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 * Other server kernels known to build and work at some point (others may also work):
9 3.10.0-862.14.4.el7 (RHEL7.5)
10 3.10.0-957.27.2.el7 (RHEL7.6)
11 4.18.0-80.11.2.el8 (RHEL8.0)
12 4.18.0-147.8.1.el8 (RHEL8.1)
13 4.18.0-193.6.3.el8 (RHEL8.2)
14 4.4.120-92.70 (SLES12 SP2)
15 4.4.180-94.100 (SLES12 SP3)
16 4.4.0-131 (Ubuntu 16.04)
17 4.15.0-32 (Ubuntu 18.04)
18 vanilla linux 4.6.7 (ZFS only)
19 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
20 * Client primary kernels built and tested during release cycle:
21 3.10.0-1062.18.1.el7 (RHEL7.7)
22 3.10.0-1127.19.1.el7 (RHEL7.8)
23 4.12.14-95.48 (SLES12 SP4)
24 4.12.14-122.20 (SLES12 SP5)
25 4.12.14-197.37 (SLES15 SP1)
26 4.15.0-48 (Ubuntu 18.04)
27 * Other clients known to build on these kernels at some point (others may also work):
28 3.10.0-862.14.4.el7 (RHEL7.5)
29 3.10.0-957.27.2.el7 (RHEL7.6)
30 4.14.0-49.13.1.el7a (RHEL7.5)
31 4.14.0-115.2.2.el7a (RHEL7.6)
32 4.18.0-80.11.2.el8 (RHEL8.0)
33 4.18.0-147.8.1.el8 (RHEL8.1)
34 4.18.0-193.6.3.el8 (RHEL8.2)
35 4.4.120-92.70 (SLES12 SP2)
36 4.4.180-94.100 (SLES12 SP3)
37 4.4.0-131 (Ubuntu 16.04)
38 5.4.0-37 (Ubuntu 20.04)
39 5.4.0 (vanilla kernel.org)
40 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
41 * Recommended ZFS version: 0.8.4
42 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
43 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
44 stack overflow. For more information, please refer to bugzilla 17630.
45 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
46 the re-exporting nodes to be mounted with "32bitapi" mount option
48 --------------------------------------------------------------------------------
52 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
53 for currently supported client and server kernel versions.
54 * Server primary kernels built and tested during release cycle:
55 3.10.0-1062.1.1.el7 (RHEL7.7)
56 * Other server kernels known to build and work at some point (others may also work):
57 3.10.0-862.14.4.el7 (RHEL7.5)
58 3.10.0-957.27.2.el7 (RHEL7.6)
59 4.18.0-80.7.1.el8 (RHEL8.0)
60 4.4.120-92.70 (SLES12 SP2)
61 4.4.180-94.100 (SLES12 SP3)
62 4.4.0-131 (Ubuntu 16.04)
63 4.15.0-32 (Ubuntu 18.04)
64 vanilla linux 4.6.7 (ZFS only)
65 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
66 * Client primary unpatched kernels built and tested during release cycle:
67 3.10.0-1062.1.1.el7 (RHEL7.7)
68 4.12.14-95.29 (SLES12 SP4)
69 4.15.0-48 (Ubuntu 18.04)
70 * Other clients known to build on unpatched kernels at some point (others may also work):
71 3.10.0-862.14.4.el7 (RHEL7.5)
72 3.10.0-957.27.2.el7 (RHEL7.6)
73 4.14.0-49.13.1.el7a (RHEL7.5)
74 4.14.0-115.2.2.el7a (RHEL7.6)
75 4.18.0-80.7.1.el8 (RHEL8.0)
76 4.4.120-92.70 (SLES12 SP2)
77 4.4.180-94.100 (SLES12 SP3)
78 4.4.0-131 (Ubuntu 16.04)
79 vanilla linux 4.15.0 (ZFS only)
80 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
81 * Recommended ZFS / SPL version: 0.7.13
82 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
83 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
84 stack overflow. For more information, please refer to bugzilla 17630.
