity : enhancement
316 Description: control OST-index in IDIF via ROCOMPAT flag.
317 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
318 last_rcvd file. For new formatted OST device, it will be auto set;
319 for the case of upgrading from old OST device, you can enable it
320 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
321 enabled, for new created OST-object, its IDIF-in-LMA will contain
322 the OST-index; for the existing OST-object, the OSD will convert
323 old format IDIF as new format IDIF with OST-index stored in the
324 LMA EA when accessing such OST-object or via OI scrub. Once such
325 flag is enabled, it cannot be reverted back, so the system cannot
326 be downgraded to the orignal incompatible version.
328 --------------------------------------------------------------------------------
330 07-30-2014 Intel Corporation
332 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
333 for currently supported client and server kernel versions.
334 * Server known to build on patched kernels:
335 2.6.32-431.20.3.el6 (RHEL6.5)
336 3.0.101-0.31 (SLES11 SP3)
337 * Client known to build on unpatched kernels:
338 2.6.32-431.20.3.el6 (RHEL6.5)
339 3.0.101-0.31 (SLES11 SP3)
342 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
343 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
344 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
345 stack overflow. For more information, please refer to bugzilla 17630.
346 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
347 the re-exporting nodes to be mounted with "32bitapi" mount option
349 --------------------------------------------------------------------------------
351 10-31-2013 Intel Corporation
353 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
354 * Server builds for kernels:
355 2.6.32-358.18.1.el6 (RHEL6.4)
356 * Client builds for unpatched kernels:
357 2.6.32-358.18.1.el6 (RHEL6.4)
358 3.0.92-0.8 (SLES11 SP3)
360 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
361 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
362 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
363 stack overflow. For more information, please refer to bugzilla 17630.
364 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
365 the re-exporting nodes to be mounted with "32bitapi" mount option
367 --------------------------------------------------------------------------------
369 05-30-2013 Intel Corporation
371 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
372 * Server builds for kernels:
373 2.6.32-358.6.2.el6 (RHEL6.4)
374 * Client builds for unpatched kernels:
375 2.6.32-358.6.2.el6 (RHEL6.4)
376 3.0.74-0.6.6 (SLES11 SP2)
378 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
379 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
380 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
381 stack overflow. For more information, please refer to bugzilla 17630.
382 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
383 the re-exporting nodes to be mounted with "32bitapi" mount option
385 --------------------------------------------------------------------------------
387 9-30-2012 Whamcloud, Inc.
389 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
390 * Server support for kernels:
391 2.6.32-279.5.1.el6 (RHEL6)
392 * Client support for unpatched kernels:
393 2.6.32-279.5.1.el6 (RHEL6)
395 2.6.18-238.19.1.el5 (RHEL5)
396 2.6.16.60-0.69.1 (SLES10)
397 * Recommended e2fsprogs version: 1.41.90.wc4
398 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
399 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
400 stack overflow. For more information, please refer to bugzilla 17630.
401 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
402 the re-exporting nodes to be mounted with "32bitapi" mount option
404 Severity : enhancement
406 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
408 Severity : enhancement
410 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
412 Severity : enhancement
414 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
416 Severity : enhancement
418 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
420 Severity : enhancement
422 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
424 --------------------------------------------------------------------------------
426 04-30-2012 Whamcloud, Inc.
428 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
429 * Server support for kernels:
430 2.6.32-220.4.2.el6 (RHEL6)
431 * Client support for unpatched kernels:
432 2.6.32-220.4.2.el6 (RHEL6)
434 2.6.18-238.19.1.el5 (RHEL5)
435 2.6.16.60-0.69.1 (SLES10)
436 * Recommended e2fsprogs version: 1.41.90.wc4
437 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
438 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
439 stack overflow. For more information, please refer to bugzilla 17630.
440 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
441 the re-exporting nodes to be mounted with "32bitapi" mount option
443 Severity : enhancement
445 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
447 Severity : enhancement
449 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
451 Severity : enhancement
453 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
455 Severity : enhancement
457 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
459 --------------------------------------------------------------------------------
461 09-30-2011 Whamcloud, Inc.
463 * Server support for kernels:
464 2.6.18-238.19.1.el5 (RHEL5)
465 2.6.32-131.6.1.el6 (RHEL6)
466 * Client support for unpatched kernels:
467 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
468 2.6.18-238.19.1.el5 (RHEL5)
469 2.6.32-131.6.1.el6 (RHEL6)
470 2.6.29.4-167.fc11 (FC11)
471 2.6.16.60-0.69.1 (SLES10)
473 * Recommended e2fsprogs version: 1.41.90.wc3
474 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
475 Lustre filesystem with 4K stack may cause a stack overflow. For more
476 information, please refer to bugzilla 17630.
477 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
478 mounted with "32bitapi" mount option
480 Severity : enhancement
481 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
483 Severity : enhancement
484 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
486 Severity : enhancement
487 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
490 Frequency : only with SLES10
492 Description: Use OFED "KMP" provided by Novell
493 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
494 Lustre is now built against this package. That means you need to
495 install the ofed-kmp package from Novell for the patchless client.
496 Note that the ofed-kmp that Novell ships may not exactly match the
497 kernel version but should still be compatible.
499 Severity : enhancement
501 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
502 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
504 --------------------------------------------------------------------------------
506 2010-07-15 Oracle, Inc.
508 * Server support for kernels:
509 2.6.18-164.11.1.el5 (RHEL 5)
510 2.6.18-164.11.1.0.1.el5 (OEL 5)
511 * Client support for unpatched kernels:
512 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
513 2.6.18-164.11.1.el5 (RHEL 5),
514 2.6.18-164.11.1.0.1.el5 (OEL 5)
515 2.6.16.60-0.42.8 (SLES 10),
517 2.6.29.4-167.fc11 (FC11)
518 * Recommended e2fsprogs version: 1.41.10-sun2
520 Severity : enhancement
522 Description: Distribute lfs_migrate script with lustre
523 Details : The lfs_migrate mentioned in the manual is now included in the
524 lustre tree in lustre/scripts and will distributed in lustre RPMs.
526 Severity : enhancement
528 Description: Failover nid list added to import proc file.
530 Severity : enhancement
532 Description: "som_preview" mount option required to use SOM
533 Details : The clients will need to use som_preview mount option
537 Description: Lustre does not do 1MB IOs to HW RAID
538 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
540 Severity : enhancement
542 Description: lfs quota output is not very convenient for awk/sed-parsing
543 Details : Some positions in lfs quota output table could be empty or
544 non-empty which made it hard to parse it with scripts, now
545 a dash is put instead of space where there is not supposed
550 Description: Update to OFED 1.5.1
551 Details : OFED 1.5.1 will be used in environments that don't use
552 vendor supplied OFED.
556 Description: Remove lnet. prefix from lctl params display
557 Details : Also renamed lnet.memused to lnet_memused to
562 Description: Use conf_param -d to remove permanant settings.
566 Description: Add list_param -R
567 Details : lctl list_param now supports recursive mode.
570 Bugzilla : 21829, 22312
571 Description: Fix llobdstat.
572 Details : 1. Now updating the write_bytes counter
573 2. Port script fixes from 1.8.
577 Description: decrease the usage of memory on clients.
578 Details : 1. On clients, recycle dentries and inodes unused.
579 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
584 Description: llobdstat fix and enhancement
585 Details : add a counter to set a limit to how many samples will be returned
586 fix a wildcard in the path to limit to obdfilter stats only
590 Description: lfs check only allowed for root
591 Details : Code cleanups and added a regresion test of non-root use of
596 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
597 Details : Fix hr->hr_index overload.
601 Description: Fix FIEMAP functionality.
605 Description: FC11 patchless client support added
609 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
610 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
614 Description: New acc-sm script and various test-framework enhancments.
618 Description: Previously landed patch was reverted.
622 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
626 Description: Add ldiskfs maxdirsize mount option.
630 Description: Fix mkfs.lustre support for 16TB LUNs.
634 Description: Print mmp_check_interval
635 Details : print mmp_check_interval and make it possible to abort
636 mount operation in case it takes too long.
640 Description: Add cascading_rw to lustre mpi tests.
644 Description: Store SOM data in EA.
645 Details : Also several SOM related cleanups.
649 Description: Add refcount to md_open_data
653 Description: Use refcount on lov_request_set to prevent memory corruption.
657 Description: Statahead: use parent dir's i_mutex for synchronization.
661 Description: loadgen improvements.
665 Description: Enable peer health detection in ksocklnd and ko2iblnd.
669 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
673 Description: Fix race on obd_devcie:md_stats
677 Description: Fix SLES 11 client builds.
678 Details : Resolved uninitalized pointer gcc warnings.
682 Description: Allow non-root access for "lfs check".
683 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
687 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
689 Severity : enhancement
691 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
692 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
696 Description: Sanity test_27p fails.
697 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
698 Sanity tests 27 were cleaned up and improved.
702 Description: Unify req format on client/servers
703 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
704 buffers and swabbers approach..
709 Description: drop join file stuff.
711 Severity : enhancement
713 Description: Add 2.6.27 and SLES11 (patchless client) support
717 Description: Fix in ptlrpc_expire_one_request() to print the signed time
720 Severity : enhancement
722 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
725 Severity : enhancement
727 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
729 Severity : enhancement
731 Description: correctly handle big reply message.
732 Details : send LNet event if reply is bigger than buffer and adjust this buffer
737 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
739 Severity : enhancement
741 Description: Add support for OEL5.
745 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
747 Severity : enhancement
749 Description: "weak-modules" support
750 Details : Implement "weak-modules" support which enables kernel modules
751 to be used with any kernel that implements the same kABI. In
752 order to achieve this modules are now installed in
753 /lib/modules/$(uname -r)/updates/kernel on all distributions.
757 Description: client could not reconnect to OST because of an active request.
758 Details : The client now retries to reconnect to the same server, if a
759 connect request failed with EBUSY or -EAGAIN. Test case 35b
760 conf-sanity.sh has been created.
764 Description: Don't allow make backward step on assignin osc next id.
765 Details : race between allocation next id and ll_sync thread can be cause
766 of set wrong osc next id and can be kill valid ost objects.
770 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
774 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
778 Description: Error handling in osc_statfs_interpret() has been improved.
779 Details : Check in osc_statfs_interpret() for EBADR.
783 Description: Do not send statfs() requests to OSTs disabled by administrator.
784 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
786 Severity : enhancement
788 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
790 Severity : enhancement
792 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
794 Severity : enhancement
796 Description: Enhancement for lfs(1) command to use numeric uid/gid.
800 Description: Improved cleanup procedure for conf-sanity test_46a.
804 Description: Do not start bulk transfer, if client tried to reconnect.
805 Details : Do not start bulk transfer, if client tried to reconnect,
806 but failed due to active rpcs. Check for exp_abort_active_req
807 has been implemented.
811 Description: Prevent inconsistences between linux and lustre mount structures.
812 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
813 Make the sleep interruptible.
815 Severity : enhancement
817 Description: provide server to client comms path
818 Details : server to client communications path via new LDLM_SET_INFO rpc
822 Description: truncate starts GFP_FS allocation under transaction causing deadlock
823 Details : ldiskfs_truncate calls grab_cache_page which may start page
824 allocation under an open transaction. This may lead to
825 calling prune_icache with consequent lustre reentrance.
829 Description: send correctly lsm on open replay
830 Details : MDS is trust to LSM size on replay open, but client can set wrong size
835 Description: enable client interface failover
836 Details : When a child reconnects from another NID, properly update export
837 nid hash position and ldlm reverse import.
839 Severity : enhancement
841 Description: Communicate OST degraded/readonly state via statfs to MDS
842 Details : Flags in the statfs returned from OSTs indicate whether the
843 OST is in a degraded RAID state, or if the filesystem has
844 turned read-only after a filesystem error is detected.
849 Description: don't panic if EPROTO was hit when reading symlink
850 Details : correctly handling request reference in error cases.
852 Severity : enhancement
854 Description: Add LustreNetLink, a kernel-userspace communcation path.
856 Severity : enhancement
858 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
863 Description: access to llog context before init.
864 Details : move handling CATALOGS file at osc layer and forbid access to llog
868 Frequency : rare, only when client is writing to OST on local node
870 Description: Avoid deadlock for local client writes
871 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
872 memory freeing context. This allows OST threads to set the
873 PF_MEMALLOC flag on task structures in order to allocate memory
874 from reserved pools and complete IO.
875 Use GFP_HIGHUSER for OST allocations for non-local client writes,
876 so that the OST threads generate memory pressure and allow
877 inactive pages to be reclaimed.
879 Severity : enhancement
881 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
886 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
887 Details : move ldlm namespace creation in setup phase to avoid grab
888 _lprocfs_lock with cli_sem held.
892 Description: Temporarily disable grant shrink.
893 Details : Disable the feature for debugging.
896 Frequency : only when formatting test filesystems
898 Description: Unable to run several mkfs.lustre on loop devices at the same time
899 Details : mkfs.lustre returns error 256 on the concurrent loop devices
900 formatting. The solution is to proper handle the error.
902 Severity : enhancement
904 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
906 Severity : enhancement
908 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
913 Description: resolve race between obd_disconnect and class_disconnect_exports
914 Details : if obd_disconnect will be called to already disconnected export he
915 forget release one reference and osc module can't unloaded.
917 Severity : enhancement
919 Description: Allow tuning service thread via /proc
920 Details : For each service a new
921 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
922 created that can be used to set min/max thread counts, and get the
923 current number of running threads.
928 Descriptoin: don't return error if have partially created objects for file.
929 Details : lov_update_create_set uses set->set_success as index for created
930 objects, so if some requests will be failed, they will have hole
931 at end of array and qos_shrink_lsm for allocate correct lsm.
933 Severity : enhancement
935 Description: Update OFED support to 1.4.1
940 Description: Reduce small size read RPC
941 Details : Set read-ahead limit for every file and only do read-ahead when
942 available read-ahead pages are bigger than 1M to avoid small size
945 Severity : enhancement
947 Description: move AT tunable parameters for more consistent usage
948 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
950 Severity : enhancement
952 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
953 Details : allow skip disconnected ost for send statfs request and hide error
959 Description: fix racy locking of mballoc block bitmaps causing BUG
960 Details : The locking of the mballoc buddy bitmap and the in-memory
961 block bitmap was using two different spin locks in some
962 cases. This made it possible to incorrectly access the
963 mballoc bitmap while another process was modifying it,
964 causing a sanity assertion to fail. While no on-disk corruption
965 was reported, there was some risk of this happening.
968 Frequency : rare, on llog test 6
970 Descriptoin: don't allow connect to already connected import
971 Details : allowing connect to already connected import is hide connecting problem.
976 Descriptoin: allow kill process which wait statahead result
977 Details : for some reasons 'ls' can stick in waiting result from statahead,
978 in this case need way for kill this process.
980 Severity : enhancement
982 Description: Add state history info file, enhance import info file
983 Details : Track import connection state changes in a new osc/mdc proc file;
984 add overview-type data to the osc/mdc import proc file.
986 Severity : enhancement
988 Description: MDS create should not wait for statfs RPC while holding DLM lock.
990 Severity : enhancement
992 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
997 Descriptoin: OST grant too much space to client even there are not enough space.
998 Details : Client will shrink its grant cache to OST if there are no write
999 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1000 this grant cache if there are already not enough avaible space
1001 (left_space < total_clients * 32M).
1004 Frequency : start MDS on uncleanly shutdowned MDS device
1006 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1007 Details : stay in waiting mds<>ost recovery finished produce random bugs
1008 due race between two ll_sync thread for one lov target. send
1009 ACTIVATE event only if connect realy finished and import have
1013 Frequency : rare, connect and disconnect target at same time
1015 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1016 Details : don't call obd_disconnect under lov_lock. this long time
1017 operation and can block ptlrpcd which answer to connect request.
1022 Descriptoin: don't lose wakeup for imp_recovery_waitq
1023 Details : recover_import_no_retry or invalidate_import and import_close can
1024 both sleep on imp_recovery_waitq, but we was send only one wakeup
1028 Frequency : always with long access acl
1030 Descriptoin: mds can't pack reply with long acl.
1031 Details : mds don't control size of acl but they limited by reint/getattr
1034 Severity : enhancement
1036 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1038 Severity : enhancement
1040 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1043 Frequency : start MDS on uncleanly shutdowned MDS device
1045 Descriptoin: aborting recovery hang on MDS
1046 Details : don't throttle destroy RPCs for the MDT.
1049 Frequency : on remount
1051 Description: external journal device not working after the remount
1052 Details : clear dev_rdonly flag for external journal devices in
1058 Description: shutdown vs evict race
1059 Details : client_disconnect_export vs connect request race.
1060 if client will evicted at this time - we start invalidate
1061 thread without referece to import and import can be freed
1065 Frequency : rare, need acl's on inode.
1067 Description: client can't handle ost additional correctly
1068 Details : if ost was added after client connected to mds client can have
1069 hit lnet_try_match_md ... to big messages to wide striped files.
1070 in this case need teach client to handle config events about add
1071 lov target and update client max ea size at that event.
1073 Severity : enhancement
1075 Description: Changelogs
1076 Details : Changelogs are a lightweight mechanism to track filesystem
1077 metadata and namespace changes. The changelog is recorded
1078 permanently on the MDTs, and is periodically "consumed" / purged
1079 when records are no longer needed.
1081 Severity : enhancement
1083 Description: compact fld format with extents
1084 Details : Store range of seq rather than every seq in FLD. Seq
1085 controller update FLD rather than clients. In Case of CMD, mdt0
1086 has FLD, all other metadata server act as non persistent proxy
1087 for FLD queries and cache fld entries in fld cache.
1092 Description: don't skip ost target if they assigned to file
1093 Details : Drop slow OSCs if we can, but not for requested start idx.
1094 This means "if OSC is slow and it is not the requested
1095 start OST, then it can be skipped, otherwise skip it only
1096 if it is inactive/recovering/out-of-space.
1100 Description: more cleanup in mds_lov
1101 Details : not send LOV EA under replay, we can't know about they size at this
1102 time. Don't allow client connect to mds before any ost connected,
1103 for avoid problems with LOV EA size and returning EIO to client.
1105 Severity : enhancement
1107 Description: Interoperability at server side (Disk interoperability)
1109 Severity : enhancement
1111 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1113 Severity : enhancement
1115 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1117 Severity : enhancement
1119 Description: New client IO stack (CLIO).
1121 Severity : enhancement
1123 Description: Commit on sharing. Eliminate inter-client dependencies between
1124 uncommitted transactions by doing transaction commits.
1125 Thereby clients may recovery independently.
1128 Frequency : Create a symlink file with a very long name
1130 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1131 Details : If there is no extra space in the request for early cancels,
1132 ldlm_req_handles_avail() returns 0 instead of a negative value.
1134 Severity : enhancement
1136 Description: Add /proc entry for import status
1137 Details : The mdc, osc, and mgc import directories now have
1138 an import directory that contains useful import data for debugging
1139 connection problems.
1141 Severity : enhancement
1143 Description: Re-disable certain /proc logging
1144 Details : Enable and disable client's offset_stats, extents_stats and
1145 extents_stats_per_process stats logging on the fly.
1148 Frequency : Only on FC kernels 2.6.22+
1150 Description: oops in statahead
1151 Details : Do not drop reference count for the dentry from VFS when lookup,
1152 VFS will do that by itself.
1154 Severity : enhancement
1156 Description: Generic /proc file permissions
1157 Details : Set /Proc file permissions in a more generic way to enable non-
1158 root users operate on some /proc files.
1162 Description: Hitting mdc_commit_close() ASSERTION
1163 Details : Properly handle request reference release in
1164 ll_release_openhandle().
1168 Description: quota recovery deadlock during mds failover
1169 Details : This patch includes att18982, att18236, att18237 in bz14840.
1171 1. fix osts hang when mds does failover with quotaon
1172 2. prevent watchdog storm when osts threads wait for the
1177 Frequency : only patchless client
1178 Description: add workaround for race between add/remove dentry from hash
1180 Severity : enhancement
1182 Description: Allow OST glimpses to return PW locks
1186 Description: LBUG when llog conf file is full
1187 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1191 Description: Prevent import from entering FULL state when server in recovery
1195 Description: service mount cannot take device name with ":"
1196 Details : Only when device name contains ":/" will mount treat it as
1202 Description: replace ptlrpcd with the statahead thread to interpret the async
1203 statahead RPC callback
1207 Frequency : on recovery
1208 Description: I/O failures after umount during fail back
1209 Details : if client reconnected to restarted server we need join to recovery
1210 instead of find server handler is changed and process self
1211 eviction with cancel all locks.
1213 Severity : enhancement
1215 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1217 Severity : enhancement
1219 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1221 Severity : enhancement
1223 Description: Upcall on Lustre log has been dumped
1224 Details : Allow for a user mode script to be called once a Lustre log has
1225 been dumped. It passes the filename of the dumped log to the
1226 script, the location of the script can be specified via
1227 /proc/sys/lnet/debug_log_upcall.
1232 Description: avoid idr_remove called for id which is not allocated.
1233 Details : Move assigment s_dev for clustered nfs to end of initialization,
1234 to avoid problem with error handling.
1239 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1240 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1241 into CONN_USED_HASH and this prodice warning when put connection
1242 again in unused hash.
1245 Severity : enhancement
1247 Description: File striping can now be set to use an arbitrary pool of OSTs.
1249 Severity : enhancement
1251 Description: Export bytes_read/bytes_write count on OSC/OST.
1255 Description: Early reply size mismatch, MGC loses connection
1256 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1257 the connect flags are properly negotiated.
1262 Description: quotas are not honored with O_DIRECT
1263 Details : all writes with the flag O_DIRECT will use grants which leads to
1264 this problem. Now using OBD_BRW_SYNC to guard this.
1268 Description: add quota statistics
1269 Details : 1. sort out quota proc entries and proc code.
1270 2. add quota statistics
1272 Severity : enhancement
1274 Description: enable quota support for HEAD.
1278 Description: Properly propagate oinfo flags from lov to osc for statfs
1279 Details : restore missing copy oi_flags to lov requests.
1281 Severity : enhancement
1283 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1284 : l_getgroups(8), lst(8), routerstat(8)
1285 Details : included man pages for llobdstat(8), llstat(8),
1286 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1288 Severity : enhancement
1290 Description: configure's --enable-quota should check the
1291 : kernel .config for CONFIG_QUOTA
1292 Details : configure is terminated if --enable-quota is passed but
1293 : no quota support is in kernel
1297 Description: Remove portals compatibility
1298 Details : Remove portals compatibility, not interoperable with releases
1303 Description: Resolve device initialization race
1304 Details : Prevent proc handler from accessing devices added to the
1305 obd_devs array but yet be intialized.
1307 Severity : enhancement
1309 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1311 Severity : enhancement
1313 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1318 Description: Using wrong pointer in osc_brw_prep_request
1319 Details : Access to array[-1] can produce panic if kernel compiled with
1320 CONFIG_PAGE_ALLOC enabled
1324 Description: Client runs out of low memory
1325 Details : Consider only lowmem when counting initial number of llap pages
1329 Description: *optional* service tags registration
1330 Details : if the "service tags" package is installed on a Lustre node
1331 When the filesystem is mounted, a local-node service tag will
1332 be created. See http://inventory.sun.com/ for more information
1333 about the Service Tags asset management system.