85 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
86 the re-exporting nodes to be mounted with "32bitapi" mount option
87 * Note for Ubuntu 14.04.5:
88 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
89 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
91 --------------------------------------------------------------------------------
95 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
96 for currently supported client and server kernel versions.
97 * Server known to build on the following kernels (others may also work):
98 2.6.32-696.18.7.el6 (RHEL6.9)
99 3.10.0-862.14.4.el7 (RHEL7.5)
100 3.10.0-957.el7 (RHEL7.6)
101 4.4.120-92.70 (SLES12 SP2)
102 4.4.162-94.69 (SLES12 SP3)
103 4.4.0-131 (Ubuntu 16.04)
104 4.15.0-32 (Ubuntu 18.04)
105 vanilla linux 4.6.7 (ZFS only)
106 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
107 * Client known to build on unpatched kernels (others may also work):
108 2.6.32-696.18.7.el6 (RHEL6.9)
109 2.6.32-754.6.3.el6 (RHEL6.10)
110 3.10.0-862.14.4.el7 (RHEL7.5)
111 3.10.0-957.el7 (RHEL7.6)
112 4.14.0-49.13.1.el7a (RHEL7.5)
113 4.14.0-115.2.2.el7a (RHEL7.6)
114 4.4.120-92.70 (SLES12 SP2)
115 4.4.162-94.69 (SLES12 SP3)
116 4.4.0-131 (Ubuntu 16.04)
117 4.15.0-32 (Ubuntu 18.04)
118 vanilla linux 4.15.0 (ZFS only)
119 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
120 * Recommended ZFS / SPL version: 0.7.9
121 * Tested with ZFS / SPL version: 0.7.9
122 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
123 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
124 stack overflow. For more information, please refer to bugzilla 17630.
125 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
126 the re-exporting nodes to be mounted with "32bitapi" mount option
127 * Note for Ubuntu 14.04.5:
128 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
129 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
131 --------------------------------------------------------------------------------
133 2018-04-03 Intel Corporation
135 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
136 for currently supported client and server kernel versions.
137 * Server known to build on patched kernels:
138 2.6.32-431.29.2.el6 (RHEL6.5)
139 2.6.32-504.30.3.el6 (RHEL6.6)
140 2.6.32-573.26.1.el6 (RHEL6.7)
141 2.6.32-642.15.1.el6 (RHEL6.8)
142 2.6.32-696.18.7.el6 (RHEL6.9)
143 3.10.0-693.21.1.el7 (RHEL7.4)
144 3.0.101-0.47.71 (SLES11 SP3)
145 3.0.101-107 (SLES11 SP4)
146 3.12.74-60.64.40 (SLES12 SP1)
147 4.4.114-92.64 (SLES12 SP2)
148 4.4.114-94.11 (SLES12 SP3)
149 3.13.0-101 (Ubuntu 14.04, ZFS only)
150 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
151 4.4.0-116 (Ubuntu 16.04)
152 vanilla linux 4.6.7 (ZFS only)
153 * Client known to build on unpatched kernels:
154 2.6.32-431.29.2.el6 (RHEL6.5)
155 2.6.32-504.30.3.el6 (RHEL6.6)
156 2.6.32-573.26.1.el6 (RHEL6.7)
157 2.6.32-642.15.1.el6 (RHEL6.8)
158 2.6.32-696.18.7.el6 (RHEL6.9)
159 3.10.0-693.21.1.el7 (RHEL7.4)
160 3.0.101-0.47.71 (SLES11 SP3)
161 3.0.101-107 (SLES11 SP4)
162 3.12.74-60.64.40 (SLES12 SP1)
163 4.4.114-92.64 (SLES12 SP2)
164 4.4.114-94.11 (SLES12 SP3)
165 3.13.0-101 (Ubuntu 14.04)
166 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
167 4.4.0-116 (Ubuntu 16.04)
169 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
170 * Recommended ZFS / SPL version: 0.7.6
171 * Tested with ZFS / SPL version: 0.7.6
172 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
173 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
174 stack overflow. For more information, please refer to bugzilla 17630.
175 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
176 the re-exporting nodes to be mounted with "32bitapi" mount option
177 * Note for Ubuntu 14.04.5:
178 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
179 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
181 --------------------------------------------------------------------------------
183 2017-07-05 Intel Corporation
185 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
186 for currently supported client and server kernel versions.
187 * Server known to build on patched kernels:
188 2.6.32-431.29.2.el6 (RHEL6.5)
189 2.6.32-504.30.3.el6 (RHEL6.6)
190 2.6.32-573.26.1.el6 (RHEL6.7)
191 2.6.32-642.15.1.el6 (RHEL6.8)
192 2.6.32-696.3.1.el6 (RHEL6.9)
193 3.10.0-514.21.1.el7 (RHEL7.3)
194 3.0.101-0.47.71 (SLES11 SP3)
195 3.0.101-100 (SLES11 SP4)
196 3.12.74-60.64.40 (SLES12 SP1)
197 4.4.59-92.17 (SLES12 SP2)
198 vanilla linux 4.6.7 (ZFS only)
199 * Client known to build on unpatched kernels:
200 2.6.32-431.29.2.el6 (RHEL6.5)
201 2.6.32-504.30.3.el6 (RHEL6.6)
202 2.6.32-573.26.1.el6 (RHEL6.7)
203 2.6.32-642.15.1.el6 (RHEL6.8)
204 2.6.32-696.3.1.el6 (RHEL6.9)
205 3.10.0-514.21.1.el7 (RHEL7.3)
206 3.0.101-0.47.71 (SLES11 SP3)
207 3.0.101-100 (SLES11 SP4)
208 3.12.74-60.64.40 (SLES12 SP1)
209 4.4.59-92.17 (SLES12 SP2)
211 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
212 * Recommended ZFS / SPL version: 0.6.5.9
213 * Tested with ZFS / SPL version: 0.6.5.9
214 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
215 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
216 stack overflow. For more information, please refer to bugzilla 17630.
217 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
218 the re-exporting nodes to be mounted with "32bitapi" mount option
220 --------------------------------------------------------------------------------
222 12-07-2016 Intel Corporation
224 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
225 for currently supported client and server kernel versions.
226 * Server known to build on patched kernels:
227 2.6.32-431.29.2.el6 (RHEL6.5)
228 2.6.32-504.30.3.el6 (RHEL6.6)
229 2.6.32-573.26.1.el6 (RHEL6.7)
230 2.6.32-642.11.1.el6 (RHEL6.8)
231 3.10.0-514.el7 (RHEL7.3)
232 3.0.101-0.47.71 (SLES11 SP3)
233 3.0.101-80 (SLES11 SP4)
234 3.12.62-60.64.8 (SLES12 SP1)
235 vanilla linux 4.5.7 (ZFS only)
236 * Client known to build on unpatched kernels:
237 2.6.32-431.29.2.el6 (RHEL6.5)
238 2.6.32-504.30.3.el6 (RHEL6.6)
239 2.6.32-573.26.1.el6 (RHEL6.7)
240 2.6.32-642.11.1.el6 (RHEL6.8)
241 3.10.0-514.el7 (RHEL7.3)
242 3.0.101-0.47.71 (SLES11 SP3)
243 3.0.101-80 (SLES11 SP4)
244 3.12.62-60.64.8 (SLES12 SP1)
246 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
247 * Recommended ZFS / SPL version: 0.6.5.8
248 * Tested with ZFS / SPL version: 0.6.5.8
249 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
250 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
251 stack overflow. For more information, please refer to bugzilla 17630.
252 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
253 the re-exporting nodes to be mounted with "32bitapi" mount option
255 --------------------------------------------------------------------------------
257 02-29-2016 Intel Corporation
259 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
260 for currently supported client and server kernel versions.