1337 Description: Kernel BUG tries to release flock
1338 Details : Lustre does not destroy flock lock before last reference goes
1339 away. So always drop flock locks when client is evicted and
1340 perform unlock regardless of successfulness of speaking to MDS.
1344 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1348 Description: sanity test 65a fails if stripecount of -1 is set
1349 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1354 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1355 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1358 Severity : enhancement
1360 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1364 Description: do not process already freed flock
1365 Details : flock can possibly be freed by another thread before it reaches
1366 to ldlm_flock_completion_ast.
1370 Description: LBUG during stress test
1371 Details : Need properly lock accesses the flock deadlock detection list.
1375 Description: oops in page fault handler
1376 Details : kernel page fault handler can return two special 'pages' in error
1377 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1381 Description: timeout with invalidate import.
1382 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1383 should be handled by ptlrpcd. This produce long age waiting and
1384 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1386 Severity : enhancement
1388 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1392 Description: enable MGS and MDT services start separately
1393 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1394 a co-located MGS without starting the MGS, which is a complement
1395 to 'nosvc' mount option.
1399 Frequency : after recovery
1400 Description: precreate to many object's after del orphan.
1401 Details : del orphan st in oscc last_id == next_id and this triger growing
1402 count of precreated objects. Set flag LOW to skip increase count
1403 of precreated objects.
1407 Frequency : rare, on clear nid stats
1408 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1409 Details : when clean nid stats sometimes try destroy live entry,
1410 and this produce panic in free.
1414 Description: Stack overflow during MDS log replay
1415 ease stack pressure by using a thread dealing llog_process.
1419 Description: wait until IO finished before start new when do lock cancel.
1420 Details : VM protocol want old IO finished before start new, in this case
1421 need wait until PG_writeback is cleared until check dirty flag and
1422 call writepages in lock cancel callback.
1424 Severity : enhancement
1426 Description: using special macro for print time and cleanup in includes.
1430 Description: mds_mfd_close() ASSERTION(rc == 0)
1431 Details : In mds_mfd_close(), we need protect inode's writecount change
1432 within its orphan write semaphore to prevent possible races.
1436 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1440 Frequency : rare, on shutdown ost
1441 Description: don't hit live lock with umount ost.
1442 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1443 use shrink_dcache_sb instead.
1447 Description: don't panic with use echo client
1448 Details : echo client pass NULL as client nid pointer and this produce null
1449 pointer dereference.
1453 Description: fix build on ppc32
1454 Details : compile code with -m64 flag produce wrong object file for ppc32.
1458 Description: add message levels for liblustreapi
1462 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1463 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1464 will result in an -EAGAIN and in the exit call path we call
1465 alloc_rr() which will with increasing aggressiveness attempt to
1466 aquire precreated objects on the minimum number of required OSCs.
1470 Description: Use old size assignment to avoid deadlock
1471 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1472 the scheduling while holding a spinlock. We do not need locking
1473 for size in ll_update_inode() because size is only updated from
1474 the MDS for directories or files without objects, so there is no
1475 other place to do the update, and concurrent access to such inodes
1476 are protected by the inode lock.
1480 Description: resolve "_IOWR redefined" build error on SLES10
1484 Description: dump the memory debugging after all modules are unloaded to
1485 suppress false negative in conf_sanity test 39
1487 Severity : enhancement
1489 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1494 Description: fix SLES kernel versioning
1495 Details : the kernel version for our SLES 10 kernel did not include a "-"
1496 before the "smp" at the end. while this was not a problem in
1497 general, it did mean that software trying to use the kernel
1498 version to try to detect a vendor specific kernel would fail.
1499 this was most evident by the OFED build scripts.
1503 Description: Don't update lov_desc members until making sure they are valid
1504 Details : When updating lov_desc members via proc fs, need fix their
1505 validities before doing the real update.
1509 Description: don't put request into delay list while invalidate in flight.
1510 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1511 import in flight. this produce timeout for invalidate and sometimes
1512 can cause stale data.
1516 Frequency : on ppc only
1517 Description: not convert ost objects for directory because it's not exist.
1518 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1522 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1524 Severity : enhancement
1526 Description: Update to support 2.6.22.14 vanilla kernel.
1530 Frequency : rare, on recovery
1531 Description: read procfs can produce deadlock in some situation
1532 Details : Holding lprocfs lock which send rpc can produce block for destroy
1533 obd objects and this also block reconnect with -EALREADY. This
1534 isn't fix all lprocfs bugs - but make it rare.
1536 Severity : enhancement
1538 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1541 Frequency : frequent on X2 node
1543 Description: mdc_set_open_replay_data LBUG
1544 Details : Set replay data for requests that are eligible for replay.
1548 Description: lustre_mgs: operation 101 on unconnected MGS
1549 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1550 MGC, and late on MGC cannot successfully connect to MGS and a lot
1551 of the error messages complaining that MGS is not connected.
1555 Frequency : on network error
1556 Description: panic with double free request if network error
1557 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1558 but it's true only for synchronus enqueue, for async enqueue
1559 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1562 Severity : enhancement
1564 Description: client-side metadata stat-ahead during readdir(directory readahead)
1565 Details : perform client-side metadata stat-ahead when the client detects
1566 readdir and sequential stat of dir entries therein
1569 Frequency : on start mds
1571 Description: Implement get_info(last_id) in obdfilter.
1574 Frequency : occasional
1576 Description: Correctly check stale fid, not start epoch if ost not support SOM
1577 Details : open with flag O_CREATE need set old fid in op_fid3 because
1578 op_fid2 was overwritten with new generated fid, but mds can answer
1579 with one of these two fids and both is not stale. Setattr
1580 incorrectly started an epoch and assume will be called
1581 done_writing, but without SOM done_writing ever being called.
1584 Frequency : rare, depends on device drivers and load
1586 Description: MDS or OSS nodes crash due to stack overflow
1587 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1588 In some cases, in conjunction with device drivers that use a lot
1589 of stack the MDS (or possibly OSS) service threads could overflow
1590 the stack. One change which was identified to consume additional
1591 stack has been reworked to avoid the extra stack usage.
1594 Frequency : occasional
1596 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1597 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1598 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1599 oscc_has_objects and see OSC has no objests and call
1600 oscc_internal_create to resend create request.
1602 Severity : enhancement
1604 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1606 Severity : enhancement
1608 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1611 Frequency : very rare
1613 Description: Fix replay if there is an un-replied request and open
1614 Details : In some cases, older replay request will revert the
1615 mcd->mcd_last_xid on MDS which is used to record the client's
1616 latest sent request.
1618 Severity : enhancement
1620 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1622 Severity : enhancement
1624 Description: Add rhel5 support to HEAD.
1626 Serverity : enhancement
1628 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1633 Description: Don't allow skipping OSTs if index has been specified.
1634 Details : Don't allow skipping OSTs if index has been specified, make
1635 locking in internal create lots better.
1639 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1640 Details : access to bitfield in structure is always rounded to long
1641 and this produce problem with not atomic change any bit.
1645 Description: Lustre make rpms failed.
1646 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1647 builds Lustre rpms from tarball.
1650 Frequency : rare on shutdown ost
1652 Description: If llog cancel was not send before clean_exports phase, this can
1653 produce deadlock in llog code.
1654 Details : If llog thread has last reference to obd and call class_import_put
1655 this produce deadlock because llog_cleanup_commit_master wait when
1656 last llog_commit_thread exited, but this never success because was
1657 called from llog_commit_thread.
1661 Description: allow userland application know is lost one of stripes.
1662 Details : fill lvb_blocks with error code on ost and return it to
1663 application if error flag found.
1667 Description: NULL lov_tgts causing MDS oops
1668 Details : more safe checks for NULL lov_tgts for avoid oops.
1670 Severity : enhancement
1672 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1676 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1677 Details : If an OST is down the MDS will hang indefinitely in
1678 obd_statfs_rqset() waiting for the statfs data. While for
1679 MDS QOS usage of statfs, it should not stuck in waiting.
1681 Severity : enhancement
1683 Description: remote_acl support
1684 Details : Support ACL-based permission check for remote user.
1685 Support setfacl/getfacl for remote user with the utils
1686 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1687 the system "{s,g}etfacl" utils.
1689 Severity : enhancement
1691 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1693 Severity : enhancement
1695 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1699 Description: make mgs_setparam() handle fsname containing dash
1700 Details : fsname containing a dash does not work with lctl conf_param
1702 Severity : enhancement
1704 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1706 Severity : enhancement
1708 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1712 Frequency : rare, at shutdown
1713 Description: access already free / zero obd_namespace.
1714 Details : if client_disconnect_export was called without force flag set,
1715 and exist connect request in flight, this can produce access to
1716 NULL pointer (or already free pointer) when connect_interpret
1717 store ocd flags in obd_namespace.
1721 Frequency : only at startup
1722 Description: not alloc memory with spinlock held.
1723 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1724 if any spinlock held.
1726 Severity : enhancement
1728 Description: make lustre randomly fail allocating memory
1729 Details : Make lustre randomly failed allocating memory for testing purpose.
1731 Severity : enhancement
1733 Description: lost problems with lov objid file
1734 Details : Fixes some scability and access to not inited memory problems
1735 in work with lov objdid file.
1740 Description: lfs find does not continue on file error
1741 Details : Continue other files processing when a file/dir is absent.
1745 Description: Inconsistent usage of lustre_pack_reply()
1746 Details : Standardize the usage of lustre_pack_reply() such that it
1747 always generate a CERROR on failure.
1750 Frequency : occasional
1752 Description: MDS hang or stay in waiting lock
1753 Details : If client receive lock with CBPENDING flag ldlm need send lock
1754 cancel as separate rpc, to avoid situation when cancel request
1755 can't processed due all i/o threads stay in wait lock.
1759 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1763 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1767 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1771 Description: Update to RHEL4 latest kernel.
1773 Severity : enhancement
1775 Description: Build SLES10 patchless client fails
1776 Details : The configure was broken by run ./configure with
1777 --with-linux-obj=.... argument for patchless client. When the
1778 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1779 can't search header adequately. Use absolute path such as
1780 -I($LINUX)/include instead.
1784 Description: interrupt oig_wait produce painc on resend.
1785 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1786 private set, and this produce situation when rq_ptlrpcd_data not
1787 copyed to new allocated request and triggered LBUG on assert
1788 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1789 ptlrpcd set if request is changed and can be safety changed to use
1794 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1795 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1796 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1797 these macros and replace them with their difinitions instead.
1801 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1802 Details : Update which_patch & target file for SLES10 latest kernel.
1805 Description: Remove client patches from SLES 10 kernel.
1806 Details : This causes SLES 10 clients to behave as patchless clients
1807 even on a Lustre-patched (server) kernel.
1809 Severity : enhancement
1811 Description: self-adjustable client's lru lists
1812 Details : use adaptive algorithm for managing client cached locks lru
1813 lists according to current server load, other client's work
1814 pattern, memory activities, etc. Both, server and client
1815 side namespaces provide number of proc tunables for controlling
1818 Severity : enhancement
1820 Description: light-weight GSS support
1821 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1822 and reduce performance overhead.
1824 Severity : enhancement
1826 Description: Linux keyring support
1827 Details : Support using service of Linux keyring for Lustre GSS internal
1828 context refresh/cache mechanism.
1832 Description: Fix errors in lfs documentation
1833 Details : Fixes man pages
1837 Description: don't use GFP_* in generic Lustre code.
1838 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1842 Description: obdclass is limited by single OBD_ALLOC(idarray)
1843 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1847 Description: warnings with build patchless client with vanila 2.6.19 and up
1848 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1849 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1853 Frequency : Only for SLES
1854 Description: sanity_quota fail test_1
1855 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1856 file, which makes TSTID[2] inunique.
1860 Frequency : Only for SLES9
1861 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1862 Details : Update target/ChangeLog/which_patch .
1866 Frequency : Only for test_10 in sanity-quota.sh
1867 Description: a bug in quota test code
1868 Details : checking a test flag in a wrong place
1872 Frequency : Only for test_13 in sanity-quota.sh
1873 Description: a bug in quota test script
1874 Details : shouldn't check overfull allocated quota
1878 Description: jbd statistics
1879 Details : Port older jbd statistics patch for sles10
1881 Severity : enhancement
1883 Description: Add printing inode info into message about error in writepage.
1888 Description: reply_lock_interpret crash due to race with it and lock cancel.
1889 Details : Do not replay locks that are being cancelled. Do not reference
1890 locks by their address during replay, just by their handle.
1894 Frequency : When flocks are used.
1895 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1896 Details : Flock locks might destroy just granted lock if it could be merged
1897 with another existing flock, this is done in completion handler,
1898 so teach ldlm_cli_enquque_fini that this is a valid case for
1904 Description: Oops in read and write path when failing to allocate lock.
1905 Details : Check if lock allocation failed and return error back.
1909 Description: lstripe command fails for valid OST index
1910 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1911 instead of lov->desc.ld_active_tgt_count.
1915 Description: sanity.sh failed test 103
1916 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1921 Description: df doesn't work properly if diskfs blocksize != 4K
1922 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1926 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1927 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1929 Severity : enhancement
1931 Description: Adaptive timeouts
1932 Details : RPC timeouts adapt to changing server load and network
1933 conditions to reduce resend attempts and improve recovery time.
1937 Description: llapi_file_create() does not allow some changes
1938 Details : add llapi_file_open() that allows specifying the mode and
1939 open flags, and also returns an open file handle.
1943 Description: merge and cleanup kernel patches.
1944 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1948 Description: Add journal checksum support.(Kernel part)
1949 Details : The journal checksum feature adds two new flags i.e
1950 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1951 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1952 indicates that the commit block contains the checksum for
1953 the blocks described by the descriptor blocks. Now commit
1954 record can be sent to disk without waiting for descriptor
1955 blocks to be written to disk. This behavior is controlled
1956 using JBD2_FEATURE_ASYNC_COMMIT flag.
1960 Description: OSS needs mutliple precreate threads
1961 Details : Add ability to start more than one create thread per OSS.
1965 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1966 Details : execute lfs setstripe on client
1970 Description: mds_obd_create error creating tmp object
1971 Details : When the user sets quota on root, llog will be affected and can't
1972 create files and write files.
1975 Frequency : Always on ia64 patchless client, and possibly others.
1977 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1978 Details : This allows the patchless client to be loaded on architectures
1979 without this export.
1983 Description: RedHat Update kernel for RHEL5
1984 Details : Add the kernel config file for RHEL5.
1988 Description: RedHat Update kernel for RHEL5
1989 Details : Modify the kernel config file more closer RHEL5.
1993 Description: Build failure against Centos5 (RHEL5)
1994 Details : Define PAGE_SIZE when it isn't present.
1998 Description: test_7 on sanity-quota.sh can't be used on separate machines
1999 Details : add facet to handle it
2003 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2004 Details : with NFS, the anon dentry's parent was set to itself in
2005 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2006 corresponding dentry other than use rec->ur_name.
2008 Severity : enhancement
2010 Description: lfs setstripe enhancement
2011 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2015 Description: enable data checksumming by default
2016 Details : enable checksum by default, allow --disable-checksum
2017 configure option and "-o nochecksum" mount option. Checksums
2018 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2022 Description: lustre support for RHEL5
2023 Details : Add support for RHEL5.
2027 Description: Client eviction due to failover config
2028 Details : after a connection loss, the lustre client should attempt to
2029 reconnect to the last active server first before trying the
2030 other potential connections.
2034 Description: when mds and osts use different quota unit(32bit and 64bit),
2035 quota will be released repeatly.
2036 Details : void sending multiple quota reqs to mds, which will keep the
2037 status between the reqs.
2041 Description: osts not allocated evenly to files
2042 Details : change the condition to increase offset_idx
2045 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2047 Description: Data corruption for OSTs that are formatted larger than 2TB
2049 Details : When generating the bio request for lustre file writes the
2050 sector number would overflow a temporary variable before being
2051 used for the IO. The data reads correctly from Lustre (which
2052 will overflow in a similar manner) but other file data or
2053 filesystem metadata may be corrupted in some cases.
2057 Description: Tune the kernel for good SCSI performance.
2058 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2059 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2064 Description: rewrite ext2-derived code in obdclass/uuid.c
2065 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2066 from scratch preserving functionality.
2070 Description: "lfs find -obd UUID" prints directories
2071 Details : "lfs find -obd UUID" will return all directory names instead
2072 of just file names. It is incorrect because the directories
2073 do not reside on the OSTs.
2077 Description: use i_size_read and i_size_write in 2.6 port
2078 Details : replace inode->i_size access with i_size_read/write()
2081 Frequency : when removing large files
2083 Description: scheduling issue during removal of large Lustre files
2084 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2085 It causes scheduling issues when removing large files (17TB in the
2089 Frequency : only with liblustre clients on XT3
2091 Description: evictions taking too long
2092 Details : allow llrd to evict clients directly on OSTs
2095 Frequency : only on ppc
2097 Description: /proc/fs/lustre/devices broken on ppc
2098 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2099 We should make sure the type of 'index' is loff_t and then cast
2100 explicitly as needed below. Do not assign an explicitly cast
2104 Frequency : only for rhel5
2106 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2107 Details : Modify the target file & which_kernel.
2109 Severity : enhancement
2111 Description: omit set fsid for export NFS
2112 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2113 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2117 Frequency : Always, for kernels after 2.6.16
2118 Description: Fix warning idr_remove called for id=.. which is not allocated.
2119 Details : Last kernels save old s_dev before kill super and not allow
2120 to restore from callback - restore it before call kill_anon_super.
2124 Description: buffer overruns could theoretically occur
2125 Details : llapi_semantic_traverse() modifies the "path" argument by
2126 appending values to the end of the origin string, and a
2127 overrun may occur. Adding buffer overrun check in liblustreapi.
2131 Description: Fix error on 'ls .' at the top of the Lustre mount.
2132 Details : Don't revalidate dentry if it is a root dentry.
2136 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2137 Details : Modify vm-tunables-rhel4.patch.
2141 Description: Kernel config for 2.6.18-vanilla.
2142 Details : Modify targets/2.6-vanilla.target.in.
2143 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2144 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2145 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2146 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2150 Description: improve handling recoverable errors
2151 Details : if request processig with error which can be recoverable on server
2152 request should be resend, otherwise page released from cache and
2157 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2158 Details : Modify target file & which_patch.
2159 A flaw was found in the IA32 system call emulation provided
2160 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2161 value could be stored in the %RAX register, which could trigger an
2162 out-of-bounds system call table access. An untrusted local user
2163 could exploit this flaw to run code in the kernel
2164 (ie a root privilege escalation). (CVE-2007-4573).
2168 Description: O_DIRECT bypasses client statistics.
2169 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2170 read_ahead_stats were updated. copied the stats section from
2171 osc_send_oap_rpc() into async_internal().
2175 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2179 Description: change order of libsysio includes
2180 Details : '#include sysio.h' should always come before '#include xtio.h'
2184 Description: handle "serious error: objid * already exists" more gracefully
2185 Details : If LAST_ID value on disk is smaller than the objects existing in
2186 the O/0/d* directories, it indicates disk corruption and causes an
2187 LBUG(). If the object is 0-length, then we should use the existing
2188 object. This will help to avoid a full fsck in most cases.
2190 Severity : enhancement
2192 Description: adapt the lustre_config script to support the upgrade case
2193 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2194 server targets to 1.6 in parallel.
2198 Description: To avoid grant space > avaible space when the disk is almost
2199 full. Without this patch you might see the error "grant XXXX >
2200 available" or some LBUG about grant, when the disk is almost
2202 Details : In filter_check_grant, for non_grant cache write, we should
2203 check the left space by if (*left > ungranted + bytes), instead
2204 of (*left > ungranted), because only we are sure the left space
2205 is enough for another "bytes", then the ungrant space should be
2206 increase. In client, we should update cl_avail_grant only there
2207 is OBD_MD_FLGRANT in the reply.
2211 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2212 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2213 described in CVE-2007-4573. This problem could allow untrusted
2214 local users to gain root access.
2217 Frequency : when using O_DIRECT and quotas
2219 Description: Incorrect file ownership on O_DIRECT output files
2220 Details : block usage reported by 'lfs quota' does not take into account
2221 files that have been written with O_DIRECT.
2226 Description: touch file failed when fs is not full
2227 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2228 otherwise we can get ENOSP while fs is not full.
2232 Description: parallel lock callbacks
2233 Details : Instead of sending blocking and completion callbacks as separated
2234 requests, adding them to a set and sending in parallel.
2237 Frequency : only for Cray XT3
2238 Bugzilla : 12829/13455
2239 Description: Changing primary group doesn't change the group lustre assigns to
2241 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2242 overridden with the primary group provided by the group upcall,
2243 whereas the supplied fsgid can be trusted if it is in the list of
2244 supplementary groups returned by the group upcall.
2246 Severity : enhancement
2248 Description: Allow masking D_WARNING, D_ERROR messages from console
2249 Details : Console messages can now be disabled via lnet.printk.
2253 Description: User code with malformed file open parameter crashes client node
2254 Details : Before packing join_file req, all the related reference should be
2255 checked carefully in case some malformed flags cause fake
2256 join_file req on client.
2260 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2261 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2262 mode is equal to the granted mode. However, it should also check
2263 that LDLM_FL_CP_REQD is not set.
2267 Description: Heavy nfs access might result in deadlocks
2268 Details : After ELC code landed, it is now improper to enqueue any mds
2269 locks under och_sem, because enqueue might want to decide to
2270 cancel open locks for same inode we are holding och_sem for.
2274 Description: Client eviction while running blogbench
2275 Details : A lot of unlink operations with concurrent I/O can lead to a
2276 deadlock causing evictions. To address the problem, the number of
2277 oustanding OST_DESTROY requests is now throttled to
2278 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2279 callbacks are processed in priority.
2283 Description: enable ACLs on MDS by default
2284 Details : ACLs must be enabled on MDS by default.
2287 Frequency : PPC/PPC64 only
2289 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2290 Details : fix duplicated definitions between asm-ppc64/types.h and
2291 lustre_types.h on PPC.
2294 Frequency : PPC/PPC64 only
2296 Description: asm-ppc/segment.h does not exist
2297 Details : fix compile issue on PPC.
2301 Description: better handle error messages in extents code
2304 Frequency : RHEL4 only
2306 Description: mkfs is very slow on IA64/RHEL4
2307 Details : A performance regression has been discovered in the MPT Fusion
2308 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2309 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2310 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2312 Severity : enhancement
2314 Description: SNMP support enhancement
2315 Details : Adding total number of sampled request for an MDS node in snmp
2318 Severity : enhancement
2320 Description: Optimize ldlm waiting list processing for PR extent locks
2321 Details : When processing waiting list for read extent lock and meeting read
2322 lock that is same or wider to it that is not contended, skip
2323 processing rest of the list and immediatelly return current
2324 status of conflictness, since we are guaranteed there are no
2325 conflicting locks in the rest of the list.
2329 Description: Time out and refuse to reconnect
2330 Details : When the failover node is the primary node, it is possible
2331 to have two identical connections in imp_conn_list. We must
2332 compare not conn's pointers but NIDs, otherwise we can defeat
2333 connection throttling.
2337 Description: port llog fixes from b1_6 into HEAD
2338 Details : Port llog reference couting and some llog cleanups from b1_6
2339 (bug 10800) into HEAD, for protect from panic and access to already
2340 free llog structures.
2344 Description: Detect stride IO mode in read-ahead
2345 Details : When a client does stride read, read-ahead should detect that and
2346 read-ahead pages according to the detected stride pattern.
2350 Description: data checksumming impacts single node performance
2351 Details : add support for several checksum algorithm. Currently, only CRC32
2352 and Adler-32 are supported. The checksum type can be changed on
2353 the fly via /proc/fs/lustre/osc/*/checksum_type.
2357 Description: use adler32 for page checksums
2358 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2363 Description: build for x2 fails
2364 Details : fix compile issue on Cray systems.
2368 Description: Properly match for duplicate locks
2369 Details : Due to different lock order from skiplists code, we need to
2370 traverse entire list for now
2373 Frequency : only on PPC/SLES10
2375 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2376 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2377 be defined. Add a hack in mkfs_lustre.c to work around this compile
2382 Description: LASSERT on MDS when client holding flock lock dies
2383 Details : ldlm pool logic depends on number of granted locks equal to
2384 number of released locks which is not true for flock locks, so
2385 just exclude such locks from consideration.
2389 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2390 Details : Use fsfilt_sync() for both the whole filesystem sync and
2391 individual file sync to eliminate dangerous inode locking
2392 with I_LOCK that can lead to a deadlock.
2396 Description: performance in 1.6.3
2397 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2398 sure that 1MB requests are not fragmented by the block layer.
2400 Severity : enhancement
2402 Description: organize the server-side client stats on per-nid basis
2403 Details : Change the structure of stats under obdfilter and mds to
2411 The "uuid"s file would list the uuids of _active_ exports.
2412 And the clear entry is to clear all stats and stale nids.
2414 Severity : enhancement
2416 Description: eliminate client locks in face of contention
2417 Details : file contention detection and lockless i/o implementation
2418 for contended files.
2422 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2425 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2427 Description: MDS may lose file striping (and hence file data) in some cases
2428 Details : If there are additional extended attributes stored on the MDS,
2429 in particular ACLs, SELinux, or user attributes (if user_xattr
2430 is specified for the client mount options) then there is a risk
2431 of attribute loss. Additionally, the Lustre file striping
2432 needs to be larger than default (e.g. striped over all OSTs),
2433 and an additional attribute must be stored initially in the
2434 inode and then increase in size enough to be moved to the
2435 external attribute block (e.g. ACL growing in size) for file
2440 Description: skiplist implementation simplification
2441 Details : skiplists are used to group compatible locks on granted list
2442 that was implemented as tracking first and last lock of each
2443 lock group the patch changes that to using doubly linked lists
2447 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2448 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2449 because LASSERT on both the data supplied by a client, and the
2450 data on disk is dangerous and incorrect.
2452 Severity : enhancement
2454 Description: Slow truncate/writes to huge files at high offsets.
2455 Details : Directly associate cached pages to lock that protect those pages,
2456 this allows us to quickly find what pages to write and remove
2457 once lock callback is received.
2461 Description: more ldlm soft lockups
2462 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2463 starve other threads from the resource lock for a long time in
2464 case of long waiting queue, so change the debug level from
2465 D_OTHER to the less frequently used D_INFO.
2467 Severity : enhancement
2469 Description: add -gid, -group, -uid, -user options to lfs find
2473 Description: Hung threads in invalidate_inode_pages2_range
2474 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2475 once one is completed. As a result, some RPCs are stuck in the
2476 queue and are never sent.
2480 Description: filter threads hungs on waiting journal commit
2481 Details : Cleanup filter group llog code, then only filter group llog will
2482 be only created in the MDS/OST syncing process.
2486 Description: Procfs and llog threads access destoryed import sometimes.
2487 Details : Sync the import destoryed process with procfs and llog threads by
2488 the import refcount and semaphore.
2490 Severity : enhancement
2492 Description: openlock cache of b1_6 port to HEAD
2497 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2498 Details : If insertion of an extent fails, then discard the inode
2499 preallocation and free data blocks else it can lead to duplicate
2504 Description: don't always update ctime in ext3_xattr_set_handle()
2505 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2506 In some cases the ctime should not be updated, for example for
2507 2.0->1.8 compatibility it is necessary to delete an xattr and it
2508 should not update the ctime.
2512 Bugzilla : 15713/16362
2513 Description: Assertion in iopen_connect_dentry in 1.6.3
2514 Details : looking up an inode via iopen with the wrong generation number can
2515 populate the dcache with a disconneced dentry while the inode
2516 number is in the process of being reallocated. This causes an
2517 assertion failure in iopen since the inode's dentry list contains
2518 both a connected and disconnected dentry.
2522 Description: assertion failure in ldlm_handle2lock()
2523 Details : fix a race between class_handle_unhash() and class_handle2object()
2524 introduced in lustre 1.6.5 by bug 13622.
2529 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2530 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2535 Description: oops in sd_iostats_seq_show()
2536 Details : unloading/reloading the scsi low level driver triggers a kernel
2537 bug when trying to access the sd iostat file.
2542 Description: Kernel panics during QLogic driver reload
2543 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2544 patch, causing memory corruption.
2549 Description: journal_dev option does not work in b1_6
2550 Details : pass mount option during pre-mount.
2552 Severity : enhancement
2554 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2555 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2556 extent information of a file. It can be used to map logical blocks
2557 in a file to physical blocks in the block device.
2561 Description: LDLM soft lockups - improvement
2562 Details : It is be possible to send the lock handle along with each read
2563 or write request because the client is already doing a lock match
2564 itself so there isn't any reason the OST should have to re-do that
2568 Frequency : only on Cray X2
2570 Description: X2 build failures
2571 Details : fix build failures on Cray X2.
2575 Description: xid & resent requests
2576 Details : Initialize RPC XID from clock at startup (randomly if clock is
2579 Severity : enhancement
2581 Description: Add lustre_start utility to start or stop multiple Lustre servers
2586 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2587 Details : In case of memory pressure, list_del() can be called twice on
2588 req->rq_history_list, causing a kernel oops.
2592 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2593 Details : fix stack overflow in the distributed lock manager by defering
2594 export eviction after a failed AST to the elt thread instead of
2595 handling it in the dlm interpret routine.
2599 Description: Convert some comments to new format.
2600 Details : Update documenting comments to match doxygen conventions.
2604 Description: Grammar fixes.
2605 Details : A couple of trivial spelling fixes.
2609 Description: OSD_COUNTERS-mandatory
2610 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2611 to simplify the code.
2615 Description: simplify cmm_device freeing logic.
2616 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2620 Description: Add lockdep support to dt_object_operations locking interface.
2621 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2622 parameter indicating lock ordering. Update mdd code to use new
2627 Description: Introduce failloc constants for lockless IO tests.
2628 Details : Add two new failloc constants to test lockless IO. Only one of
2629 them in implemented---another is checked in yet to be landed
2634 Description: Add lockdep support for inode mutex.
2635 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2636 in the situations where Lustre has to lock more than one inode
2641 Description: Add optional invariants checking support.
2642 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2643 configure switch. This macro is to be used for consistency and
2644 sanity checks that are too expensive to be left in `production'
2649 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2650 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2655 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2656 Details : Use spin_lock_nested() in (the only) situation where more than
2657 one ldlm_lock is locked simultaneously. Also, fix possible
2658 dead-lock in ldlm_lock_change_resource() by enforcing particular
2663 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2664 Details : Instead of specifying each ldlm_lock call-back through separate
2665 parameter, wrap them into struct ldlm_callback_suite.
2669 Description: Kill join_lru obd method and its callers.
2670 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2671 in memory---a requirement implemented through ->o_join_lru() obd
2672 method in HEAD. Remove this method and its users.
2676 Description: Add asynchronous ldlm ENQUEUE completion handler.
2677 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2678 case that stock ldlm_completion_ast() cannot handle as it waits
2679 until lock is granted. Introduce new ldlm_completion_ast_async()
2680 for this. Also comment ldlm_completion_ast().
2684 Description: ldlm_error <-> errno conversion.
2685 Details : Add functions to map (rather arbitrary) between LDLM error codes
2686 and standard errno values. CLIO needs this to prevent LDLM specific
2687 constants from escaping ldlm and osc.
2691 Description: Kill unused ldlm_handle2lock_ns() function.
2692 Details : Kill unused ldlm_handle2lock_ns() function.
2696 Description: Add lu_ref support to ldlm_lock
2697 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2698 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2699 and ldlm_resource. LDLM interface has to be changed, because code that
2700 releases a reference on a lock, has to "know" what reference this is.
2701 In the most frequent case
2703 lock = ldlm_handle2lock(handle);
2705 LDLM_LOCK_PUT(lock);
2707 no changes are required. When any other reference (received _not_ from
2708 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2709 instead of LDLM_LOCK_PUT().
2711 Arguably, changes are pervasive, and interface requires some discipline
2712 for proper use. On the other hand, it was very instrumental in finding
2713 a few leaked lock references.
2717 Description: Add ldlm_lock_addref_try().
2718 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2719 attempts to addref a lock that might be being canceled
2724 Description: Add ldlm_weigh_callback().
2725 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2726 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2727 hard-coded `number of pages' logic.
2731 Description: Add lockdep annotations to llog code.
2732 Details : Use appropriately tagged _nested() locking calls in the places
2733 where llog takes more than one ->lgh_lock lock.
2737 Description: Add loi_kms_set().
2738 Details : Wrap kms updates into a helper function.
2742 Description: Constify instances of struct lsm_operations.
2743 Details : Constify instances of struct lsm_operations.
2747 Description: lu_conf support.
2748 Details : On a server, a file system object is uniquely identified
2749 by a fid, which is sufficient to locate and load all object
2750 state (inode). On a client, on the other hand, more data are
2751 necessary instantiate an object. Change lu_object_find() and
2752 friends to take additional `lu_conf' argument describing object.
2753 Typically this includes layout information.
2757 Description: lu_context fixes.
2758 Details : Introduce new lu_context functions that are needed on the client
2759 side, where some system threads (ptlrpcd) are shared by multiple
2760 modules, and so cannot be stopped during module shutdown.
2764 Description: Add start and stop methods to lu_device_type_operations.
2765 Details : Introduce two new methods in lu_device_type_operations, that are
2766 invoked when first instance of a given type is created and last one
2767 is destroyed respectively. This is need by CLIO.
2771 Description: Add lu_ref support to struct lu_device.
2772 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2773 track leaked references.
2777 Description: Introduce lu_kmem_descr.
2778 Details : lu_kmem_descr and its companion interface allow to create
2779 and destroy a number of kmem caches at once.
2783 Description: Fix lu_object finalization race.
2784 Details : Fix a race between lu_object_find() finding an object and its
2785 concurrent finalization. This race is (most likely) not possible
2786 on the server, but might happen on the client.
2790 Description: Introduce lu_ref interface.
2791 Details : lu_ref is a debugging module allowing to track references to
2792 a given object. It is quite cpu expensive, and has to be
2793 explicitly enabled with --enable-lu_ref. See usage description
2798 Description: Factor lu_site procfs stats into a separate function.
2799 Details : Separate lu_site stats printing code into a separate function
2800 that can be reused on a client.
2804 Description: Constify instances of struct {lu,dt,md}_device_operations.
2805 Details : Constify instances of struct {lu,dt,md}_device_operations.
2809 Description: Introduce struct md_site and move meta-data specific parts of
2810 struct lu_site here.
2811 Details : Move md-specific fields out of struct lu_site into special struct
2812 md_site, so that lu_site can be used on a client.
2816 Description: Kill mdd_lov_destroy().
2817 Details : Remove unused mdd code.
2821 Description: Add st_block checking to multistat.c.
2822 Details : Add st_block checking to multistat.c.
2826 Description: Add lu_ref support to struct obd_device.
2827 Details : Add lu_ref tracking to obd_device.
2831 Description: Kill obd_set_fail_loc().
2832 Details : Remove unused code.
2836 Description: Add special type for ptlrpc_request interpret functions.
2837 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2838 there. Actual usage will be in CLIO.
2842 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2843 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2844 doesn't work with lockdep.
2848 Description: Add rwv.c test program.
2849 Details : New testing program exercising readv(2) and writev(2) (Qian).
2853 Description: Add sendfile.c test program.
2854 Details : New testing program exercising sendfile(2) (Jay).
2858 Description: Ratelimit a message that can be very frequent.
2859 Details : Ratelimit a memory allocation failure message that can
2864 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2865 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2869 Description: Liblustre build fixes.
2870 Details : Liblustre build fixes.
2874 Description: libcfs: add cfs_{need,cond}_resched() interface.
2875 Details : libcfs: add cfs_{need,cond}_resched() definition and
2876 implementations for Linux, NT, and liblustre.
2878 Severity : enhancement
2880 Description: More exported tunables for mballoc
2881 Details : Add support for tunable preallocation window and new tunables for
2882 large/small requests
2886 Description: Detect corruption of block bitmap and checking for preallocations
2887 Details : Checks validity of on-disk block bitmap. Also it does better
2888 checking of number of applied preallocations. When corruption is
2889 found, it turns filesystem readonly to prevent further corruptions.
2893 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2894 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2895 window to 0. If it does hit cache miss, and read-pattern is still
2896 stride-io mode, does not reset the stride window, but also does not
2897 increase the stride window length in this case.
2901 Frequency : only for big-endian servers
2902 Description: Check if system is big-endian while mounting fs with extents feature
2903 Details : Mounting a filesystem with extents feature will fail on big-endian
2904 systems since ext3-based ldiskfs is not supported on big-endian
2905 systems. This can be over-riden with "bigendian_extents" mount option.
2907 Severity : enhancement
2909 Description: The root squash functionality
2910 Details : A security feature, which is to prevent users from being able
2911 to mount lustre on their desktop, run as root, and delete
2912 all of the files in the filesystem. The goal is accomplished by
2913 remapping user id (UID) and group id (GID) of the root user to
2914 a UID and GID specified by the system administartor via Lustre
2915 configuration management server (MGS). The functionality also
2916 allows to specify sets of clients for which the remapping does
2921 Description: Excessive recovery window
2922 Details : With AT enabled, the recovery window can be excessively long (6000+
2923 seconds). To address this problem, we no longer use
2924 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2925 timeout no longer depends on the service time, it is set to
2926 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2927 time via pb_service_time.
2931 Description: Watchdog triggered on MDS failover
2932 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2933 the OSTs know that the MDS "UUID" can be reused for the same export
2934 from a different NID, so we do not need to wait for the export to be
2938 Frequency : rare, only if using MMP with Linux RAID
2940 Description: MMP doesn't work with Linux RAID
2941 Details : While using HA for Lustre servers with Linux RAID, it is possible
2942 that MMP will not detect multiple mounts. To make this work we
2943 need to unplug the device queue in RAID when the MMP block is being
2944 written. Also while reading the MMP block, we should read it from
2945 disk and not the cached one.
2947 Severity : enhancement
2949 Description: open file using fid
2950 Details : A file can be opened using just its fid, like
2951 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2954 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2957 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2958 Details : add proper locking for creation of jbd_4k slab cache
2962 Description: MMP check in ext3_remount() fails without displaying any error
2963 Details : When multiple mount protection fails during remount, proper error
2966 Severity : enhancement
2968 Description: Allow stripe size to be up to 4G-64k
2969 Details : Fix math logic to allow large stripe sizes.
2973 Description: add check for >8TB ldiskfs filesystems
2974 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2975 Don't allow >8TB ldiskfs filesystems to be mounted without
2976 force_over_8tb mount option
2981 Descriptoin: OSS complains "too many blocks" and crashes
2982 Details : In filter_commitrw_write():
2983 1. lquota_chkquota();
2984 2. obdo_from_inode();
2985 3. lquota_pending_commit();
2986 "2." has changed uid and gid in this case, then caused
2987 lquota_chkquota() and lquota_pending_commit() to use
2988 different uid and gid. That is the root of the bug.
2992 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2993 Details : Detect and print out the 'stripe count = -1' case properly.
2997 Description: Out or order replies might be lost on replay
2998 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2999 request with tid smaller than one currently being added, add it
3000 to the start, not end of the list.
3002 --------------------------------------------------------------------------------
3004 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3006 * Support for kernels:
3007 2.6.5-7.283 (SLES 9),
3008 2.6.9-55.EL (RHEL 4),
3009 2.6.16.46-0.14 (SLES 10),
3010 2.6.18.8 vanilla (kernel.org)
3011 * Client support for unpatched kernels:
3012 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3013 2.6.16 - 2.6.22 vanilla (kernel.org)
3014 * Due to recently discovered recovery problems, we do not recommend
3015 using patchless RHEL 4 clients with this or any earlier release.
3016 * Recommended e2fsprogs version: 1.39.cfs8
3017 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3018 * Starting with this release, the ldiskfs backing filesystem required
3019 by Lustre is now in its own package, lustre-ldiskfs. This package
3020 should be installed. It is versioned separately from Lustre and
3021 may be released separately in future.
3025 Description: block reactivating mgc import until all deactivates complete
3026 Details : Fix race when failing back MDT/MGS to itself (testing)
3029 Frequency : at statup only
3031 Description: mds_lov_synchronize race leads to various problems
3032 Details : simultaneous MDT->OST connections at startup can cause the
3033 sync to abort, leaving the OSC in a bad state.
3035 Severity : enhancement
3037 Description: add optional extra BUILD_VERSION info
3038 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3039 to override the lustre version.
3042 Frequency : 2.6.18 servers only
3044 Description: ll_kern_mount() doesn't release the module reference
3045 Details : The ldiskfs module reference count never drops down to 0
3046 because ll_kern_mount() doesn't release the module reference.
3051 Description: server LBUG when using old ost_num_threads parameter
3052 Details : Accept the old ost_num_threads parameter but warn that it
3053 is deprecated, and fix an off-by-one error that caused an LBUG.
3058 Description: Transient SCSI error results in persistent IO issue
3059 Details : iobuf->dr_error is not reinitialized to 0 between two
3063 Frequency : sometimes when underlying device returns I/O errors
3065 Description: OSTs not going read-only during write failures
3066 Details : OSTs are not remounted read-only when the journal commit threads
3067 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3068 instead of the ext3 wrappers.
3072 Description: poor connect scaling with increasing client count
3073 Details : Don't run filter_grant_sanity_check for more than 100 exports
3074 to improve scaling for large numbers of clients.
3077 Frequency : SLES10 only
3079 Description: sanity-quota.sh quotacheck failed: rc = -22
3080 Details : Quotas cannot be enabled on SLES10.
3083 Frequency : liblustre clients only
3085 Description: getdirentries does not give error when run on compute nodes
3086 Details : getdirentries does not fail when the size specified as an argument
3087 is too small to contain at least one entry
3089 Severity : enhancement
3091 Description: Add LNET router traceability for debug purposes
3092 Details : If a checksum failure occurs with a router as part of the
3093 IO path, the NID of the last router that forwarded the bulk data
3094 is printed so it can be identified.
3099 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3100 Details : Due to a race condition, liblustre clients were occasionally
3101 evicted incorrectly.
3103 Severity : enhancement
3105 Description: lfs setstripe use optional parameters instead of postional
3108 Severity : enhancement
3110 Description: Nanosecond timestamp support for ldiskfs
3111 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3112 resolution timestamps. There is not yet support for this at
3113 the Lustre filesystem level.
3116 Frequency : during server recovery
3118 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3119 Details : request with rq_no_resend flag not awake l_wait_event if they get
3123 Frequency : nfs export on patchless client
3125 Description: connectathon hang when test nfs export over patchless client
3126 Details : Disconnected dentry cannot be found with lookup, so we do not need
3127 to unhash it or make it invalid
3130 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3131 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3132 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3137 Description: open req refcounting wrong on reconnect
3138 Details : If reconnect happened between getting open reply from server and
3139 call to mdc_set_replay_data in ll_file_open, we will schedule
3140 replay for unreferenced request that we are about to free.
3141 Subsequent close will crash in variety of ways.
3142 Check that request is still eligible for replay in
3143 mdc_set_replay_data().
3148 Description: disable writes to filesystem when reading health_check file
3149 Details : the default for reading the health_check proc file has changed
3150 to NOT do a journal transaction and write to disk, because this
3151 can cause reads of the /proc file to hang and block HA state
3152 checking on a healthy but otherwise heavily loaded system. It
3153 is possible to return to the previous behaviour during configure
3154 with --enable-health-write.
3156 Severity : enhancement
3158 Description: 64-bit inode version
3159 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3160 to the inode. This will be required for version-based recovery.
3165 Description: MDS fails to start if a duplicate client export is detected
3166 Details : in some rare cases it was possible for a client to connect to
3167 an MDS multiple times. Upon recovery the MDS would detect this
3168 and fail during startup. Handle this more gracefully.
3170 Severity : enhancement
3172 Description: Add -o localflock option to simulate old noflock
3174 Details : This will achieve local-only flock/fcntl locks
3180 Description: log_commit_thread vs filter_destroy race leads to crash
3181 Details : Take import reference before releasing llog record semaphore
3186 Description: Wrong request locking in request set processing
3187 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3188 imp_delayed_list, in this place should be used imp_lock.
3191 Frequency : when reconnection
3193 Description: Grant Leak when osc reconnect to OST
3194 Details : When osc reconnect ost, OST(filter) should check whether it
3195 should grant more space to client by comparing fed_grant and
3196 cl_avail_grant, and return the granted space to client instead
3197 of "new granted" space, because client will call osc_init_grant
3198 to update the client grant space info.
3201 Frequency : when client reconnect to OST
3203 Description: Grant Leak when osc do resend and replay bulk write
3204 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3205 bulk write request, because the grant info while be sync between
3206 OSC and OST when reconnect, and we should ignore the grant info
3207 these of resend/replay write req.
3212 Description: Grant space more than available left space sometimes.
3213 Details : When then OST is about to be full, if two bulk writing from
3214 different clients came to OST. Accord the avaliable space of the
3215 OST, the first req should be permitted, and the second one
3216 should be denied by ENOSPC. But if the seconde arrived before
3217 the first one is committed. The OST might wrongly permit second
3218 writing, which will cause grant space > available space.
3221 Frequency : when client is evicted
3223 Description: Grant might be wrongly erased when osc is evicted by OST
3224 Details : when the import is evicted by server, it will fork another
3225 thread ptlrpc_invalidate_import_thread to invalidate the
3226 import, where the grant will be set to 0. While the original
3227 thread will update the grant it got when connecting. So if
3228 the former happened latter, the grant will be wrongly errased
3229 because of this race.