261 * Server known to build on patched kernels:
262 2.6.32-431.29.2.el6 (RHEL6.5)
263 2.6.32-504.30.3.el6 (RHEL6.6)
264 2.6.32-573.12.1.el6 (RHEL6.7)
265 3.10.0-327.3.1.el7 (RHEL7.2)
266 3.0.101-0.47.71 (SLES11 SP3)
267 3.0.101-68 (SLES11 SP4)
268 vanilla linux 4.2.1 (ZFS only)
269 * Client known to build on unpatched kernels:
270 2.6.32-431.29.2.el6 (RHEL6.5)
271 2.6.32-504.30.3.el6 (RHEL6.6)
272 2.6.32-573.12.1.el6 (RHEL6.7)
273 3.10.0-327.3.1.el7 (RHEL7.2)
274 3.0.101-0.47.71 (SLES11 SP3)
275 3.0.101-68 (SLES11 SP4)
278 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
279 * Recommended ZFS / SPL version: 0.6.4.2
280 * Tested with ZFS / SPL version: 0.6.4.2
281 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
282 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
283 stack overflow. For more information, please refer to bugzilla 17630.
284 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
285 the re-exporting nodes to be mounted with "32bitapi" mount option
287 --------------------------------------------------------------------------------
289 03-10-2015 Intel Corporation
291 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
292 for currently supported client and server kernel versions.
293 * Server known to build on patched kernels:
294 2.6.32-431.29.2.el6 (RHEL6.5)
295 2.6.32-504.8.1.el6 (RHEL6.6)
296 3.0.101-0.46 (SLES11 SP3)
297 * Client known to build on unpatched kernels:
298 2.6.32-431.29.2.el6 (RHEL6.5)
299 2.6.32-504.8.1.el6 (RHEL6.6)
300 3.10.0-123.20.1.el7 (RHEL7)
301 3.0.101-0.46 (SLES11 SP3)
302 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
303 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
304 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
305 stack overflow. For more information, please refer to bugzilla 17630.
306 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
307 the re-exporting nodes to be mounted with "32bitapi" mount option
309 Severity : enhancement
311 Description: control OST-index in IDIF via ROCOMPAT flag.
312 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
313 last_rcvd file. For new formatted OST device, it will be auto set;
314 for the case of upgrading from old OST device, you can enable it
315 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
316 enabled, for new created OST-object, its IDIF-in-LMA will contain
317 the OST-index; for the existing OST-object, the OSD will convert
318 old format IDIF as new format IDIF with OST-index stored in the
319 LMA EA when accessing such OST-object or via OI scrub. Once such
320 flag is enabled, it cannot be reverted back, so the system cannot
321 be downgraded to the orignal incompatible version.
323 --------------------------------------------------------------------------------
325 07-30-2014 Intel Corporation
327 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
328 for currently supported client and server kernel versions.
329 * Server known to build on patched kernels:
330 2.6.32-431.20.3.el6 (RHEL6.5)
331 3.0.101-0.31 (SLES11 SP3)
332 * Client known to build on unpatched kernels:
333 2.6.32-431.20.3.el6 (RHEL6.5)
334 3.0.101-0.31 (SLES11 SP3)
337 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
338 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
339 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
340 stack overflow. For more information, please refer to bugzilla 17630.
341 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
342 the re-exporting nodes to be mounted with "32bitapi" mount option
344 --------------------------------------------------------------------------------
346 10-31-2013 Intel Corporation
348 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
349 * Server builds for kernels:
350 2.6.32-358.18.1.el6 (RHEL6.4)
351 * Client builds for unpatched kernels:
352 2.6.32-358.18.1.el6 (RHEL6.4)
353 3.0.92-0.8 (SLES11 SP3)
355 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
356 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
357 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
358 stack overflow. For more information, please refer to bugzilla 17630.
359 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
360 the re-exporting nodes to be mounted with "32bitapi" mount option
362 --------------------------------------------------------------------------------
364 05-30-2013 Intel Corporation
366 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
367 * Server builds for kernels:
368 2.6.32-358.6.2.el6 (RHEL6.4)
369 * Client builds for unpatched kernels:
370 2.6.32-358.6.2.el6 (RHEL6.4)
371 3.0.74-0.6.6 (SLES11 SP2)
373 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
374 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
375 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
376 stack overflow. For more information, please refer to bugzilla 17630.
377 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
378 the re-exporting nodes to be mounted with "32bitapi" mount option
380 --------------------------------------------------------------------------------
382 9-30-2012 Whamcloud, Inc.