3234 Description: Checking Stale with correct fid
3235 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3236 to check whether it is stale, because sometimes, we want the
3237 enqueue happened anyway, and op_data.fid2 will not be initialized.
3239 Severity : enhancement
3241 Description: update patchless client
3242 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3245 Frequency : only with 2.4 kernel
3247 Description: random memory corruption
3248 Details : size of struct ll_inode_info is to big for union inode.u and this
3249 can be cause of random memory corruption.
3254 Description: Memory leak in recovery
3255 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3256 It should also check obd_fail before fsfilt_start, otherwise if
3257 fsfilt_start return -EROFS,(failover mds during mds recovery).
3258 then the req will return with repmsg->transno = 0 and rc = EROFS.
3259 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3260 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3261 be freed no matter whether fsfilt_commit success or not.
3264 Frequency : only with huge count clients
3266 Description: Prevents from taking the superblock lock in llap_from_page for
3268 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3269 for llap_from_page call in ll_removepage prevents from taking the
3270 superblock lock for a soon died page.
3275 Description: Not check open intent error before release open handle
3276 Details : in some rare cases, the open intent error is not checked before
3277 release open handle, which may cause
3278 ASSERTION(open_req->rq_transno != 0), because it tries to release
3279 the failed open handle.
3284 Description: Set cat log bitmap only after create log success.
3285 Details : in some rare cases, the cat log bitmap is set too early. and it
3286 should be set only after create log success.
3290 Description: Accessing a block bevice can re-enable I/O when Lustre is
3291 tearing down a device.
3292 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3297 Description: service threads may hog cpus when there are a lot of requests
3298 Details : Insert cond_resched to give other threads a chance to use some CPU
3303 Description: the cat log was not initialized in recovery
3304 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3305 unlink log on mds will not be initialized until mds post
3306 recovery. And also in mds post recovery, the unlink log will
3307 initialization will be done asynchronausly, so there will be race
3308 between add unlink log and unlink log initialization.
3312 Description: brw_stats were being printed incorrectly
3313 Details : brw_stats were being printed as log2 but all of them were not
3314 recorded as log2. Also remove some code duplication arising from
3315 filter_tally_{read,write}.
3319 Frequency : rare, only in recovery.
3320 Description: ASSERTION(req->rq_type != LI_POISON) failed
3321 Details : imp_lock should be held while iterating over imp_sending_list for
3322 prevent destroy request after get timeout in ptlrpc_queue_wait.
3326 Description: replay-single.sh test 52 fails
3327 Details : A lock's skiplist need to be cleanup when it being unlinked
3328 from its resource list.
3332 Description: Short directio read returns full requested size rather than
3334 Details : Direct I/O operations should return actual amount of bytes
3335 transferred rather than requested size.
3337 Severity : enhancement
3339 Description: metadata RPC reduction (e.g. for rm performance)
3340 Details : decrease the amount of synchronous RPC between clients and servers
3341 by canceling conflicing lock before the operation on the client
3342 and packing thier handles into the main operation RPC to server.
3344 Severity : enhancement
3346 Description: Async OSC create to avoid the blocking unnecessarily.
3347 Details : If an OST has no remaining object, system will block on the
3348 creation when it needs to create a new object on this OST. Now,
3349 ways use pre-created objects when available, instead of blocking on
3350 an empty osc while others are not empty. If we must block, we
3351 block for the shortest possible period of time.
3355 Description: improve handling recoverable errors
3356 Details : if request processig with error which can be recoverable on server
3357 request should be resend, otherwise page released from cache and
3360 Severity : enhancement
3362 Description: refine locking for avoid write wrong info into lov_objid file
3363 Details : fix possible races with add new target and write/update data in
3366 --------------------------------------------------------------------------------
3368 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3373 Frequency : on some architectures
3375 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3376 Details : Uninitialized flags sometimes cause configuration commands to
3380 Frequency : patchless clients only
3382 Description: missing __iget() symbol export
3383 Details : The __iget() symbol export is missing. To avoid the need for
3384 this on patchless clients the deathrow inode reaper is turned
3385 off, and we depend on the VM to clean up old inodes. This
3386 dependency was during via the fix for bug 12181.
3388 --------------------------------------------------------------------------------
3390 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3392 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3393 INTEROPERATE with older versions automatically. In many cases a
3394 special upgrade step is needed. Please read the
3395 user documentation before upgrading any part of a live system.
3396 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3397 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3398 * WARNING: Lustre configuration and startup changes are required with
3399 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3403 Severity : enhancement
3405 Description: MountConf
3406 Details : Lustre configuration is now managed via mkfs and mount
3407 commands instead of lmc and lconf. New obd types (MGS, MGC)
3408 are added for dynamic configuration management. See
3409 https://mail.clusterfs.com/wikis/lustre/MountConf for
3412 Severity : enhancement
3414 Description: dynamic OST addition
3415 Details : OSTs can now be added to a live filesystem
3417 Severity : enhancement
3419 Description: startup order invariance
3420 Details : MDTs and OSTs can be started in any order. Clients only
3421 require the MDT to complete startup.
3423 Severity : enhancement
3425 Description: parallel, asynchronous orphan cleanup
3426 Details : orphan cleanup is now performed in separate threads for each
3427 OST, allowing parallel non-blocking operation.
3429 Severity : enhancement
3431 Description: optimized stripe assignment
3432 Details : stripe assignments are now made based on ost space available,
3433 ost previous usage, and OSS previous usage, in order to try
3434 to optimize storage space and networking resources.
3436 Severity : enhancement
3438 Description: Permanently set tunables
3439 Details : All writable /proc/fs/lustre tunables can now be permanently
3440 set on a per-server basis, at mkfs time or on a live system.
3442 Severity : enhancement
3444 Description: Lustre message v2
3445 Details : Add lustre message format v2.
3447 Severity : enhancement
3449 Description: client OST exclusion list
3450 Details : Clients can be started with a list of OSTs that should be
3451 declared "inactive" for known non-responsive OSTs.
3455 Description: SPEC SFS validation failure on NFS v2 over lustre.
3456 Details : Changes the blocksize for regular files to be 2x RPC size,
3457 and not depend on stripe size.
3459 Severity : enhancement
3461 Description: Multiple MD RPCs in flight.
3462 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3463 To avoid overly-overloading MDS, introduce a limit on number of
3464 MDS RPCs in flight for a single client and add /proc controls
3465 to adjust this limit.
3467 Severity : enhancement
3469 Description: client read/write statistics
3470 Details : Add client read/write call usage stats for performance
3471 analysis of user processes.
3472 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3473 file access. extents_stats shows chunk size distribution.
3474 extents_stats_per_process show chunk size distribution per
3477 Severity : enhancement
3479 Description: mds statistics
3480 Details : Add detailed mds operations statistics in
3481 /proc/fs/lustre/mds/*/stats.
3485 Description: Failure of copying files with lustre special EAs.
3486 Details : Client side always return success for setxattr call for lustre
3487 special xattr (currently only "trusted.lov").
3492 Description: Refcount LNET uuids
3493 Details : The global LNET uuid list grew linearly with every startup;
3494 refcount repeated list entries instead of always adding to
3497 Severity : enhancement
3499 Description: Dynamic service threads
3500 Details : Within a small range, start extra service threads
3501 automatically when the request queue builds up.
3504 Frequency : mixed-endian client/server environments
3506 Description: mixed-endian crashes
3507 Details : The new msg_v2 system had some failures in mixed-endian
3510 Severity : enhancement
3512 Description: Easy OST removal
3513 Details : OSTs can be permanently deactivated with e.g. 'lctl
3514 conf_param lustre-OST0001.osc.active=0'
3516 Severity : enhancement
3518 Description: MGS proc entries
3519 Details : Added basic proc entries for the MGS showing what filesystems
3522 Severity : enhancement
3524 Description: provide MGS failover
3525 Details : Added config lock reacquisition after MGS server failover.
3527 Severity : enhancement
3529 Description: add Linux 2.4 support
3530 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3534 Description: a large application tries to do I/O to the same resource and dies
3535 in the middle of it.
3536 Details : Check the req->rq_arrival time after the call to
3537 ost_brw_lock_get(), but before we do anything about
3538 processing it & sending the BULK transfer request. This
3539 should help move old stale pending locks off the queue as
3540 quickly as obd_timeout.
3543 Frequency : when an incorrect nid is specified during startup
3545 Description: ptlrpc connect to non-existant node causes kernel crash
3546 Details : LNET can't be re-entered from an event callback, which
3547 happened when we expire a message after the export has been
3548 cleaned up. Instead, hand the zombie cleanup off to another
3551 Severity : enhancement
3553 Description: plain/inodebits lock performance improvement
3554 Details : Grouping plain/inodebits in granted list by their request modes
3555 and bits policy, thus improving the performance of search through
3559 Frequency : only if OST filesystem is corrupted
3561 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3562 Details : for a short time RPCs with bulk IO are in the replay list,
3563 but replay of bulk IOs is unimplemented. If the OST filesystem
3564 is corrupted due to disk cache incoherency and then replay is
3565 started it is possible to trip an assertion. Avoid putting
3566 committed RPCs into the replay list at all to avoid this issue.
3569 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3571 Description: System hang on startup
3572 Details : This bug allowed the liblustre (e.g. catamount) client to
3573 return to the app before handling all startup RPCs. This
3574 could leave the node unresponsive to lustre network traffic
3575 and manifested as a server ptllnd timeout.
3577 Severity : enhancement
3579 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3580 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3581 boolean to control whether to print peer debug info when a
3582 client's RPC times out.
3585 Frequency : only for kernels with patches from Lustre below 1.4.3
3587 Description: Remove old rdonly API
3588 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3591 Frequency : only for devices with external journals
3593 Description: Set external device read-only also
3594 Details : During a commanded failover stop, we set the disk device
3595 read-only while the server shuts down. We now also set any
3596 external journal device read-only at the same time.
3599 Frequency : when upgrading from 1.4 while trying to change parameters
3601 Description: The wrong (new) MDC name was used when setting parameters for
3602 upgraded MDT's. Also allows changing of OSC (and MDC)
3603 parameters if --writeconf is specified at tunefs upgrade time.
3606 Frequency : when setting specific ost indicies
3608 Description: QOS code breaks on skipped indicies
3609 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3610 created with --index need not be sequential.
3612 Severity : enhancement
3614 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3615 Details : The uninit_groups feature works in conjunction with the kernel
3616 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3617 up the pass1 processing of e2fsck. This is a read-only feature
3618 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3619 mount filesystems that have had this feature enabled.
3621 Severity : enhancement
3623 Description: Improve multi-block allocation algorithm to avoid fragmentation
3624 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3625 allocation locality and avoid filesystem fragmentation.
3627 ------------------------------------------------------------------------------
3629 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3631 * Support for kernels:
3632 2.4.21-47.0.1.EL (RHEL 3)
3633 2.6.5-7.283 (SLES 9)
3634 2.6.9-42.0.10.EL (RHEL 4)
3635 2.6.12.6 vanilla (kernel.org)
3636 2.6.16.27-0.9 (SLES 10)
3637 * Recommended e2fsprogs version: 1.39.cfs5
3639 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3643 Frequency : occasional, depends on client load and configuration
3644 Bugzilla : 12181, 12203
3645 Description: data loss for recently-modified files
3647 Details : In some cases it is possible that recently written or created
3648 files may not be written to disk in a timely manner (this should
3649 normally be within 30s unless client IO load is very high).
3650 The problem appears as zero-length files or files that are a
3651 multiple of 1MB in size after a client crash or client eviction
3652 that are missing data at the end of the file.
3654 This problem is more likely to be hit on clients where files are
3655 repeatedly created and unlinked in the same directory, clients
3656 have a large amount of RAM, have many CPUs, the filesystem has
3657 many OSTs, the clients are rebooted frequently, and/or the files
3658 are not accessed by other nodes after being written.
3660 The presence of the problem can be detected by looking at
3661 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3662 smaller than the second (nr_unused) then dirty files will not
3663 be flushed automatically to disk. "sync; sleep 10" should be
3664 run several times on the node before unmounting it to update
3665 Lustre (this is also safe to run on nodes without this problem).
3667 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3668 that can cause this same problem, so customers using that kernel
3669 also need to update the kernel in addition to Lustre. In order
3670 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3672 It is normal that files written just before a client crash (less
3673 than 30s) may not yet have been flushed to disk, even for local
3677 Frequency : frequent on thin XT3 nodes
3679 Description: UUID collision on thin XT3 Linux nodes
3680 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3681 randomly, since we relied on an insufficiently-seeded PRNG.
3686 Description: OSS hangs after "All ost request buffers busy"
3687 Details : A deadlock between quota and journal operations caused OSS
3688 hangs after printing "All ost request buffers busy."
3691 Frequency : always on liblustre builds
3693 Description: Cleanup compiler warnings on liblustre
3696 Frequency : always on liblustre builds on XT3
3698 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3702 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3703 > 32000 subdirectories
3704 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3705 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3706 subdirectory count crosses 32000. This will aid e2fsck to
3707 correctly handle more than 32000 subdirectories.
3710 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3712 Description: System hang on startup
3713 Details : This bug allowed the liblustre (e.g. catamount) client to
3714 return to the app before handling all startup RPCs. This
3715 could leave the node unresponsive to lustre network traffic
3716 and manifested as a server ptllnd timeout.
3718 Severity : enhancement
3720 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3721 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3722 boolean to control whether to print peer debug info when a
3723 client's RPC times out.
3728 Description: make O_SYNC working on 2.6 kernels
3729 Details : 2.6 kernels use different method for mark pages for write,
3730 so need add a code to lustre for O_SYNC work.
3735 Description: Failure to close file and release space on NFS
3736 Details : Put inode details into lock acquired in ll_intent_file_open.
3737 Use mdc_intent_lock in ll_intent_open to properly
3738 detect all kind of errors unhandled by mdc_enqueue.
3743 Description: proc file read during shutdown sometimes raced obd removal,
3745 Details : Add lock to prevent obd access after proc file removal.
3748 Frequency : Only for files larger than 4GB on 32-bit clients.
3750 Description: improperly doing page alignment of locks
3751 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3752 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3757 Description: RPCs being resent when they shouldn't be
3758 Details : Some RPCs that should not be resent are being resent. This
3759 can cause inconsistencies in the RPC state machine. Do not
3760 resend such requests.
3763 Frequency : rare, only with NFS export
3765 Description: Crash on NFS re-export node
3766 Details : under very unusual load conditions an assertion is hit in
3767 ll_intent_file_open()
3770 Frequency : only if OST filesystem is corrupted
3772 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3773 Details : for a short time RPCs with bulk IO are in the replay list,
3774 but replay of bulk IOs is unimplemented. If the OST filesystem
3775 is corrupted due to disk cache incoherency and then replay is
3776 started it is possible to trip an assertion. Avoid putting
3777 committed RPCs into the replay list at all to avoid this issue.
3782 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3783 Details : Large single O_DIRECT read and write calls can fail to allocate
3784 a sufficiently large buffer to process the request. In case of
3785 allocation failure the allocation is retried with a smaller
3786 buffer and broken into smaller requests.
3788 Severity : enhancement
3790 Description: Add -o localflock option to simulate old noflock behaviour.
3791 Details : This will achieve local-only flock/fcntl locks coherentness.
3796 Description: versioning check is incomplete
3797 Details : Checking the version difference of client vs. server, report
3798 error if the gap is too big.
3803 Description: add support PG_writeback bit
3804 Details : add support for PG_writeback bit for Lustre, for more carefull
3805 work with page cache in 2.6 kernel. This also fix some deadlocks
3806 and remove hack for work O_SYNC with 2.6 kernel.
3808 Severity : enhancement
3810 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3811 Details : The uninit_groups feature works in conjunction with the kernel
3812 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3813 up the pass1 processing of e2fsck. This is a read-only feature
3814 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3815 mount filesystems that have had this feature enabled.
3817 Severity : enhancement
3819 Description: Improve multi-block allocation algorithm to avoid fragmentation
3820 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3821 allocation locality and avoid filesystem fragmentation.
3823 ------------------------------------------------------------------------------
3825 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3827 * Support for kernels:
3828 2.6.9-42.0.3.EL (RHEL 4)
3829 2.6.5-7.276 (SLES 9)
3830 2.4.21-47.0.1.EL (RHEL 3)
3831 2.6.12.6 vanilla (kernel.org)
3832 2.6.16.21-0.8 (SLES10)
3833 * Recommended e2fsprogs version: 1.39.cfs2-0
3835 * The backwards-compatible /proc/sys/portals symlink has been removed
3836 in this release. Before upgrading, please ensure that you change
3837 any configuration scripts or /etc/sysctl.conf files that access
3838 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3839 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3840 in advance of the upgrade on any system running Lustre 1.4.6 or
3841 newer, since /proc/sys/lnet was added in that version.
3842 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3846 Frequency : only when quota is used
3848 Description: avoid scanning export list for quota master
3849 Details : Change the algorithms to avoid scanning export list in order
3850 to improve the efficiency.
3853 Frequency : MDS failover only, very rarely
3855 Description: "went back in time" messages on mds failover
3856 Details : The greatest transno may be lost when the current operation
3857 finishes with an error (transno==0) and the client's last_rcvd
3858 record is over-written. Save the greatest transno in the
3859 mds_last_transno for this case.
3862 Frequency : always for specific kernels and striping counts
3864 Description: client may get "Matching packet too big" without ACL support
3865 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3866 when trying to access files in certain configurations. The
3867 clients should in fact be denied when mounting because they do
3868 not understand ACLs.
3871 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3873 Description: many clients connecting with IO in progress causes connect timeouts
3874 Details : Avoid synchronous journal commits to avoid delays caused by many
3875 clients connecting/disconnecting when bulk IO is in progress.
3876 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3877 OST_IO_PORTAL to avoid delays behind potentially many pending
3881 Frequency : occasionally with multiple writers to a single file
3883 Description: shared writes to file may result in wrong size reported by stat()
3884 Details : Allow growing of kms when extent lock is cancelled
3887 Frequency : always with random mmap IO to multi-striped file
3889 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3890 Details : Only if the hole is at the end of OST object so that kms is too
3891 small. Fix is to increase kms accordingly in ll_nopage.
3894 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3896 Description: writes to a missing object would leak memory on the OST
3897 Details : If there is an inconsistency between the MDS and OST filesystems,
3898 such that the MDS references an object that doesn't exist, writes
3899 to that object will leak memory due to incorrect cleanup in the
3900 error handling path, eventually running out of memory on the OST.
3905 Description: Creating too long symlink causes lustre errors
3906 Details : Check symlink and name lengths before sending requests to MDS.
3909 Frequency : only if flock is enabled (not on by default)
3911 Description: posix locks not released on fd closure on 2.6.9+
3912 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3913 kernels, this caused such locks not to be released on fd close and
3914 then assertions on fs unmount about still used locks.
3917 Frequency : MDS failover only, very rarely
3919 Description: clients may get ASSERTION(granted_lock != NULL)
3920 Details : When request was taking a long time, and a client was resending
3921 a getattr by name lock request. The were multiple lock requests
3922 with the same client lock handle and
3923 mds_getattr_name->fixup_handle_for_resent_request found one of the
3924 lock handles but later failed with ASSERTION(granted_lock != NULL).
3929 Description: handle->h_buffer_credits > 0, assertion failure
3930 Details : h_buffer_credits is zero after truncate, causing assertion
3931 failure. This patch extends the transaction or creates a new
3935 Frequency : NFS re-export or patchless client
3936 Bugzilla : 11179, 10796
3937 Description: Crash on NFS re-export node (__d_move)
3938 Details : We do not want to hash the dentry if we don't have a lock.
3939 But if this dentry is later used in d_move, we'd hit uninitialised
3940 list head d_hash, so we just do this to init d_hash field but
3941 leave dentry unhashed.
3944 Frequency : NFS re-export or patchless client
3946 Description: NFS exports has problem with symbolic link
3947 Details : lustre client didn't properly install dentry when re-exported
3948 to NFS or running patchless client.
3951 Frequency : NFS re-export or patchless client
3953 Description: Various nfs/patchless fixes.
3954 Details : fixes reuse disconected alias for lookup process - this fixes
3955 warning "find_exported_dentry: npd != pd",
3956 fix permission error with open files at nfs.
3957 fix apply umask when do revalidate.
3960 Frequency : occasional
3962 Description: Crash on NFS re-export node
3963 Details : calling clear_page() on the wrong pointer triggered oops in
3964 generic_mapping_read().
3967 Frequency : rarely, using O_DIRECT IO
3969 Description: unaligned directio crashes client with LASSERT
3970 Details : check for unaligned buffers before trying any requests.
3973 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3975 Description: stale data returned from RAID cache
3976 Details : If only a small amount of IO is done to the RAID device before
3977 reading it again it is possible to get stale data from the RAID
3978 cache instead of reading it from disk.
3981 Frequency : always for sles10 kernel
3983 Description: sles10 support
3984 Details : ll_follow_link: compile fixes and using of nd_set_link
3985 under newer kernels.
3988 Frequency : depends on arch, kernel and compiler version, always on sles10
3991 Description: recursive or deep enough symlinks cause stack overflow
3992 Details : getting rid of large stack-allocated variable in
3996 Frequency : depends on hardware
3998 Description: lustre write performance loss in the SLES10 kernel
3999 Details : the performance loss is caused by using of write barriers in the
4000 ext3 code. The SLES10 kernel turns barrier support on by
4001 default. The fix is to undo that change for ldiskfs.
4003 ------------------------------------------------------------------------------
4005 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4007 * Support for kernels:
4008 2.6.9-42.0.3EL (RHEL 4)
4009 2.6.5-7.276 (SLES 9)
4010 2.4.21-47.0.1.EL (RHEL 3)
4011 2.6.12.6 vanilla (kernel.org)
4015 Frequency : quota enabled and large files being deleted
4017 Description: releasing more than 4GB of quota at once hangs OST
4018 Details : If a user deletes more than 4GB of files on a single OST it
4019 will cause the OST to spin in an infinite loop. Release
4020 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4025 Description: statfs data retrieved from /proc may be stale or zero
4026 Details : When reading per-device statfs data from /proc, in the
4027 {kbytes,files}_{total,free,avail} files, it may appear
4028 as zero or be out of date.
4031 Frequency : systems with MD RAID1 external journal devices
4033 Description: lconf's call to blkid is confused by RAID1 journal devices
4034 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4035 of returning all block devices that match the journal UUID.
4038 Frequency : always, for aggregate stripe size over 4GB
4040 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4041 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4042 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4045 Frequency : always if "lfs find" used on a local file/directory
4047 Description: "lfs find" segfaults if used on a local file/directory
4048 Details : The case where a directory component was not specified wasn't
4049 handled correctly. Handle this properly.
4052 Frequency : always on ppc64
4054 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4055 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4056 perform endianness conversion.
4059 Frequency : rarely (truncate to non-zero file size after write under load)
4060 Bugzilla : 10730, 10687
4061 Description: Files padded with zeros to next 4K multiple
4062 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4063 it is possible that files that are truncated to a non-zero size
4064 immediately after being written are filled with zero bytes beyond
4065 the truncated size. No file data is lost.
4067 Severity : enhancement
4069 Description: Allow recovery/failover for liblustre clients.
4070 Details : liblustre clients were unaware of failover configurations until
4073 Severity : enhancement
4075 Description: user file locks should fail when not mounting with flock option
4076 Details : Set up an error-returning stub in ll_file_operations.lock field
4077 to prevent incorrect behaviour when client is mounted without
4078 flock option. Also, set up properly f_op->flock field for
4082 Frequency : always on ia64
4084 Description: "lfs df" loops on printing out MDS statfs information
4085 Details : The obd_ioctl_data was not initialized and in some systems
4086 this caused a failure during the ioctl that did not return
4087 an error. Initialize the struct and return an error on failure.
4090 Frequency : SLES 9 only
4092 Description: Error of copying files with lustre special EAs as root
4093 Details : Client side always return success for setxattr call for lustre
4094 special xattr (currently only "trusted.lov").
4097 Frequency : rarely on clusters with both ia64+i386 clients
4099 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4100 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4101 clients are doing concurrent writes to the same file it is
4102 possible that the ia64 clients may LASSERT because the OST
4103 extent locks are not PAGE_SIZE aligned. Ensure that grown
4104 locks are always aligned on the request boundary.
4107 Frequency : specific use, occasional
4109 Description: Overwriting in use executable truncates on-disk binary image
4110 Details : If one node attempts to overwrite an executable in use by
4111 another node, we now correctly return ETXTBSY instead of
4112 truncating the file.
4114 Severity : enhancement
4116 Description: Async OSC create to avoid the blocking unnecessarily.
4117 Details : If an OST has no remaining object, system will block on the creating
4118 when need to create a new object on this OST. Now, ways use
4119 pre-created objects when available, instead of blocking on an
4120 empty osc while others are not empty. If we must block, we block
4121 for the shortest possible period of time.
4126 Description: chmod on Lustre root is propagated to other clients
4127 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4128 invalid by the follow_mount time.
4131 Frequency : liblustre clients only
4133 Description: Race in 'instant cancel' lock handling could lead to such locks
4134 never to be granted in case of SMP MDS
4135 Details : Do not destroy not yet granted but cbpending locks in
4139 Frequency : replay/resend of open
4141 Description: non null lock assetion failure in mds_intent_policy
4142 Details : Trying to replay/resend lockless open requests resulted in
4143 mds_open() returning 0 with no lock. Now it sets a flag if
4144 a lock is going to be returned.
4146 Severity : enhancement
4148 Description: Checksum enhancements
4149 Details : New checksum enhancements allow for resending RPCs that failed
4152 Severity : enhancement
4154 Description: Tunables on number of dirty pages in cacche
4155 Details : Allow to set limit on number of dirty pages cached.
4160 Description: client crash on unmount - lock still has references
4161 Details : In some error handling cases it was possible to leak a lock
4162 reference on a client while accessing a file. This was not
4163 harmful to the client during operation, but would cause the
4164 client to crash when the filesystem is unmounted.
4167 Frequency : specific case, rare
4169 Description: ETXTBSY on mds though file not in use by client
4170 Details : ETXTBSY is no longer incorrectly returned when attempting to
4171 chmod or chown a directory that the user previously tried to
4172 execute or a currently-executing binary.
4175 Frequency : extremely rare except on liblustre-based clients
4177 Description: Lustre space not freed when files are deleted
4178 Details : Clean up open-unlinked files after client eviction. Previously
4179 the unlink was skipped and the files remained as orphans.
4184 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4185 Details : In some cases it was possible to send a blocking callback to
4186 a client doing a glimpse, even though that client didn't get
4187 a lock granted. When the glimpse lock is cancelled on the OST
4188 the freed lock is left on the waiting list and corrupted the list.
4191 Frequency : all core dumps
4193 Description: Broke core dumps to lustre
4194 Details : Negative dentry may be unhashed if parent does not have UPDATE
4195 lock, but some callers, e.g. do_coredump, expect dentry to be
4196 hashed after successful create, hash it in ll_create_it.
4198 ------------------------------------------------------------------------------
4200 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4202 * Support for kernels:
4203 2.6.9-42.0.2.EL (RHEL 4)
4204 2.6.5-7.276 (SLES 9)
4205 2.4.21-40.EL (RHEL 3)
4206 2.6.12.6 vanilla (kernel.org)
4210 Frequency : always on RHEL 3
4212 Description: Number of open files grows over time
4213 Details : The number of open files grows over time, whether or not
4214 Lustre is started. This was due to a filp leak introduced
4215 by one of our kernel patches.
4217 ------------------------------------------------------------------------------
4219 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4221 * Support for kernels:
4222 2.6.9-42.EL (RHEL 4)
4223 2.6.5-7.276 (SLES 9)
4224 2.4.21-40.EL (RHEL 3)
4225 2.6.12.6 vanilla (kernel.org)
4230 Bugzilla : 5719, 9635, 9792, 9684
4231 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4232 Details : If a server is under heavy load and cannot reply to new
4233 connection requests before the client resends the (re)connect,
4234 the connection handling code can behave badly if two service
4235 threads are concurrently handing separate (re)connections from
4236 the same client. Add better locking to the connection handling
4237 code, and ensure that only a single connection will be processed
4238 for a given client UUID, even if the lock is dropped.
4240 Severity : enhancement
4242 Description: add TCP zero-copy support to kernel
4243 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4244 sends if the hardware supports scatter-gather and checksumming.
4245 This allows socklnd to do client-write and server-read more
4246 efficiently and reduce CPU utilization from skbuf copying.
4249 Frequency : only if NFS exporting from client
4251 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4252 Details : If setting attributes on a file created under NFS that had
4253 never been opened it would be possible to oops the client
4254 if the file had no objects.
4257 Frequency : always for liblustre
4259 Description: liblustre client does MDS+OSTs setattr RPC for each write
4260 Details : When doing a write from a liblustre client, the client
4261 incorrectly issued an RPC to the MDS and each OST the file was
4262 striped over in order to update the timestamps. When writing
4263 with small chunks and many clients this could overwhelm the MDS
4264 with RPCs. In all cases it would slow down the write because
4265 these RPCs are unnecessary.
4267 Severity : enhancement
4269 Description: allow number of MDS service threads to be changed at module load
4270 Details : It is now possible to change the number of MDS service threads
4271 running. Adding "options mds mds_num_threads={N}" to the MDS's
4272 /etc/modprobe.conf will set the number of threads for the next
4273 time Lustre is restarted (assuming the "mds" module is also
4274 reloaded at that time). The default number of threads will
4275 stay the same, 32 for most systems.
4280 Description: OST crash if filesystem is unformatted or corrupt
4281 Details : If an OST is started on a device that has never been formatted
4282 or if the filesystem is corrupt and cannot even mount then the
4283 error handling cleanup routines would dereference a NULL pointer.
4288 Description: NULL pointer deref in llap_from_page.
4289 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4290 incorrect) mapping if the page was truncated/reclaimed while it was
4291 searched for. Check for this condition and skip such pages when
4292 doing readahead. Introduce extra check to llap_from_page() to
4293 verify page->mapping->host is non-NULL (so page is not anonymous).
4296 Frequency : Sometimes when using sys_sendfile
4298 Description: "page not covered by a lock" warnings from ll_readpage
4299 Details : sendfile called ll_readpage without right page locks present.
4300 Now we introduced ll_file_sendfile that does necessary locking
4301 around call to generic_file_sendfile() much like we do in
4305 Frequency : with certain MDS communication failures at client mount time
4307 Description: NULL pointer deref after failed client mount
4308 Details : a client connection request may delayed by the network layer
4309 and not be sent until after the PTLRPC layer has timed out the
4310 request. If the client fails the mount immediately it will try
4311 to clean up before the network times out the request. Add a
4312 reference from the request import to the obd device and delay
4313 the cleanup until the network drops the request.
4316 Frequency : occasionally during client (re)connect
4318 Description: assertion failure during client (re)connect
4319 Details : processing a client connection request may be delayed by the
4320 client or server longer than the client connect timeout. This
4321 causes the client to resend the connection request. If the
4322 original connection request is replied in this interval, the
4323 client may trip an assertion failure in ptlrpc_connect_interpret()
4324 which thought it would be the only running connect process.
4327 Frequency : only with obd_echo servers and clients that are rebooted
4329 Description: kernel BUG accessing uninitialized data structure
4330 Details : When running an obd_echo server it did not start the ping_evictor
4331 thread, and when a client was evicted an uninitialized data
4332 structure was accessed. Start the ping_evictor in the RPC
4333 service startup instead of the OBD startup.
4335 Severity : enhancement
4336 Bugzilla : 10193 (patchless)
4337 Description: Remove dependency on various unexported kernel interfaces.
4338 Details : No longer need reparent_to_init, exit_mm, exit_files,
4339 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4342 Frequency : rare (only users of deprecated and unsupported LDAP config)
4344 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4345 Details : LDAP apparently contains 'lustreName' attributes instead of
4346 'name'. A simple remapping of the name is sufficient.
4349 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4351 Description: waiting_locks_callback trips kernel BUG if client is evicted
4352 Details : Running with the dump_on_timeout debug flag turned on makes
4353 it possible that the waiting_locks_callback() can try to dump
4354 the Lustre kernel debug logs from an interrupt handler. Defer
4355 this log dumping to the expired_lock_main() thread.
4357 Severity : enhancement
4359 Description: Support NFS exporting on 2.6 kernels.
4360 Details : Implement non-rawops metadata methods for NFS server to use without
4361 changing NFS server code.
4364 Frequency : very rare (synthetic metadata workload only)
4366 Description: two racing renames might cause an MDS thread to deadlock
4367 Details : Running the "racer" program may cause one MDS thread to rename
4368 a file from being the source of a rename to being the target of
4369 a rename at exactly the same time that another thread is doing
4370 so, and the second thread has already enqueued these locks after
4371 doing a lookup of the target and is trying to relock them in
4372 order. Ensure that we don't try to re-lock the same resource.
4375 Frequency : only very large systems with liblustre clients
4377 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4378 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4379 all OSTs in parallel. This allows the eviction of stale liblustre
4380 clients to proceed much faster than if they were done in series,
4381 and also offers similar improvements for other set_info RPCs.
4386 Description: excessive CPU usage during initial read phase on client
4387 Details : During the initial read phase on a client, it would agressively
4388 retry readahead on the file, consuming too much CPU and impacting
4389 performance (since 1.4.5.8). Improve the readahead algorithm
4390 to avoid this, and also improve some other common cases (read
4391 of small files in particular, where "small" is files smaller than
4392 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4397 Description: MDS crash when receiving packet with unknown intent.
4398 Details : Do not LBUG in unknown intent case, just return -EFAULT
4400 Severity : enhancement
4401 Bugzilla : 9293, 9385
4402 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4403 Details : Do not serialize getattr (non-intent version) and statfs.
4406 Frequency : occasional, when OST network is overloaded/intermittent
4408 Description: client evicted by OST after bulk IO timeout
4409 Details : If a client sends a bulk IO request (read or write) the OST
4410 may evict the client if it is unresposive to its data GET/PUT
4411 request. This is incorrect if the network is overloaded (takes
4412 too long to transfer the RPC data) or dropped the OST GET/PUT
4413 request. There is no need to evict the client at all, since
4414 the pinger and/or lock callbacks will handle this, and the
4415 client can restart the bulk request.
4418 Frequency : Always when mmapping file with no objects
4420 Description: client crashes when mmapping file with no objects
4421 Details : Check that we actually have objects in a file before doing any
4422 operations on objects in ll_vm_open, ll_vm_close and
4428 Description: Request leak when working with deleted CWD
4429 Details : Introduce advanced request refcount tracking for requests
4430 referenced from lustre intent.
4432 Severity : Enhancement
4434 Description: Cache open file handles on client.
4435 Details : MDS now will return special lock along with openhandle, if
4436 requested and client is allowed to hold openhandle, even if unused,
4437 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4438 closing files for every read/write openration.
4440 Severity : Enhancement
4442 Description: Cache open negative dentries on client when possible.
4443 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4444 negative dentries on lock revocation.
4449 Description: Remounting a client read-only wasn't possible with a zconf mount
4450 Details : It wasn't possible to remount a client read-only with llmount.
4452 Severity : enhancement
4453 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4454 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4455 MPICH is included to add Lustre-specific ADIO interfaces.
4456 This is based closely on the UFS ADIO layer and only differs
4457 in file creation, in order to allow the OST striping to be set.
4458 This is user-contributed code and not supported by CFS.
4463 Description: extended inode attributes (immutable, append-only) work improperly
4464 when 2.4 and 2.6 kernels are used on client/server or vice versa
4465 Details : Introduce kernel-independent values for these flags.
4467 Severity : enhancement
4470 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4471 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4472 granularity. Now, Fractional MB granularity is be supported,
4473 this is very useful for low memory system.
4475 Severity : enhancement
4477 Description: Getattr by fid
4478 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4479 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4480 have locks and avoids getattr rpc requests for it on every stat.
4483 Frequency : Always, for filesystems larger than 2TB
4485 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4486 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4487 filesystems larger than 2TB to be mounted with mballoc enabled.
4490 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4492 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4493 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4494 filesystems larger than 2TB is mounted then the kernel will
4495 silently corrupt the start of the filesystem. CONFIG_LBD is
4496 enabled for all CFS-supported kernels, but the possibility of
4497 this happening with a modified kernel config exists.
4499 Severity : enhancement
4501 Description: add client O_DIRECT support for 2.6 kernels
4502 Details : It is now possible to do O_DIRECT reads and writes to files
4503 in the Lustre client mountpoint on 2.6 kernel clients.
4505 Severity : enhancement
4507 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4508 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4509 parallel, not waiting for response from every OST before sending
4510 a rpc to the next OST.
4515 Description: setattr vs write race when updating file timestamps
4516 Details : Client processes that update a file timestamp into the past
4517 right after writing to the file (e.g. tar) it is possible that
4518 the updated file modification time can be reset to the current
4519 time due to a race between processing the setattr and write RPC.
4521 Severity : enhancement
4523 Description: Bring 'lfs find' closer in line with regular Linux find.
4524 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4525 -print0 options and obtains all the needed info through the lustre
4528 Severity : enhancement
4530 Description: support up to 1024 configured devices on one node
4531 Details : change obd_dev array from statically allocated to dynamically
4532 allocated structs as they are first used to reduce memory usage
4537 Description: Flush dirty partially truncated pages during truncate
4538 Details : Immediatelly flush partially truncated pages in filter_setattr,
4539 this way we completely avoid having any pages in page cache on OST
4540 and can retire ugly workarounds during writes to flush such pages.
4545 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4546 Details : protect inode from truncation within vfs_unlink() context
4547 just take a reference before calling vfs_unlink() and release it
4548 when parent's i_sem is free.
4553 Description: last_id value checked outside lock on OST caused LASSERT failure
4554 Details : If there were multiple MDS->OST object precreate requests in
4555 flight, it was possible that the OST's last object id was checked
4556 outside a lock and incorrectly tripped an assertion. Move checks
4557 inside locks, and discard old precreate requests.
4560 Frequency : always, if extents are used on OSTs
4562 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4563 Details : OSTs using the extents format would not zero the high 16 bits of
4564 the index physical block number. This is not a problem for any
4565 OST filesystems smaller than 16TB, and no kernels support ext3
4566 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4567 new/modified files) and can be fixed for existing filesystems
4568 with e2fsprogs-1.39-cfs1.
4573 Description: import connection selection may be incorrect if timer wraps
4574 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4575 import connections to be ignored if the 32-bit jiffies counter
4576 wraps. Use a 64-bit jiffies counter.
4579 Frequency : very large clusters immediately after boot
4581 Description: LNET request buffers exhausted under heavy short-term load
4582 Details : If a large number of client requests are generated on a service
4583 that has previously never seen so many requests it is possible
4584 that the request buffer growth cannot keep up with the spike in
4585 demand. Instead of dropping incoming requests, they are held in
4586 the LND until the RPC service can accept more requests.
4589 Frequency : Sometimes during replay
4591 Description: Assertion failure in ll_local_open after replay.
4592 Details : If replay happened on an open request reply before we were able
4593 to set replay handler, reply will become not swabbed tripping the
4594 assertion in ll_local_open. Now we set the handler right after
4595 recognising of open request
4598 Frequency : very rare
4600 Description: kernel reports "badness in vsnprintf"
4601 Details : Reading from the "recovery_status" /proc file in small chunks
4602 may cause a negative length in lprocfs_obd_rd_recovery_status()
4603 call to vsnprintf() (which is otherwise harmless). Exit early
4604 if there is no more space in the output buffer.
4606 Severity : enhancement
4608 Description: clear OBD RPC statistics by writing to them
4609 Details : It is now possible to clear the OBD RPC statistics by writing
4610 to the "stats" file.
4615 Description: Client mtime is not the same on different clients after utimes
4616 Details : In some cases, the client was using the utimes() syscall on
4617 a file cached on another node. The clients now validate the
4618 ctime from the MDS + OSTs to determine which one is right.
4623 Description: Inability to activate failout mode
4624 Details : lconf script incorrectly assumed that in pythong string's numeric
4625 value is used in comparisons.
4628 Frequency : always with multiple stripes per file
4630 Description: Inefficient object allocation for mutli-stripe files
4631 Details : When selecting which OSTs to stripe files over, for files with
4632 a stripe count that divides evenly into the number of OSTs,
4633 the MDS is always picking the same starting OST for each file.
4634 Return the OST selection heuristic to the original design.
4639 Description: mount failures may take full timeout to return an error
4640 Details : Under some heavy load conditions it is possible that a
4641 failed mount can wait for the full obd_timeout interval,
4642 possibly several minutes, before reporting an error.
4643 Instead return an error as soon as the status is known.
4645 Frequency : quota enabled and large files being deleted
4647 Description: releasing more than 4GB of quota at once hangs OST
4648 Details : If a user deletes more than 4GB of files on a single OST it
4649 will cause the OST to spin in an infinite loop. Release
4650 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4655 Description: statfs data retrieved from /proc may be stale or zero
4656 Details : When reading per-device statfs data from /proc, in the
4657 {kbytes,files}_{total,free,avail} files, it may appear
4658 as zero or be out of date.
4661 Frequency : systems with MD RAID1 external journal devices
4663 Description: lconf's call to blkid is confused by RAID1 journal devices
4664 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4665 of returning all block devices that match the journal UUID.
4668 Frequency : always, for aggregate stripe size over 4GB
4670 Description: assertion fails when trying to use 4GB stripe size
4671 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4672 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4675 Frequency : always on ppc64
4677 Description: the first write on an ext3 filesystem with mballoc got stuck
4678 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4679 perform endianness conversion.
4681 ------------------------------------------------------------------------------
4683 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4685 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4686 INTEROPERATE with older versions automatically. Please read the
4687 user documentation before upgrading any part of a live system.
4688 * WARNING: Lustre networking configuration changes are required with
4689 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4692 * Support for kernels:
4693 2.6.9-22.0.2.EL (RHEL 4)
4694 2.6.5-7.244 (SLES 9)
4695 2.6.12.6 vanilla (kernel.org)
4698 Severity : enhancement
4699 Bugzilla : 7981/8208
4700 Description: Introduced Lustre Networking (LNET)
4701 Details : LNET is new networking infrastructure for Lustre, it includes
4702 a reorganized network configuration mode (see the user
4703 documentation for full details) as well as support for routing
4704 between different network fabrics. Lustre Networking Devices
4705 (LNDs) for the supported network fabrics have also been
4706 created for this new infrastructure.
4708 Severity : enhancement
4709 Description: Introduced Access control lists
4710 Details : clients can set ACLs on files and directories in order to have
4711 more fine-grained permissions than the standard Unix UGO+RWX.
4712 The MDS must be started with the "-o acl" mount option.
4714 Severity : enhancement
4715 Description: Introduced filesystem quotas
4716 Details : Administrators may now establish per-user quotas on the
4719 Severity : enhancement
4721 Description: Configuration change for the XT3
4722 The PTLLND is now used to run Lustre over Portals on the XT3
4723 The configure option(s) --with-cray-portals are no longer used.
4724 Rather --with-portals=<path-to-portals-includes> is used to
4725 enable building on the XT3. In addition to enable XT3 specific
4726 features the option --enable-cray-xt3 must be used.
4731 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4732 Details : A race between allocating a new llcd and re-getting the llcd_lock
4733 allowed another thread to grab newly-allocated llcd.
4735 Severity : enhancement
4737 Description: 2.6 OST async journal commit and locking fix to improve performance
4738 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4739 2.6 kernels are now async as they already were in 2.4 kernels so
4740 that they can commit concurrently with the network bulk transfer.
4741 For block-allocated files the filter allocation semaphore is held
4742 to avoid filesystem fragmentation during allocation. BKL lock
4743 removed for 2.6 xattr operations where it is no longer needed.
4748 Description: lconf incorrectly determined whether two IP networks could talk
4749 Details : In some more complicated routing and multiple-network
4750 configurations, lconf will avoid trying to make a network
4751 connection to a disjoint part of the IP space. It was doing the
4752 math incorrectly for one set of cases.
4757 Description: Fix for potential infinite loop processing records in an llog.
4758 Details : If an llog record is corrupted/zeroed, it is possible to loop
4759 forever in llog_process(). Validate the llog record length
4760 and skip the remainder of the block on error.
4763 Frequency : occasional (liblustre only)
4765 Description: liblustre could not open files whose last component is a symlink
4766 Details : sysio_path_walk() would incorrectly pass the open intent to
4767 intermediate path components.
4770 Frequency : rare (liblustre only with non-standard tuning)
4771 Bugzilla : 7201 (7350)
4772 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4773 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4774 after it has been released, only do this for referenced locks
4776 Severity : enhancement
4778 Description: specifying an (invalid) directory default stripe_size of -1
4779 would reset the directory default striping
4780 Details : stripe_size -1 was used internally to signal directory stripe
4781 removal, now use "all default" to signal dir stripe removal
4782 as a directory striping of "all default" is not useful
4785 Frequency : common for large clusters running liblustre clients
4787 Description: doing an ls when liblustre clients are running is slow
4788 Details : sending a glimpse AST to a liblustre client waits for every AST
4789 to time out, as liblustre clients will not respond. Since they
4790 cannot cache data we refresh the OST lock LVB from disk instead.
4792 Severity : enhancement
4794 Description: doing an ls at the same time as file IO can be slow
4795 Details : enqueue and other "small" requests can be blocked behind many
4796 large IO requests. Create a new OST IO portal for non-IO
4797 requests so they can be processed faster.
4800 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4802 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4803 created file with mode 000
4804 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4805 allow an ftruncate() as a non-root user to a file with mode 000.
4806 Lustre now respects this flag to disable mode checks when
4807 truncating a file owned by the user
4810 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4813 Description: Revoking locks from clients that went dead or catatonic might take
4815 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4816 cancellation of such locks instant on servers without waiting for
4817 any reply from clients. Clients drops these locks when cancel
4818 notification from server is received without replying.