384 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
385 * Server support for kernels:
386 2.6.32-279.5.1.el6 (RHEL6)
387 * Client support for unpatched kernels:
388 2.6.32-279.5.1.el6 (RHEL6)
390 2.6.18-238.19.1.el5 (RHEL5)
391 2.6.16.60-0.69.1 (SLES10)
392 * Recommended e2fsprogs version: 1.41.90.wc4
393 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
394 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
395 stack overflow. For more information, please refer to bugzilla 17630.
396 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
397 the re-exporting nodes to be mounted with "32bitapi" mount option
399 Severity : enhancement
401 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
403 Severity : enhancement
405 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
407 Severity : enhancement
409 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
411 Severity : enhancement
413 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
415 Severity : enhancement
417 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
419 --------------------------------------------------------------------------------
421 04-30-2012 Whamcloud, Inc.
423 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
424 * Server support for kernels:
425 2.6.32-220.4.2.el6 (RHEL6)
426 * Client support for unpatched kernels:
427 2.6.32-220.4.2.el6 (RHEL6)
429 2.6.18-238.19.1.el5 (RHEL5)
430 2.6.16.60-0.69.1 (SLES10)
431 * Recommended e2fsprogs version: 1.41.90.wc4
432 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
433 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
434 stack overflow. For more information, please refer to bugzilla 17630.
435 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
436 the re-exporting nodes to be mounted with "32bitapi" mount option
438 Severity : enhancement
440 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
442 Severity : enhancement
444 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
446 Severity : enhancement
448 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
450 Severity : enhancement
452 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
454 --------------------------------------------------------------------------------
456 09-30-2011 Whamcloud, Inc.
458 * Server support for kernels:
459 2.6.18-238.19.1.el5 (RHEL5)
460 2.6.32-131.6.1.el6 (RHEL6)
461 * Client support for unpatched kernels:
462 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
463 2.6.18-238.19.1.el5 (RHEL5)
464 2.6.32-131.6.1.el6 (RHEL6)
465 2.6.29.4-167.fc11 (FC11)
466 2.6.16.60-0.69.1 (SLES10)
468 * Recommended e2fsprogs version: 1.41.90.wc3
469 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
470 Lustre filesystem with 4K stack may cause a stack overflow. For more
471 information, please refer to bugzilla 17630.
472 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
473 mounted with "32bitapi" mount option
475 Severity : enhancement
476 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
478 Severity : enhancement
479 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
481 Severity : enhancement
482 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
485 Frequency : only with SLES10
487 Description: Use OFED "KMP" provided by Novell
488 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
489 Lustre is now built against this package. That means you need to
490 install the ofed-kmp package from Novell for the patchless client.
491 Note that the ofed-kmp that Novell ships may not exactly match the
492 kernel version but should still be compatible.
494 Severity : enhancement
496 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
497 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
499 --------------------------------------------------------------------------------
501 2010-07-15 Oracle, Inc.
503 * Server support for kernels:
504 2.6.18-164.11.1.el5 (RHEL 5)
505 2.6.18-164.11.1.0.1.el5 (OEL 5)
506 * Client support for unpatched kernels:
507 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
508 2.6.18-164.11.1.el5 (RHEL 5),
509 2.6.18-164.11.1.0.1.el5 (OEL 5)
510 2.6.16.60-0.42.8 (SLES 10),
512 2.6.29.4-167.fc11 (FC11)
513 * Recommended e2fsprogs version: 1.41.10-sun2
515 Severity : enhancement
517 Description: Distribute lfs_migrate script with lustre
518 Details : The lfs_migrate mentioned in the manual is now included in the
519 lustre tree in lustre/scripts and will distributed in lustre RPMs.