4821 Frequency : liblustre-only, when liblustre client dies or becomes busy
4823 Description: Doing ls on Linux clients can take a long time with active
4825 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4826 granting such locks to it in the first place if possible. Locks
4827 are taken by proxy on the OST during the read or write and
4828 dropped immediately afterward. Add connect flags handling, do
4829 not grant locks to liblustre clients for glimpse ASTs.
4831 Severity : enhancement
4833 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4834 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4835 sequential reads and behaves badly with applications doing random
4836 reads. Improve it by reading ahead at least read region, and
4837 avoiding excessive large RPC for small reads.
4839 Severity : enhancement
4841 Description: Creating more than 1000 files for a single job may cause a load
4842 imbalance on the OSTs if there are also a large number of OSTs.
4843 Details : qos_prep_create() uses an OST index reseed value that is an
4844 even multiple of the number of available OSTs so that if the
4845 reseed happens in the middle of the object allocation it will
4846 still utilize the OSTs as uniformly as possible.
4851 Description: OST or MDS may oops in ping_evictor_main()
4852 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4853 but doesn't restart at beginning of obd_exports_timed list
4856 Severity : enhancement
4858 Description: improve by-nid export eviction on the MDS and OST
4859 Details : allow multiple exports with the same NID to be evicted at one
4860 time without re-searching the exports list.
4863 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4865 Description: MDS may oops in groups_free()
4866 Details : in rare race conditions a newly allocated group_info struct is
4867 freed again, and this can be NULL. The 2.4 compatibility code
4868 for groups_free() checked for a NULL pointer, but 2.6 did not.
4871 Frequency : common for liblustre clients doing little filesystem IO
4872 Bugzilla : 9352, 7313
4873 Description: server may evict liblustre clients accessing contended locks
4874 Details : if a client is granted a lock or receives a completion AST
4875 with a blocking AST already set it would not reply to the AST
4876 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4879 Frequency : lfs setstripe, only systems with more than 160 OSTs
4881 Description: unable to set striping with a starting offset beyond OST 160
4882 Details : llapi_create_file() incorrectly limited the starting stripe
4883 index to the maximum single-file stripe count.
4886 Frequency : LDAP users only
4888 Description: lconf did not handle in-kernel recovery with LDAP properly
4889 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4891 Severity : enhancement
4893 Description: bind OST threads to NUMA nodes to improve performance
4894 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4895 node and do their allocations there to localize memory access
4897 Severity : enhancement
4899 Description: llmount can determine client NID directly from Myrinet (GM)
4900 Details : the client NID code from gmnalnid was moved directly into
4901 llmount, removing the need to use this or specifying the
4902 client NID explicitly when mounting GM clients with zeroconf
4905 Frequency : if client is started with down MDS
4907 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4908 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4909 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4910 ptlrpc_import_delayed_req() because we didn't check if the
4911 request was interrupted, and we also didn't break out of the
4912 event loop if there was no timeout
4917 Description: data loss during non-page-aligned writes to a single file from
4918 both multiple nodes and multiple threads on one node at same time
4919 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4920 inconsistency led to false short-reads, that were cached and later
4921 used by ->prepare_write() to fill in partially written page,
4922 leading to data loss.
4925 Frequency : always, if lconf --abort_recovery used
4927 Description: lconf --abort_recovery fails with 'Operation not supported'
4928 Details : lconf was attempting to abort recovery on the MDT device and not
4931 Severity : enhancement
4933 Description: remove cleanup logs
4934 Details : replace lconf-generated cleanup logs with lustre internal
4935 cleanup routines. Eliminates the need for client-cleanup and
4938 Severity : enhancement
4940 Description: add support for EAs (user and system) on lustre filesystems
4941 Details : it is now possible to store extended attributes in the Lustre
4942 client filesystem, and with the user_xattr mount option it
4943 is possible to allow users to store EAs on their files also
4945 Severity : enhancement
4947 Description: Add possibility (config option) to show minimal available OST free
4949 Details : When compiled with --enable-mindf configure option, statfs(2)
4950 (and so, df) will return least minimal free space available from
4951 all OSTs as amount of free space on FS, instead of summary of
4952 free spaces of all OSTs.
4954 Severity : enhancement
4956 Description: do not expand extent locks acquired on OST-side
4957 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4958 by server: they are not cached anyway.
4961 Frequency : when mmap is used/binaries executed from Lustre
4963 Description: Unmmap pages before throwing them away from read cache.
4964 Details : llap_shrink cache now attempts to unmap pages before discarding
4965 them (if unmapping failed - do not discard). SLES9 kernel has
4966 extra checks that trigger if this unmapping is not done first.
4971 Description: lconf didn't resolve symlinks before checking to see whether a
4972 given mountpoint was already in use
4975 Frequency : when migrating failover services
4976 Bugzilla : 6395, 9514
4977 Description: When migrating a subset of services from a node (e.g. failback
4978 from a failover service node) the remaining services would
4979 time out and evict clients.
4980 Details : lconf --force (implied by --failover) sets the global obd_timeout
4981 to 5 seconds in order to quickly disconnect, but this caused
4982 other RPCs to time out too quickly. Do not change the global
4983 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4985 Severity : enhancement
4986 Frequency : if MDS is started with down OST
4987 Bugzilla : 9439,5706
4988 Description: Allow startup/shutdown of an MDS without depending on the
4989 availability of the OSTs.
4990 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4991 Add appropriate locking and lov-osc refcounts for safe
4992 cleaning. Add osc abort_inflight calls in case the
4993 synchronize never started.
4996 Frequency : occasional (Cray XT3 only)
4998 Description: root not authorized to access files in CRAY_PORTALS environment
4999 Details : The client process capabilities were not honoured on the MDS in
5000 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5001 been accessed by an authorized user then root was able to access
5002 the file on the local client also. The root user capabilities
5003 are now allowed on the MDS, as this environment has secure UID.
5006 Frequency : occasional
5008 Description: ldiskfs "too long searching" message happens too often
5009 Details : A debugging message (otherwise harmless) prints too often on
5010 the OST console. This has been reduced to only happen when
5011 there are fragmentation problems on the filesystem.
5016 Description: Division by zero in statfs when all OSCs are inactive
5017 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5018 lov_statfs divides by value returned by lov_get_stripecnt().
5022 Bugzilla : 9489, 3273
5023 Description: First write from each client to each OST was only 4kB in size,
5024 to initialize client writeback cache, which caused sub-optimal
5025 RPCs and poor layout on disk for the first written file.
5026 Details : Clients now request an initial cache grant at (re)connect time
5027 and so that they can start streaming writes to the cache right
5028 away and always do full-sized RPCs if there is enough data.
5029 If the OST is rebooted the client also re-establishes its grant
5030 so that client cached writes will be honoured under the grant.
5035 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5036 Details : Now I/O RPCs go to different portal number and (presumably) fast
5037 lock requests (and glimses) and other RPCs get their own service
5038 threads pool that should be able to service those RPCs
5041 Severity : enhancement
5043 Description: Ability to exchange lustre version between client and servers and
5044 issue warnings at client side if client is too old. Also for
5045 liblustre clients there is ability to refuse connection of too old
5047 Details : New 'version' field is added to connect data structure that is
5048 filled with version info. That info is later checked by server and
5052 Frequency : rare, liblustre only.
5053 Bugzilla : 9296, 9581
5054 Description: Two simultaneous writes from liblustre at offset within same page
5055 might proceed at the same time overwriting eachother with stale
5057 Details : I/O lock withing llu_file_prwv was released too early, before data
5058 actually was hitting the wire. Extended lock-holding time until
5059 server acknowledges receiving data.
5062 Frequency : extremely rare. Never observed in practice.
5064 Description: avoid generating lustre_handle cookie of 0.
5065 Details : class_handle_hash() generates handle cookies by incrementing
5066 global counter, and can hit 0 occasionaly (this is unlikely, but
5067 not impossible, because initial value of cookie counter is
5068 selected randonly). Value of 0 is used as a sentinel meaning
5069 "unassigned handle" --- avoid it. Also coalesce two critical
5070 sections in this function into one.
5072 Severity : enhancement
5074 Description: allow liblustre clients to delegate truncate locking to OST
5075 Details : To avoid overhead of locking, liblustre client instructs OST to
5076 take extent lock in ost_punch() on client's behalf. New connection
5077 flag is added to handle backward compatibility.
5079 Severity : enhancement
5080 Bugzilla : 4928, 7341, 9758
5081 Description: allow number of OST service threads to be specified
5082 Details : a module parameter allows the number of OST service threads
5083 to be specified via "options ost ost_num_threads={N}" in the
5084 OSS's /etc/modules.conf or /etc/modprobe.conf.
5088 Bugzilla : 6146, 9635, 9895
5089 Description: servers crash with bad pointer in target_handle_connect()
5090 Details : In rare cases when a client is reconnecting it was possible that
5091 the connection request was the last reference for that export.
5092 We would temporarily drop the export reference and get a new
5093 one, but this may have been the last reference and the export
5094 was just destroyed. Get new reference before dropping old one.
5096 Severity : enhancement
5097 Frequency : if client is started with failover MDS
5099 Description: Allow multiple MDS hostnames in the mount command
5100 Details : Try to read the configuration from all specified MDS
5101 hostnames during a client mount in case the "primary"
5104 Severity : enhancement
5106 Description: Stop sending data to evicted clients as soon as possible.
5107 Details : Check if the client we are about to send or are sending data to
5108 was evicted already. (Check is done every second of waiting,
5109 for which l_wait_event interface was extended to allow checking
5110 of exit condition at specified intervals).
5113 Frequency : rare, normally only when NFS exporting is done from client
5115 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5117 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5120 Frequency : always, if srand() is called before liblustre initialization
5122 Description: Liblustre uses system PRNG disturbing its usage by user application
5123 Details : Introduce internal to lustre fast and high-quality PRNG for
5124 lustre usage and make liblustre and some other places in generic
5125 lustre code to use it.
5127 Severity : enhancement
5128 Bugzilla : 9477, 9557, 9870
5129 Description: Verify that the MDS configuration logs are updated when xml is
5130 Details : Check if the .xml configuration logs are newer than the config
5131 logs stored on the MDS and report an error if this is the case.
5132 Request --write-conf, or allow starting with --old_conf.
5134 Severity : enhancement
5136 Description: Handle symlinks in the path when checking if Lustre is mounted.
5137 Details : Resolve intermediate symlinks when checking if a client has
5138 mounted a filesystem to avoid duplicate client mounts.
5143 Description: lconf can hit an error exception but still return success.
5144 Details : The lconf command catches the Command error exception at the top
5145 level script context and will exit with the associated exit
5146 status, but doesn't ensure that this exit status is non-zero.
5151 Description: failure of ptlrpc thread startup can cause oops
5152 Details : Starting a ptlrpc service thread can fail if there are a large
5153 number of threads or the server memory is very fragmented.
5154 Handle this without oopsing.
5157 Frequency : always, only if liblustre and non-default acceptor port was used
5159 Description: liblustre cannot connect to servers with non-default acceptor port
5160 Details : tcpnal_set_default_params() was not called and was therefore
5161 ignoring the environment varaible TCPNAL_PORT, as well as other
5162 TCPNAL_ environment variables
5167 Description: two objects could be created on the same OST for a single file
5168 Details : If an OST is down, in some cases it was possible to create two
5169 objects on a single OST for a single file. No problems other
5170 than potential performance impact and spurious error messages.
5174 Bugzilla : 5681, 9562
5175 Description: Client may oops in ll_unhash_aliases
5176 Details : Client dcache may become inconsistent in race condition.
5177 In some cases "getcwd" can fail if the current directory is
5183 Description: Inode refcounting problems in NFS export code
5184 Details : link_raw functions used to call d_instantiate without obtaining
5185 extra inode reference first.
5189 Bugzilla : 9942, 9903
5190 Description: Referencing freed requests leading to crash, memleaks with NFS.
5191 Details : We used to require that call to ll_revalidate_it was always
5192 followed by ll_lookup_it. Also with revalidate_special() it is
5193 possible to call ll_revalidate_it() twice for the same dentry
5194 even if first occurence returned success. This fix changes semantic
5195 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5196 reference on a request referred from the intent.
5197 ll_intent_release() then releases such a request.
5200 Frequency : rare, normally benchmark loads only
5202 Description: unlinked inodes were kept in memory on the client
5203 Details : If a client is repeatedly creating and unlinking files it
5204 can accumulate a lot of stale inodes in the inode slab cache.
5205 If there is no other client load running this can cause the
5206 client node to run out of memory. Instead flush old inodes
5207 from client cache that have the same inode number as a new inode.
5210 Frequency : SLES9 2.6.5 kernel and long filenames only
5211 Bugzilla : 9969, 10379
5212 Description: utime reports stale NFS file handle
5213 Details : SLES9 uses out-of-dentry names in some cases, which confused
5214 the lustre dentry revalidation. Change it to always use the
5218 Frequency : rare, unless heavy write-truncate concurrency is continuous
5219 Bugzilla : 4180, 6984, 7171, 9963, 9331
5220 Description: OST becomes very slow and/or deadlocked during object unlink
5221 Details : filter_destroy() was holding onto the parent directory lock
5222 while truncating+unlinking objects. For very large objects this
5223 may block other threads for a long time and slow overall OST
5224 responsiveness. It may also be possible to get a lock ordering
5225 deadlock in this case, or run out of journal credits because of
5226 the combined truncate+unlink. Solution is to do object truncate
5227 first in one transaction without parent lock, and then do the
5228 final unlink in a new transaction with the parent lock. This
5229 reduces the lock hold time dramatically.
5232 Frequency : rare, 2.4 kernels only
5234 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5235 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5236 before unmounting their filesystem in order to avoid deadlock.
5237 The kernel_locked() function in 2.4 kernels only checks whether
5238 the kernel lock is held, not whether it is this process that is
5239 holding it as 2.6 kernels do.
5244 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5245 Details : The client ptlrpc code may be trying to reconnect to a down
5246 server before a previous connection attempt has timed out.
5247 Increase the reconnect interval to be longer than the connection
5248 timeout interval to avoid sending duplicate connections to
5252 Frequency : echo_client brw_test command
5254 Description: fix echo_client to work with OST preallocated code
5255 Details : OST preallocation code (5137) didn't take echo_client IO path
5256 into account: echo_client calls filter methods outside of any
5257 OST thread and, hence, there is no per-thread preallocated
5258 pages and buffers to use. Solution: hijack pga pages for IO. As
5259 a byproduct, this avoids unnecessary data copying.
5263 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5264 Description: Client can oops in mdc_commit_close() after open replay
5265 Details : It was possible for the MDS to return an open request with no
5266 transaction number in mds_finish_transno() if the client was
5267 evicted, but without actually returning an error. Clients
5268 would later try to replay that open and may trip an assertion
5269 Simplify the client close codepath, and always return an error
5270 from the MDS in case the open is not successful.
5273 Frequency : rare, 2.6 OSTs only
5275 Description: OST may deadlock under high load on fragmented files
5276 Details : If there was a heavy load and highly-fragmented OST filesystems
5277 it was possible to have all the OST threads deadlock waiting on
5278 allocation of biovecs, because the biovecs were not released
5279 until the entire RPC IO was completed. Instead, release biovecs
5280 as soon as they are complete to ensure forward IO progress.
5282 Severity : enhancement
5284 Description: Support for specifying external journal device at mount
5285 Details : If an OST or MDS device is formatted with an external journal
5286 device, this device major/minor is stored in the ext3 superblock
5287 and may not be valid for failover. Allow detecting and
5288 specifying the external journal at mount time.
5293 Description: Mounting an MDS with pending unlinked files may cause oops
5294 Details : target_finish_recovery() calls mds_postrecov() which returned
5295 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5296 considers this an error and immediately begins cleaning up the
5297 lov, just after starting the mds_lov process
5299 Severity : enhancement
5301 Description: Implement 'lfs df' to report actual free space on per-OST basis
5302 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5303 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5304 usage of objects; '-h' to report in human readable format.
5306 ------------------------------------------------------------------------------
5308 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5315 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5316 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5317 at startup during orphan unlinking. Instead, skip buddy update
5318 before setup, as it will be regenerated after recovery is complete.
5323 Description: If an OST is inactive, its locks might reference stale inodes.
5324 Details : lov_change_cbdata() must iterate over all namespaces, even if
5325 they are inactive to clear inode references from the lock.
5327 Severity : enhancement
5328 Frequency : occasional, if non-standard max_dirty_mb used
5330 Description: Client will block write RPCs if not enough grant
5331 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5332 then the client will block writes while waiting for another RPC
5333 to complete instead of consuming its dirty limit. With change
5334 we get improved performance when max_dirty_mb is small.
5336 Severity : enhancement
5337 Bugzilla : 3389, 6253
5338 Description: Add support for supplementary groups on the MDS.
5339 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5340 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5341 lookups for user supplementary groups into a cache.
5345 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5346 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5347 for liblustre O_EXCL is enforced by MDS.
5351 Description: readdir never returns NULL in liblustre.
5352 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5353 dentry in struct dirent.
5357 Frequency : liblustre only, depends on application IO pattern
5358 Description: liblustre clients evicted if not contacting servers
5359 Details : Don't put liblustre clients into the ping_evictor list, so
5360 they will not be evicted by the pinger ever.
5362 Severity : enhancement
5364 Description: Add ability to evict clients by NID from MDS.
5365 Details : By echoing "nid:$NID" string into
5366 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5367 $NID would be instantly evicted from this MDS and from all active
5368 OSTs connected to it.
5372 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5373 Details : lookup_it_finish() used to query file size from OSTs that was not
5378 Description: service threads change working directory to that of init
5379 Details : Starting lustre service threads may pin the working directory
5380 of the parent thread, making that filesystem busy. Threads
5381 now change to the working directory of init to avoid this.
5385 Frequency : during shutdown only
5386 Description: shutdown with a failed MDS or OST can cause unmount to hang
5387 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5392 Frequency : 2.6 only
5393 Description: chmod/chown may include an extra supplementary group
5394 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5395 supplementary group and if none is specified this is used.
5398 Bugzilla : 5479 (6816)
5400 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5401 Details : If lookup is in progress on a file that is unlinked we might try
5402 to revalidate the inode and fail in revalidate after lookup is
5403 complete and ll_file_open() enqueues the open again but
5404 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5407 Frequency : always, if lconf --abort_recovery used
5409 Description: lconf --abort_recovery fails with 'Operation not supported'
5410 Details : lconf was attempting to abort recovery on the MDT device and not
5413 ------------------------------------------------------------------------------
5415 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5420 Frequency : rare (only unsupported configurations with a node running as an
5422 Bugzilla : 6514, 5137
5423 Description: Mounting a Lustre file system on a node running as an OST could
5425 Details : OSTs now preallocates memory needed to write out data at
5426 startup, instead of when needed, to avoid having to
5427 allocate memory in possibly low memory situations.
5428 Specifically, if the file system is mounted on on OST,
5429 memory pressure could force it to try to write out data,
5430 which it needed to allocate memory to do. Due to the low
5431 memory, it would be unable to do so and the node would
5432 become unresponsive.
5434 Severity : enhancement
5436 Description: Addition of lconf --service command line option
5437 Details : lconf now accepts a '--service <arg>' option, which is
5438 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5440 Severity : enhancement
5442 Description: Failover mode is now the default for OSTs.
5443 Details : By default, OSTs will now run in failover mode. To return to
5444 the old behaviour, add '--failout' to the lmc line for OSTs.
5446 Severity : enhancement
5448 Description: Health checks are now provided for MDS and OSTs
5449 Details : Additional detailed health check information on MSD and OSTs
5450 is now provided through the procfs health_check value.
5453 Frequency : occasional, depends on IO load
5455 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5456 after numerous create/delete cycles
5457 Details : The ext3 inode allocation policy would not allocate new inodes
5458 very well on the OSTs because there are no new directories
5459 being created. Instead we look for groups with free space if
5460 the parent directories are nearly full.
5465 Description: Network or server problems during mount may cause partially
5466 mounted clients instead of returning an error.
5467 Details : The config llog parsing code may overwrite the error return
5468 code during mount error handling, returning success instead
5474 Description: MDS can fail to allocate large reply buffers
5475 Details : After long uptimes the MDS can fail to allocate large reply
5476 buffers (e.g. zconf client mount config records) due to memory
5477 fragmentation or consumption by the buffer cache. Preallocate
5478 some large reply buffers so that these replies can be sent even
5479 under memory pressure.
5483 Frequency : rare (liblustre)
5484 Description: fsx running with liblustre complained that using truncate() to
5485 extend the file doesn't work. This patch corrects that issue.
5486 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5487 ATTR_SIZE and lsm use in llu_setattr_raw.
5491 Frequency : rare, only 2.6 kernels
5492 Description: Unusual file access patterns on the MDS may result in inode
5493 data being lost in very rare circumstances.
5494 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5495 code in the 2.6 kernel caused the inode and/or EA data not to
5496 be read from disk, causing single-file corruption.
5500 Frequency : rare, only 2.6 filesystems using extents
5501 Description: Heavy concurrent write and delete load may cause data corruption.
5502 Details : It was possible under high-load situations to have an extent
5503 metadata block in the block device cache from a just-unlinked
5504 file overwrite a newly-allocated data block. We now unmap any
5505 metadata buffers that alias just-allocated data blocks.
5509 Frequency : filesystems with default stripe_count larger than 77
5510 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5511 Details : lconf specifies an inode size of 4096 bytes when the default
5512 stripe_count is larger than 77. This conflicts with the default
5513 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5514 this case to avoid pinning too much memory for large EAs.
5516 ------------------------------------------------------------------------------
5518 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5523 Frequency : rare (extremely heavy IO load with hundreds of clients)
5525 Description: Client is evicted, gets IO error writing to file
5526 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5527 caused clients to be evicted for AST timeouts. The fixes in
5528 bug 5192 mean we no longer need to have such short AST timeouts
5529 so ldlm_timeout has been increased.
5532 Frequency : occasional during --force or --failover shutdown under load
5533 Bugzilla : 5949, 4834
5534 Description: Server oops/LBUG if stopped with --force or --failover under load
5535 Details : a collection of import/export refcount and cleanup ordering
5536 issues fixed for safer force cleanup
5539 Frequency : only filesystems larger than 120 OSTs
5540 Bugzilla : 5990, 6223
5541 Description: lfs getstripe would oops on a very large filesystem
5542 Details : lov_getconfig used kfree on vmalloc'd memory
5545 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5546 Bugzilla : 6242, 6243
5547 Description: reading from files that had been truncated to a non-zero size
5548 but never opened returned no data
5549 Details : ll_file_read() reads zeros from no-object files to EOF
5554 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5555 Details : There's little harm in aborting MDS/OSS recovery and letting it
5556 try again, so I removed the LASSERT and return an error instead.
5558 Severity : enhancement
5560 Description: New debugging infrastructure for tracking down data corruption
5561 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5562 (b) cover more of the client-side code path, and (c) try to narrow
5563 down where problems occurred
5567 Bugzilla : 3819, 4364, 4397, 6313
5568 Description: Racing close and eviction MDS could cause assertion in mds_close
5569 Details : It was possible to get multiple mfd references during close and
5570 client eviction, leading to one thread referencing a freed mfd.
5572 Severity: : enhancement
5573 Bugzilla : 3262, 6359
5574 Description: Attempts to reconnect to servers are now more aggressive.
5575 Details : This builds on the enhanced upcall-less recovery that was added
5576 in 1.4.2. When trying to reconnect to servers, clients will
5577 now try each server in the failover group every 10 seconds. By
5578 default, clients would previously try one server every 25 seconds.
5583 Description: After recovery, certain operations trigger a failed
5584 assertion on a client.
5585 Details : Failing over an mds, using lconf -d --failover, while a
5586 client was doing a readdir() call would cause the client to
5587 LBUG after recovery completed and the readdir() was resent.
5589 Severity : enhancement
5591 Description: Default groups are now added by lconf
5592 Details : You can now run lconf --group <servicename> without having to
5593 manually add groups with lmc.
5596 Frequency : occasional
5598 Description: Nodes with an elan id of 0 trigger a failed assertion
5601 Frequency : always when accessing e.g. tty/console device nodes
5603 Description: tty and some other devices nodes cannot be used on lustre
5604 Details : file's private_data field is used by device data and lustre
5605 values in there got lost. New field was added to struct file to
5606 store fs-specific private data.
5609 Frequency : when exporting Lustre via NFS
5611 Description: NFSD failed occasionally when looking up a path component
5612 Details : NFSD is looking up ".." which was broken in ext3 directories
5613 that had grown large enough to become hashed.
5616 Frequency : Clusters with multiple interfaces not on the same subnet
5618 Description: Nodes will repeatedly try to reconnect to an interface which it
5619 cannot reach and report an error to the log.
5620 Details : Extra peer list entries will be created by lconf with some peers
5621 unreachable. lconf now validates the peer before adding it.
5624 Frequency : Only if a default stripe is set on the filesystem root.
5626 Description: Setting a default stripe on the filesystem root prevented the
5627 filesystem from being remounted.
5628 Details : The client was sending extra request flags in the root getattr
5629 request and did not allocate a reply buffer for the dir EA.
5632 Frequency : occasional, higher if lots of files are accessed by one client
5633 Bugzilla : 6159, 6097
5634 Description: Client trips assertion regarding lsm mismatch/magic
5635 Details : While revalidating inodes the VFS looks up inodes with ifind()
5636 and in rare cases can find an inode that is being freed.
5637 The ll_test_inode() code will free the lsm during ifind()
5638 when it finds an existing inode and then the VFS later attaches
5639 this free lsm to a new inode.
5643 Bugzilla : 6422, 7030
5644 Description: MDS deadlock between mkdir and client eviction
5645 Details : Creating a new file via mkdir or mknod (starting a transaction
5646 and getting the ns lock) can deadlock with client eviction
5647 (gets ns lock and trying to finish a synchronous transaction).
5650 Frequency : occasional
5651 Description: While starting a server, the fsfilt_ext3 module could not be
5653 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5654 kernels. Previously, lconf would still use the ext3 name
5655 when trying to load modules. Now, it will correctly use
5656 ext3 on 2.4 and ldiskfs on 2.6.
5658 Severity : enhancement
5659 Description: The default stripe count has been changed to 1
5660 Details : The interpretation of the default stripe count (0, to lfs
5661 or lmc) has been changed to mean striping across a single
5662 OST, rather than all available. For general usage we have
5663 found a stripe count of 1 or 2 works best.
5665 Severity : enhancement
5666 Description: Add support for compiling against Cray portals.
5667 Details : Conditional compiling for some areas that are different
5671 Frequency : occasional
5672 Bugzilla : 6409, 6834
5673 Description: Creating files with an explicit stripe count may lead to
5674 a failed assertion on the MDS
5675 Details : If some OSTs are full or unavailable, creating files may
5676 trigger a failed assertion on the MDS. Now, Lustre will
5677 try to use other servers or return an error to the
5681 Frequency : occasional
5683 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5684 caused lock timeout during readahead (since 1.4.2).
5685 Details : Processes doing readahead might match a lock that hasn't been
5686 granted yet if there are overlapping and conflicting lock
5687 requests. The readahead process waits on ungranted lock
5688 (original lock is CBPENDING), while OST waits for that process
5689 to cancel CBPENDING read lock and eventually evicts client.
5691 Severity : enhancement
5693 Description: Initial enabling of flock support for clients
5694 Details : Implements fcntl advisory locking and file status functions.
5695 This feature is provided as an optional mount flag (default
5696 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5697 locking are implemented yet, and those that are are not guaranteed
5698 to be completely correct in production environments.
5699 mount -t lustre -o [flock|noflock] ...
5702 Frequency : occasional
5704 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5705 assertion in the ext3 JBD (journaling) layer.
5706 Details : The b_committed_data struct is protected by the big kernel lock
5707 in 2.4 kernels, serializing journal_commit_transaction() and
5708 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5709 access to this struct. In 2.6 kernels there is finer grained
5710 locking to improve SMP performance of the JBD layer.
5714 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5719 Description: merge and cleanup kernel patches.
5722 -----------------------------------------------------------------------------
5724 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5726 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5727 versions of Lustre. Please update all servers and clients to
5728 version 1.4.2 or later at the same time. You must also run
5729 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5730 to update the configuration logs.
5732 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5733 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5734 - swab llog records of type '0' so we get proper header size/idx (5861)
5735 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5736 - fix rename of one directory over another leaking an inode (5953)
5737 - avoid SetPageDirty on 2.6 (5981)
5738 - don't re-add just-being-destroyed locks to the waiting list (5653)
5739 - when creating new directories, inherit the parent's custom
5740 striping settings if present parent (3048)
5741 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5742 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5743 - don't hold client locks on temporary worklist from l_lru (5666)
5744 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5745 - automatically evict dead clients (5921)
5746 - Update file size properly in create+truncate+fstat case (6196)
5747 - Do not unhash mountpoint dentries, do not allow removal of
5749 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5750 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5751 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5752 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5753 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5754 - Correctly handle reads of files with no objects (6243)
5755 - lctl recover will also mark a device active if deactivate used (5933)
5757 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5758 - allow --write-conf on an MDS with different nettype than client (5619)
5759 - don't write config llogs to MDS for mounts not from that MDS (5617)
5760 - lconf should create multiple TCP connections from a client (5201)
5761 - init scripts are now turned off by default; run chkconfig --on
5762 lustre and chkconfig --on lustrefs to use them
5763 - upcalls are no longer needed for clients to recover to failover
5765 - add --abort-recovery option to lconf to abort recovery on device
5767 - add support for an arbitrary number of OSTs (3026)
5768 - Quota support protocol changes.
5769 - forward compatibility changes to wire structs (6007)
5770 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5771 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5772 - improved "lustre status" script
5773 - initialize blocksize for non-regular files (6062)
5774 - added --disable-server and --disable-client configure options (5782)
5775 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5776 - Vanilla 2.4.29 support
5777 - increase maximum number of obd devices to 520 (6242)
5778 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5779 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5780 SLES 9 2.6.5 kernels
5781 - limit stripes per file to 160 (the maximum EA size) (6093)
5783 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5786 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5787 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5788 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5789 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5790 - small ext3 extents cleanups and fixes (5733)
5791 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5792 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5793 - avoid needless client->OST connect, fix handle mismatch (5317)
5794 - fix DLM error path that led to out-of-sync client, long delays (5779)
5795 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5796 - fix several locking issues related to i_size (5492,5624,5654,5672)
5797 - don't move pending lock onto export if it is already evicted (5683)
5798 - fix kernel oops when creating .foo in unlinked directory (5548)
5799 - fix deadlock in obdfilter statistics vs. object create (5811)
5800 - use time_{before,after} to avoid timer jiffies wrap (5882)
5801 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5802 - Do not leak request if server was not able to process it (5154)
5803 - If mds_open unable to find parent dir, make that negative lookup(5154)
5804 - don't create new directories with extent-mapping (5909, 5936)
5806 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5807 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5808 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5809 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5810 - support for RHEL4 kernel on i686 (5773)
5811 - provide error messages when incompatible logs are encountered (5898)
5813 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5814 * version 1.4.0.10 (1.4.1 release candidate 1)
5816 - don't keep a lock reference when lock is not granted (4238)
5817 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5818 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5819 - drop import inflight refcount on signal_completed_replay error (5255)
5820 - unlock page after async write error during send (3677)
5821 - handle missing objects in filter_preprw_read properly (5265)
5822 - no transno return for symlink open, don't save no-trasno open (3440)
5823 - don't try to complete elan receive that already failed (4012)
5824 - free RPC server reply state on error (5406)
5825 - clean up thread from ptlrpc_start_thread() on error (5160)
5826 - readahead could read extra page into cache that wasn't ejected (5388)
5827 - prevent races in class_attach/setup/cleanup/detach (5260)
5828 - don't dereference de->d_inode after l_dput of de (5458)
5829 - use "int" for stripe value returned from lock_to_stripe (5544)
5830 - mballoc allocation and error-checking fixes in 2.6 (5504)
5831 - block device patches to fix I/O request sizes in 2.6 (5482)
5832 - look up hostnames for IB nals (5602)
5833 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5834 - don't start multiple acceptors for the same port (5277)
5835 - fix incorrect LASSERT in mds_getattr_name (5635)
5836 - export a proc file for general "ping" checking (5628)
5837 - fix "lfs check" to not block when the MDS is down (5628)
5839 - service request history (4965)
5840 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5841 - create an "evict_client" /proc entry on OSTs, like the MDS has
5842 - fix mount usage message, return errors per mount(8) (5168)
5843 - change grep [] to grep "[]" in tests so they work in more UMLs
5844 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5845 - remove ancient LOV_MAGIC_V0 EA support (5047)
5846 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5847 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5848 - allow readahead allocations to fail when low on memory (5383)
5849 - mmap locking landed again, after considerable improvement (2828)
5850 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5852 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5855 - send OST transaction number in read/write reply to free req (4966)
5856 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5857 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5858 - the watchdog thread now runs as interruptible (5246)
5859 - flock/lockf fixes (but it's still disabled, pending 5135)
5860 - don't use EXT3 constants in llite code (5094)
5861 - memory shortage at startup could cause assertion (5176)
5863 - reorganization of lov code
5864 - single portals codebase
5866 - add extents/mballoc support (5025)
5867 - direct I/O reads in the obdfilter (4048)
5868 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5870 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5873 - send OST transaction number in read/write reply to free req (4966)
5874 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5875 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5876 - the watchdog thread now runs as interruptible (5246)
5877 - handle missing objects in filter_preprw_read properly (5265)
5878 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5879 - drop import inflight refcount on signal_completed_replay error (5255)
5880 - unlock page after async write error during send (3677)
5881 - return original error code on reconstructed replies (3761)
5882 - no transno return for symlink open, don't save no-trasno open (3440)
5884 - add pid to ldlm debugging output (4922)
5885 - bump the watchdog timeouts -- we can't handle 30sec yet
5886 - extra debugging for orphan dentry/inode bug (5259)
5888 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5891 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5892 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5893 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5894 - fix NULL dereference in /proc/sys/portals/routes (4827)
5895 - allow failed mdc_close() operations to be interrupted (4561)
5896 - stop precreate on OST before MDS would time out on it (4778)
5897 - don't send partial-page writes before EOF from client (4410)
5898 - discard client grant for sub-page writes on large-page clients (4520)
5899 - don't free dentries not owned by NFS code, check generation (4806)
5900 - fix lsm leak if mds_create_objects() fails (4801)
5901 - limit debug_daemon file size, always print CERROR messages (4789)
5902 - use transno after validating reply (3892)
5903 - process timed out requests if import state changes (3754)
5904 - update mtime on OST during writes, return in glimpse (4829)
5905 - add mkfsoptions to LDAP (4679)
5906 - use ->max_readahead method instead of zapping global ra (5039)
5907 - don't interrupt __l_wait_event() during strace
5909 - add software watchdogs to catch hung threads quickly (4941)
5910 - make lustrefs init script start after nfs is mounted
5911 - fix CWARN/ERROR duplication (4930)
5912 - return async write errors to application if possible (2248)
5913 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5914 - print NAL number in %x format (4645)
5915 - update barely-supported suse-2.4.21-171 series (4842)
5916 - support for sles 9 %post scripts
5917 - support for building 2.6 kernel-source packages
5918 - support for sles km_* packages
5920 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5923 - ignore -ENOENT errors in osc_destroy (3639)
5924 - notify osc create thread that OSC is being cleaned up (4600)
5925 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5926 - reconstruct ost_handle() like mds_handle() (4657)
5927 - create a new thread to do import eviction to avoid deadlock (3969)
5928 - let lconf resolve symlinked-to devices (4629)
5929 - don't unlink "objects" from directory with default EA (4554)
5930 - hold socknal file ref over connect in case target is down (4394)
5931 - allow more than 32000 subdirectories in a single directory (3244)
5932 - fix blocks count for O_DIRECT writes (3751)
5933 - OST returns ENOSPC from object create when no space left (4539)
5934 - don't send truncate RPC if file size isn't changing (4410)
5935 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5936 - bind to privileged port in socknal and tcpnal (3689)
5938 - rate limit CERROR/CWARN console message to avoid overload (4519)
5939 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5940 - basic mmap support (3918)
5941 - kernel patch series update from b1_4 (4711)
5943 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5946 - avoid crash during MDS cleanup with OST shut down (2775)
5947 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5948 - don't use bad inodes on the MDS (3744)
5949 - dynamic object preallocation to improve recovery speed (4236)
5950 - don't hold spinlock over lock dumping or change debug flags (4401)
5951 - don't zero obd_dev when it is force cleaned (3651)
5952 - print grants to console if they go negative (4431)
5953 - "lctl deactivate" will stop automatic recovery attempts (3406)
5954 - look for existing locks in ldlm_handle_enqueue() (3764)
5955 - don't resolve lock handle twice in recovery avoiding race (4401)
5956 - revalidate should check working dir is a directory (4134)
5958 - don't always mark "slow" obdfilter messages as errors (4418)
5960 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5963 - don't close LustreDB during write_conf until it is done (3860)
5964 - fix typo in lconf for_each_profile (3821)
5965 - allow dumping logs from multiple threads at one time (3820)
5966 - don't allow multiple threads in OSC recovery (3812)
5967 - fix debug_size parameters (3864)
5968 - fix mds_postrecov to initialize import for llog ctxt (3121)
5969 - replace config semaphore with spinlock (3306)
5970 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5971 - don't allow enqueue to complete on a destroyed export (3822)
5972 - down write_lock before checking llog header bitmap (3825)
5973 - recover from lock replay timeout (3764)
5974 - up llog sem before sending rpc (3652)
5975 - reduce ns lock hold times when setting kms (3267)
5976 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5977 - fix NULL deref and obd_dev leak on setup error (3312)
5978 - replace some LBUG about llog ops with error handling (3841)
5979 - don't match INVALID dentries from d_lookup and spin (3784)
5980 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5981 - fix invalid assertion in ptlrpc_set_wait (3880)
5983 - add libwrap support for the TCP acceptor (3996)
5984 - add /proc/sys/portals/routes for non-root route listing (3994)
5985 - allow setting MDS UUID in .xml (2580)
5986 - print the stack of a process that LBUGs (4228)
5988 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5991 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5992 - make sure to unset replay flag from failed open requests (3440)
5993 - if default stripe count is 0, use OST count for inode size (3636)
5994 - update parent mtime/ctime on client for create/unlink (2611)
5995 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5996 - free recovery state on server during a forced cleanup (3571)
5997 - unregister_reply for resent reqs (3063)
5998 - loop back devices mounting and status check on 2.6 (3563)
5999 - fix resource-creation race that can provoke i_size == 0 (3513)
6000 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6001 - more debugging for page-accounting assertion (3746)
6002 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6003 - avoid deadlock after precreation failure (3758)
6004 - fix race and lock order deadlock in orphan handling (3450, 3750)
6005 - add validity checks when grabbing inodes from l_ast_data (3599)
6007 - add /proc/.../recovery_status to obdfilter (3428)
6008 - lightweight CDEBUG infrastructure, debug daemon (3668)
6009 - change default OSC RPC parameters to be better on small clusters
6010 - turn off OST read cache for files smaller than 32MB
6011 - install man pages and include them in rpms (3100)
6012 - add new init script for (un)mounting lustre filesystems (2593)
6013 - run chkconfig in %post for init scripts (3701)
6014 - drop scimac NAL (unmaintained)
6016 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6019 - clean kiobufs before and after use (3485)
6020 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6021 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6022 - change init script to fail more clearly if not run as root (1528)
6023 - allow clients to reconnect during replay (1742)
6024 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6025 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6026 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6027 - fix cond_resched (3554)
6028 - only evict unfinished clients after recovery (3515)
6029 - allow bulk resend, prevent data loss (3570)
6030 - dynamic ptlrpc request buffer allocation (2102)
6031 - don't allow unlinking open directory if it isn't empty (2904)
6032 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6033 - remove extraneous obd dereference causing LASSERT failure (3334)
6034 - don't use get_cycles() when creating temp. files on the mds (3156)
6035 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6036 - handle EEXIST for set-stripe, set proper directory name (3336)
6038 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6039 - fix ksocknal_fmb_callback() error messages (2918)
6041 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6044 - don't copy lvb into (possibly NULL) reply on error (2983)
6045 - don't deref dentry after dput, don't free lvb on error (2922)
6046 - use the kms to determine writeback rpc length (2947)
6047 - increment oti_logcookies when osc is inactive (2948)
6048 - update client's i_blocks count via lvb messages (2543)
6049 - handle intent open/close of special files properly (1557)
6050 - mount MDS with errors=remount-ro, like obdfilter (2009)
6051 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6052 - don't use cancelling-locks' kms values (2947)
6053 - use highest lock extent for kms, not last one (2925)
6054 - don't dereference ERR_PTR() dentry in error handling path (3107)
6055 - fix thread race in portals_debug_dumplog() (3122)
6056 - create lprocfs device entries at setup instead of at attach (1519)
6057 - common AST error handler, don't evict client on completion race (3145)
6058 - zero nameidata in detach_mnt in 2.6 (3118)
6059 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6060 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6061 - initialize RPC timeout timer earlier for 2.6 (3219)
6062 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6063 - print nal/nid for unknown nid (3258)
6064 - additional checks for oscc recovery before doing precreate (3284)
6065 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6066 - don't crash in mdc_close for bad permissions on open (3285)
6067 - zero i_rdev for non-device files (3147)
6068 - clear page->private before handing to FS, better assertion (3119)
6069 - tune the read pipeline (3236)
6070 - fix incorrect decref of invalidated dentry (2350)
6071 - provide read-ahead stats and refine rpc in flight stats (3328)
6072 - don't hold journal transaction open across create RPC (3313)
6073 - update atime on MDS at close time (3265)
6074 - close LDAP connection when recovering to avoid server load (3315)
6075 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6076 - don't leak open file on MDS after open resend (3325)
6077 - serialize filter_precreate and filter_destroy_precreated (3329)
6078 - loop device shouldn't call sync_dev() for nul device (3092)
6079 - clear page cache after eviction (2766)
6080 - resynchronize MDS->OST in background (2824)
6081 - refuse to mount the same filesystem twice on same mountpoint (3394)
6082 - allow llmount to create routes for mounting behind routers (3320)
6083 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6084 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6085 - fix rare problem with rename on htree directories (3417)
6087 - allow default OST striping configuration per directory (1414)
6088 - fix compilation for qswnal for 2.6 kernels (3125)
6089 - increase maximum number of MDS request buffers for large systems
6090 - change liblustreapi to be useful for external progs like lfsck (3098)
6091 - increase local configuration timeout for slow disks (3353)
6092 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6094 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6097 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6098 - don't overwrite extent policy data in reply if lock was blocked (2901)
6099 - drop filter export grants atomically with removal from device (2663)
6100 - del obd_self_export from work_list in class_disconnect_exports (2908)
6101 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6102 - swab reply message in mdc_close, other PPC fixes (2464)
6103 - fix destroying of named logs (2325)
6104 - overwrite old logs when running lconf --write_conf (2264)
6105 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6106 - fix race in target_handle_connect (2898)
6107 - mds_reint_create() should take same inode create lock (2926)
6108 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6109 - don't close files for self_export to avoid uninitialized obd (2936)
6110 - allow MDS with the same name as client node (2939)
6111 - hold dentry reference for closed log files for unlink (2325)
6112 - reserve space for all logs during transactions (2059)
6113 - don't evict page beyond end of stripe extent (2925)
6114 - don't oops on a deleted current working directory (2399)
6115 - handle hard links to targets without a parent properly (2517)
6116 - don't dereference NULL lock when racing during eviction (2867)
6117 - don't grow lock extents when lots of conflicting locks (2919)
6119 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6122 - account for cache space usage on clients to avoid data loss (974)
6123 - lfsck support in lustre kernel code (2349)
6124 - reduce journal credits needed for BRW writes (2370)
6125 - orphan handling to avoid losing space on client/server crashes
6126 - ptlrpcd can be blocked, stopping ALL progress (2477)
6127 - use lock value blocks to assist in proper KMS, faster stat (1021)
6128 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6129 - recovery for initial connections (2355)
6130 - fixes for mds_cleanup_orphans (1934)
6131 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6132 - block all file creations until orphan recovery completes (1901)
6133 - client remove rq_connection from request struct (2423)
6134 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6135 - recovery timer race (2670)
6136 - mdc_close recovey bug (2532)
6137 - ptlrpc cleanup bug (2710)
6138 - mds timeout on local locks (2588)
6139 - namespace lock held during RPCs (2431)
6140 - handle interrupted sync write properly (2503)
6141 - don't try to handle a message that hasn't been replied to (2699)
6142 - client assert failure during cleanup after abort recovery (2701)
6143 - leak mdc device after failed mount (2712)
6144 - ptlrpc_check_set allows timedout requests to complete (2714)
6145 - wait for inflight reqs when ptlrpcd finishes (2710)
6146 - make sure unregistered services are removed from the srv_list
6147 - reset bulk XID's when resending them (caught by 1138 test)
6148 - unregister_bulk after timeout
6149 - fix lconf error (2694)
6150 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6151 - readahead locks pages, leaves pending causing memory pressure (2673)
6152 - increase OST request buffers to 4096 on large machines (2729)
6153 - fix up permission of existing directories in simple_mkdir (2661)
6154 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6155 - don't assign transno to errored transactions (2742)
6156 - don't delete objects on OST if given a bogus objid from MDS (2751)
6157 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6158 - if rq_no_resend, then timeout request after recovery (2432)
6159 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6160 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6161 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6162 - MDS recovery shouldn't delete objects that it has given out (2730)
6163 - if enqueue arrives after completion, don't clobber LVB (2819)
6164 - don't unlock pages twice when trigger_group_io returns error (2814)
6165 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6166 - don't write pages to disk if there was an error (1450)
6167 - don't ping imports that have recovery disabled (2676)
6168 - take buffered bytes into account when balancing socknal conn (2817)
6169 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6170 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6171 - remove little-endian swabbing of llog records (1987)
6172 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6173 - retry reposting request buffers if they fail (1191)
6174 - grow extent at grant time to avoid granting a revoked lock (2809)
6175 - lock revoke doesn't evict page if covered by a second lock (2765)
6176 - disable VM readahead to avoid reading outside lock extents (2805)
6178 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6179 - updated kernel patches for hp-2.4.20 kernel (2681)
6181 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6184 - fix truncated write corruption (2366)
6185 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6187 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6188 - ensure len(uuid) < 37 in lmc (1171)
6189 - fix ia64 OOPS in llog_test (2255)
6190 - zero end of page at obdfilter for partial page writes (2648)
6191 - don't leave stale dentries around after renames (bug 2428)
6192 - fix timeouts when evicting a client with a single lock held (2642)
6193 - set deadline for the initial HELLO message to drain (2634)
6194 - print out dotted-quad IP addresses in the socknal (2302)
6196 - additional debugging for MDS client eviction problem (2443)
6197 - fix mkfsoptions support for osts (2603, 2604)
6199 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6202 - add series for the vanilla 2.6.0 kernel
6203 - add series for the vanilla 2.4.24 kernel
6204 - add series for a cray x86/64 UL kernel drop
6205 - fix xattr patches for the vanilla 2.4.19 series
6207 - generate true UUIDs in lmc (1171)
6208 - have portals stack dumping break in UML (2466)
6209 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6210 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6211 - wake callback waiting threads on client eviction (2460)
6212 - Add --ptldebug and --subsystem to lmc (1719)
6213 - update assertion to allow safe interrupt allocation
6214 - set rq_no_resend for cancel requests (2432)
6215 - recalculate ptlrpcd timeout after resend (2494)
6216 - call vfs_rmdir when removing pending directories (2368)
6217 - fix renaming a file to itself (2429)
6218 - lmc creates a default one-stripe lov (2454)
6219 - expand procfs space to handle large clusters (2326)
6220 - increase UML stack to avoid overflow
6221 - update lconf's list of debug and subsystem masks
6222 - fix lfs find --obd (2510)
6223 - /proc tunable for disabling filter read caching (2591)
6224 - stop rpm packages from altering slapd.conf (2301)
6225 - disable nagle in the socknal under 0conf (2578)
6226 - choose mds inode size based on stripe count (2572)
6227 - fix kernel-source rpm problems (2516)
6229 - add --disable-doc to avoid pdf generation (2421)
6230 - update documentation, tests, type-os, comments
6231 - avoid format warnings on ia64
6232 - remove the TOE NAL
6233 - tiny code cleanups by removing unused fields
6235 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6238 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6239 - remove the most blatant lies from BUILDING file (2371)
6240 - change default debug level to reasonable production setting
6241 - reduce client side cache size to reduce cache flush time
6242 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6243 - make TCP zerocopy and pinger support enabled by default (2476)
6244 - sync writes completed after process exits caused crashes (2319)
6245 - maintain correct mount count on the MDS (2356)
6246 - backout 1557, because 2316 wasn't really fixed
6247 - better file I/O statistics gathering in /proc
6248 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6249 - another kernel patch to fix zero-copy TCP function export
6250 - don't take duplicate lock when processing re-sent getattr (2420)
6251 - lctl uses obd_self_export instead of creating new conn (2353)
6252 - MDS/OST recovery case which requires object creation asserted (2425)
6253 - move lfs from /usr/sbin to /usr/bin in packages
6254 - fix race between mds_client_add and mds_client_free (2417)
6255 - use kmalloc instead of slabs in portals (2430)
6256 - don't create duplicate records when a failover MDS is present (2442)
6257 - remove unnecessary mount age check (2332)
6258 - don't remove directory inodes from locks prematurely (2451)
6259 - don't break if MDS service name is the same as hostname (2103)
6260 - fix races in client write RPC generation when cache full (2482)
6262 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6265 - remove now-unused request->rq_obd (278)
6266 - if an allocation fails, print out how much memory we've used (1933)
6267 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6268 - add the "configurable stack size" patch to most series files (1256)
6269 - ability to write large log records, for 100+ OST configs (2306)
6270 - fix NULL deref when filter_prep fails (2314)
6271 - fix operator precedence error in filter_sync
6272 - dynamic allocation of socknal TX descriptors (2315)
6273 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6274 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6275 - ensure that CWARN messages in Portals always get to the syslog
6276 - __init/__exit are not for prototype decls (ldlm_init/exit)
6277 - x86-64 compile warning fixes
6278 - fix gateway LMC keyword conflict (2318)
6279 - fix MDS lock inversions in getattr/reint paths (1844)
6280 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6281 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6282 - DLM race condition prevented some lock evictions (2328)
6283 - ENOMEM detection and retry on socknal sends (2230)
6284 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6285 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6286 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6287 - better error messages when a client is rejected during recovery (1505)
6288 - avoid cancelling locks which were never granted, after failure (2330)
6289 - fix i_sem/journal inversion in mds_client_add (2333)
6290 - fix truncate/getattr lock cycle deadlock (2334)
6291 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6292 - fix two rare exit paths which could leak an l_lock() ref (2321)
6293 - fencepost error in MDS/OST orphan recovery (2226)
6294 - make log record alignment 8 bytes (1988)
6295 - lstripe now fails when requested offset > ost_count (2237)
6296 - ensure that all kernel series have a complete list.h (1607)
6297 - fix crashes in special-file operations (2316)
6298 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6300 - allow configurable automake binary, for testing new versions
6301 - small update to the lfs documentation
6303 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6305 * fix negative export reference count in fsfilt_sync (2312)
6307 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6308 * release candidate 0.9.1
6310 - orphans are moved into the PENDING directory for possible recovery
6311 - replayed opens now open by fid for orphan/rename safety (1042)
6312 - last close of an orphan inode generates a transno (683)
6313 - chdir() and mount() now pin the directory entry (1020)
6314 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6315 - discard very old requests without processing them (1502)
6316 - remove some common, well-understood CERRORs (1505)
6317 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6318 - clear "grant" flags in OST replies until OST grant code lands (1644)
6319 - fix read performance by not clobbering i_blksize on client (1598)
6320 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6321 - make LRU size a /proc tunable, clears locks when reduced (707)
6322 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6323 - server locks take references on exports now (1558)
6324 - build fixes for 2.4.20-rh trees (1663)
6325 - return an error from lov_create if all OSCs are inactive (1751)
6326 - fix import levels when a reconnect happens without a timeout (1597)
6327 - exit early from mds_open if we get a lookup error (1749)
6328 - partial page read at EOF wouldn't wait for disk before sending (1642)
6329 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6330 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6331 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6332 - write/truncate lock inversion (1639)
6333 - Don't auto-load obdclass, portals modules during cleanup (1495)
6334 - fix timestamps from jumping to "now" (1763)
6335 - extra journal assertions (1648)
6336 - add an extra multiunlink test (1771)
6337 - fix read_record/write_record API (1776)
6338 - fix leak of offset_extent, possible incorrect i_size later (1772)
6339 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6340 - Debugging for the kqswnal_get_idle_tx problems (1820)
6341 - Allow recovery to be attempted multiple times (1536)
6342 - Write out MDS last_rcvd file after it is first created (1600)
6343 - Fix tx_descriptor leak in failed transmit situations (1827)
6344 - ext3 journaling fixes for assertion failure after IO error (1871)
6345 - class_export_put() on freed export after completion AST error (1896)
6346 - Fix revalidate looping in VFS (1322)
6347 - Don't access a freed export during MDS_REINT timeout (1521)
6348 - Add open-unlink recovery support on the MDS (1673,1764)
6349 - Return an error if no MDS data was read from last_rcvd (1946)
6350 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6351 - Don't setup a disk device that doesn't match exported UUID (317)
6352 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6353 - avoid committing NULL handle in force close
6354 - local.sh is now a one-stripe LOV configuration
6355 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6356 - fix ext3 htree duplicate directory entry corruption (1516)
6357 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6358 - update to new LOV EA format (2097)
6359 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6361 2003-06-15 Phil Schwan <phil@clusterfs.com>
6364 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6365 - per-import/export recovery handling (958, 931, 959)
6366 - multiple last-rcvd slots, for serving multiple FSes (949)
6367 - connections are again shared between multiple imp/exports (963, 964)
6368 - "umount -f" would hang if any requests needed to be sent (393, 978)
6369 - avoid pinning large req buffer by copying for queued messages (989)
6370 - add "uuid" to "lctl device" command to help upcalls (991)
6371 - "open" RPCs with transnos would confuse recovery counters (1037)
6372 - do proper endian conversion of all wire messages (288, 340, 891)
6373 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6374 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6375 - fix ptlrpc_connection leak in target_handle_connect (1174)
6376 - fix import refcounting bug in OST and MDS cleanup (1134)
6377 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6378 - fix very unlikely obd_types race condition (501)
6379 - remove osc_open hack for echo_client (1187)
6380 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6381 - a failure in read_inode2 leads to deadlock (1139)
6382 - cancel ack-locks as soon as transaction is committed (1072)
6383 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6384 - make sure to commitrw after any preprw to avoid deadlock (1162)
6385 - failing to execute a file in a lustre FS would lock inode (1203)
6386 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6387 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6388 - obd_uuid2tgt would walk off the end of the list (1255)
6389 - on IA64 the osc would give portals incorrect bulk size (1258)
6390 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6391 - fix lock inversion caused by new llite matching code (1282)
6392 - limit the number of dirty pages on a client to 10MB (1286)
6393 - timed out locks were not being corrected cancelled (1289)
6394 - fix O_DIRECT above 4GB on IA-32 (1292)
6395 * major user-visible changes
6396 - fail out/fail over policy now controlled by the upcall (993)
6398 - add OBD_PING to check server availability and failure (954)
6399 - lustre messages are now sent in sending host order (288, 340, 891)
6400 - add eadatalen to MDS getattr reply (340)
6401 - OST read replies may contain second buffer, with per-page status (593)
6403 2003-03-11 Phil Schwan <phil@clusterfs.com>
6406 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6407 - failed open()s could cause deadlock; fixed (867, 869)
6408 - stop cancelling OST locks when files are closed (481)
6409 - overlapping XID spaces caused network corruption (851, 853)
6410 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6411 - setattr_raw added, to do single-RPC, server-side setattrs
6412 - lmc/lconf syntax change for OST UUIDs
6413 - fix crashy race condition between ptlrpc_free_req and osc_close
6414 - don't use request in mdc_enqueue if we hit a timeout (889)
6415 - don't set the inode i_size for regular files from the MDS (896)
6416 - handle out of order completion AST (842)
6417 - don't LBUG if a lock request times out after receiving AST (913)
6418 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6419 - if a bad lock AST arrives, send an error instead of dropping entirely
6420 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6421 - fix leak in bulk IO when only partially completed (899, 900, 926)
6422 - fix O_DIRECT for ia64 (55)
6423 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6424 - C-z after timeout could hang a process forever; fixed (977)
6426 - client-side I/O cache (678, 924, 929, 941, 970)
6428 - READPAGE and SETATTRs which don't take server-side locks get
6431 2003-02-11 Phil Schwan <phil@clusterfs.com>
6434 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6435 - Fix fsfilt_extN_readpage() to read a full page of directory
6436 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6437 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6438 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6439 - Most intent operations were removed, in favour of a new RPC mode
6440 that does a single RPC to the server and bypasses most of the VFS
6441 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6442 - Aggressively cancel local locks on DLM servers
6443 - mds_reint_unlink sends EA to the client if it's the last nlink.
6444 client uses that EA to unlink OST objects.
6445 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6446 - recursive symlinks were fixed (439)
6447 - fixed NULL deref in DEBUG_REQ
6448 - filter_update_lastobjid no longer calls sync, which annoyed extN
6449 - fixed multi-client small-writes to a single file problem (445)
6450 - fixed mtime updates during file writes (607)
6451 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6452 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6453 - fixed Linux OST size reporting problem (444, 656)
6454 - OST now updates object mtime with writes or setattr (607, 619)
6455 - client verifies file size before zeroing page past EOF (445)
6456 - OST now writes last allocated objid to disk with allocation (108)
6457 - LOV on echo now works (409)
6459 - mds_reint_unlink sends a new buffer, with the EA included. this
6460 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6461 set if a regular file was being unlinked, and it was the last link
6462 - use PtlGet from the target for bulk writes (315)
6463 - OST now updates object mtime with writes or setattr (607, 619)
6464 - LDLM now has a grant-time callback to revalidate locked items, if
6466 - Many MDS operations were reorganized to combat race conditions
6468 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6469 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6471 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6474 - Fully reactivate OST imports after reconnection (512, others)
6475 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6476 - More graceful error handling for truncating on dead OST (515)
6477 - Don't error out unless we're actually accessing dead stripes (474)
6478 - Fix garbage sizes when stripes are missing (410)
6479 - LRU counters were broken, causing constant lock purge (433, 432)
6480 - garbage on read from stripes with failed OSTs (441)
6481 - mark OSCs as active before reconnecting during recovery (438)
6482 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6483 - lfind did not preserve OST order in output (443)
6484 - symlinks cause hung clients, incorrect data (439)
6485 - stop dereferencing request after dropping refcount (457)
6486 - don't LASSERT(spin_is_locked) on non-SMP (455)
6487 - fixes for many rename() bugs
6488 - fstat didn't correctly synchronize attributes (399)
6489 - server must handle lock cancellation during blocking AST prep (487)
6490 - bulk descriptors were free()d too soon (511)
6491 - fix paths in lconf, which would load incorrect modules (451, 507)
6492 - fix confusing lconf 'host not found' error message (386)
6493 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6494 - fix race condition in mdc_blocking_ast() for inode access (526)
6495 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6496 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6497 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6498 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6499 - fix lstripe in several ways (kernel side) (527)
6500 - fix request leak in ldlm_cli_enqueue (262)
6501 - incorrect OSC was marked inactive after OST failure
6502 - call mds_fs_cleanup before unmounting filesystem (524)
6503 - fix races between taking ns_lock and ldlm_lock_change_resource
6504 - fix races updating LOV export open file list
6505 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6506 - fix recovery NULL deref in ldlm_cli_cancel_unused
6507 - fix some DLM races by using new hash table for lock handles (419)
6508 - permit the client to specify desired inodes, at replay
6509 - duplicate requests when we queue them for replay reintegration
6510 - fix last_rcvd offset calculation
6511 - sync after each recovered transaction, so we always make progress
6512 - never, not always, ERESTART requests without transnos
6513 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6515 - skip replay if the MDS says that the client is already connected
6516 - don't check for a recovery-enabled export to match lctl's UUID
6517 - don't INC_USE_COUNT for phantom exports
6518 - don't crash when cleaning up phantom exports (567)
6519 - don't double-finish or set replay data for errored mdc_open requests
6520 - abort requests when they time out, so we don't get old replies
6521 - send/receive replies for AST messages again
6522 - if the client says that it doesn't have the lock, cancel it on the
6524 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6525 mark it as destroyed, it will all work out when decref is called
6526 - fix module use counts (22, 581)
6528 - ASTs now expect a reply (server cancels lock on error reply)
6530 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6533 - fix many simultaneous client startup (392)
6534 - fix dentry->d_it clobbering
6535 - credentials weren't being shipped for readdir/getattr operations
6536 - remove invalid assertions triggered during some concurrent MD
6538 - proper Lustre versions added (336, 389)
6539 - fix memory leak for create error case (398)
6540 - fix LOV locking bug that would get cli/srv out of sync
6541 - fix echo client over LOV (409)
6542 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6543 - fix double-O_EXCL intent crash (424)
6544 - avoid sending multiple lock CANCELs (352)
6546 - MDS can do multi-client recovery (modulo bugs in new code)
6548 - many updates, edits, cleanups
6550 2002-11-18 Phil Schwan <phil@clusterfs.com>
6553 - fix null d_it dereference (346)
6554 - fix full OST/dbench hang (333)
6555 - fix permission problem with file removal (286)
6556 - fix removal of OSCs from LOV when they fail
6557 - fix NULL deref during bulk timeout (214)
6558 - fix problems related to multiple filesystems on one MDS (241)
6559 - fixed serious subtle metadata locking bugs
6560 - free locks on clients when inodes are removed due to memory
6562 - fix inode pointer in lock data (285)
6563 - partial support for multiple MDS on a single host (241)
6564 - data locks weren't cancelled at clear_inode time (290, 311)
6565 - intent locks could lead to unbounded lock growth (205)
6566 - added a maximum lock count, an LRU list, and a flusher
6567 - fix multiple rename (365)
6568 - properly abstracted the echo client
6569 - OSC locked 1 byte too many; fixed
6570 - rewrote brw callback code:
6571 - fixed recovery bugs related to LOVs (306)
6572 - fixed too-many-pages-in-one-write crash (191)
6573 - fixed (again) crash in sync_io_timeout (214)
6574 - probably fixed callback-related race (385)
6576 - Add capability to MDS protocol
6577 - LDLM cancellations and callbacks on different portals
6579 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6582 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6583 - defer open object destruction to close time (601981, 138)
6584 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6585 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6586 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6587 - don't call obd_setattr in ll_file_release for destroyed objects
6588 * protocol change to lustre_msg: move |version| and add |flags|
6589 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6590 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6591 * added replay of create, unlink, link and rename operations during
6592 MDS failover; recovery should be much more robust now
6593 * remove failed OSCs from LOVs (only lov_create uses this so far)
6594 * the lustre-HOWTO was brought (more) up to date (582544)
6596 2002-10-23 Phil Schwan <phil@clusterfs.com>
6599 - in-use dentries weren't being reused properly (617851)
6600 - prevent multiple LDLM setup (599178)
6601 - fix LOV size calculations for truncate (617853)
6602 - fix client handling of MDS intent errors (POSIX)
6603 - fix permission bug in lovstripe.c test (624321)
6604 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6605 - truncate past end of file could corrupt data
6606 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6607 - a race in recovery could return ETIMEDOUT to apps (623947)
6608 - building outside the source directory was fixed
6609 * the lustre-HOWTO was brought (more) up to date (582544)
6610 * major progress was made on recovery functionality
6612 2002-10-10 Phil Schwan <phil@clusterfs.com>
6615 - recovery deadlock fix
6616 - rm -rf causes LBUG fix (617817)
6617 - file open by multiple tasks fix (618962)
6618 - directory permissions bugs (602707 and 620007)
6619 - journal_stop fixed with locking (611313)
6620 - O_APPEND failures resolved (618273, perhaps 614459)
6621 - lconf PATH fix (619770)
6622 - IA64 build fix (621450)
6623 - RPC buffer sizes scale with amount of memory
6625 2002-10-01 Phil Schwan <phil@clusterfs.com>
6628 - locks would be cancelled without throwing away data pages,
6629 resulting in inconsistent data (605627)
6630 - inode attributes were not always being refreshed (605627, 612449)
6631 - lconf now continues to cleanup after lctl reports an error
6632 - MDS now enforces user permissions (602707)
6633 - lprocfs cleanup fixed, but not yet enabled (614157)
6634 - fixed infinite server hang, should a client not respond to an AST
6635 - avoid going into recovery if user calls readlink() with a buffer
6636 that's too small (613941)
6637 - AST RPCs no longer require replies (614867) -- this may be changed
6638 - don't crash server if client sends an IOV that's too big (611336)
6639 - fixed lock conversion deadlock (611892)
6640 - fixed the following of symlinks (614622)
6641 * recovery: the server can remove locks from a client that dies, other
6642 clients can make progress
6643 * more extN patch fixes
6644 * compile-time configurable ptlrpc buffer allocations
6646 - collaborative read cache document
6647 - Lustre Lite Performance CDR document-in-progress
6649 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6652 - fix typo in patch-2.4.18
6654 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6657 - clear ptlrpc request each time in handle_incoming_request()
6658 - unlink of files now destroys the object on the OST
6660 2002-09-19 Peter Braam <braam@clusterfs.com>
6662 * add hard link support
6663 * change obdfile creation method
6664 * kernel patch changed
6666 2002-09-19 Peter Braam <braam@clusterfs.com>
6669 - stack overflow bug in extN fixed
6671 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6673 * documentation updates
6674 - add man pages for config tools
6675 - update tests/README to describe testing with new config tools
6676 - finish metadata API descriptions
6677 * bug fixes and cleanups
6678 - statfs workaround for 16TB limit
6679 - LOV stripe allocation improved, can stripe on subset of OSTs
6680 - LOV file size/IO offset was wrong for files > 4GB in size
6681 - object EA data was being dropped, caused files to be unreadable
6682 - memory overflow with non-LOV OST caused memory corruption
6683 - fixed regression tests to work with new config tools, obdfilter
6684 - fixed bug when directory size became larger than 1 block
6685 - fixed bug (for single client case) when PWD was deleted
6686 - invalidate local directory pages when doing intent-based ops
6687 - avoid LDLM oops when lock callback contained bad data
6689 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6691 * documentation updates
6692 * bug fixes and cleanups
6693 - configuration tools
6696 - 64-bit compile warnings
6697 - 64-bit internal statfs data
6699 * test_brw on persistent OST devices
6701 * lprocfs (disabled)
6703 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6705 * documentation updates
6706 * bug fixes and cleanups
6707 * configuration tools
6709 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6712 * many small fixes to 0_5_4
6713 * io/network handling
6714 * thinkos in MDS operations
6716 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6719 * crucial basic fixes to 0.5.3
6720 * IOR, Iozone work over Elan
6723 2002-08-07 Phil Schwan <phil@clusterfs.com>
6724 * version 0_5_3, our first alpha
6725 * we use the new Portals iovs
6726 * documentation updates
6727 * bug fixes and cleanups
6728 * small changes in the DLM wire protocol
6730 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6731 * version 0_5_1 with some initial stability,
6732 * locking on MD and file I/O.
6733 * documentation updates
6734 * several bug fixes since 0.5.0
6735 * small changes in wire protocol
6737 2002-07-18 Phil Schwan <phil@clusterfs.com>
6739 * delivered as Lustre Light Alpha
6740 * fixed a crash after handling invalid MDS requests
6741 * fixed directory pages for architectures with non-4k pages sizes
6743 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6744 * release version v0_4_4
6745 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6747 * Fixes a number of interruption problems with OST operations.
6748 * Update documentation for portals header changes
6749 * Move all wire protocol structs/defines to lustre_idl.h
6750 * Fixes symlink length bug.
6751 * Add tcpdump to repository.
6753 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6754 * release version v0_4_3
6755 * Fixes statfs for inodes on extN.
6756 * Fixes bug in runtests which would delete /etc/hosts.
6757 * Use 64-bit object IDs wherever possible (not into VFS though)
6758 Remove ost_get_info, which is unused by lustre, and out of date.
6760 2002-07-03 Peter Braam <braam@clusterfs.com>
6761 * release version v0_4_2 Fixes a lookup error (type not passed)
6762 * move forward to head of Portals
6763 * move forward to latest Lustre kernel
6765 2002-06-25 Peter Braam <braam@clusterfs.com>
6766 * release version v0_4_1. Hopefully stable on single node use.