521 Severity : enhancement
523 Description: Failover nid list added to import proc file.
525 Severity : enhancement
527 Description: "som_preview" mount option required to use SOM
528 Details : The clients will need to use som_preview mount option
532 Description: Lustre does not do 1MB IOs to HW RAID
533 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
535 Severity : enhancement
537 Description: lfs quota output is not very convenient for awk/sed-parsing
538 Details : Some positions in lfs quota output table could be empty or
539 non-empty which made it hard to parse it with scripts, now
540 a dash is put instead of space where there is not supposed
545 Description: Update to OFED 1.5.1
546 Details : OFED 1.5.1 will be used in environments that don't use
547 vendor supplied OFED.
551 Description: Remove lnet. prefix from lctl params display
552 Details : Also renamed lnet.memused to lnet_memused to
557 Description: Use conf_param -d to remove permanant settings.
561 Description: Add list_param -R
562 Details : lctl list_param now supports recursive mode.
565 Bugzilla : 21829, 22312
566 Description: Fix llobdstat.
567 Details : 1. Now updating the write_bytes counter
568 2. Port script fixes from 1.8.
572 Description: decrease the usage of memory on clients.
573 Details : 1. On clients, recycle dentries and inodes unused.
574 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
579 Description: llobdstat fix and enhancement
580 Details : add a counter to set a limit to how many samples will be returned
581 fix a wildcard in the path to limit to obdfilter stats only
585 Description: lfs check only allowed for root
586 Details : Code cleanups and added a regresion test of non-root use of
591 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
592 Details : Fix hr->hr_index overload.
596 Description: Fix FIEMAP functionality.
600 Description: FC11 patchless client support added
604 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
605 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
609 Description: New acc-sm script and various test-framework enhancments.
613 Description: Previously landed patch was reverted.
617 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
621 Description: Add ldiskfs maxdirsize mount option.
625 Description: Fix mkfs.lustre support for 16TB LUNs.
629 Description: Print mmp_check_interval
630 Details : print mmp_check_interval and make it possible to abort
631 mount operation in case it takes too long.
635 Description: Add cascading_rw to lustre mpi tests.
639 Description: Store SOM data in EA.
640 Details : Also several SOM related cleanups.
644 Description: Add refcount to md_open_data
648 Description: Use refcount on lov_request_set to prevent memory corruption.
652 Description: Statahead: use parent dir's i_mutex for synchronization.
656 Description: loadgen improvements.
660 Description: Enable peer health detection in ksocklnd and ko2iblnd.
664 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
668 Description: Fix race on obd_devcie:md_stats
672 Description: Fix SLES 11 client builds.
673 Details : Resolved uninitalized pointer gcc warnings.
677 Description: Allow non-root access for "lfs check".
678 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
682 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
684 Severity : enhancement
686 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
687 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
691 Description: Sanity test_27p fails.
692 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
693 Sanity tests 27 were cleaned up and improved.
697 Description: Unify req format on client/servers
698 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
699 buffers and swabbers approach..
704 Description: drop join file stuff.
706 Severity : enhancement
708 Description: Add 2.6.27 and SLES11 (patchless client) support
712 Description: Fix in ptlrpc_expire_one_request() to print the signed time
715 Severity : enhancement
717 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
720 Severity : enhancement
722 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
724 Severity : enhancement
726 Description: correctly handle big reply message.
727 Details : send LNet event if reply is bigger than buffer and adjust this buffer
732 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
734 Severity : enhancement
736 Description: Add support for OEL5.
740 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
742 Severity : enhancement
744 Description: "weak-modules" support
745 Details : Implement "weak-modules" support which enables kernel modules
746 to be used with any kernel that implements the same kABI. In
747 order to achieve this modules are now installed in
748 /lib/modules/$(uname -r)/updates/kernel on all distributions.
752 Description: client could not reconnect to OST because of an active request.
753 Details : The client now retries to reconnect to the same server, if a
754 connect request failed with EBUSY or -EAGAIN. Test case 35b
755 conf-sanity.sh has been created.
759 Description: Don't allow make backward step on assignin osc next id.
760 Details : race between allocation next id and ll_sync thread can be cause
761 of set wrong osc next id and can be kill valid ost objects.
765 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
769 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
773 Description: Error handling in osc_statfs_interpret() has been improved.
774 Details : Check in osc_statfs_interpret() for EBADR.
778 Description: Do not send statfs() requests to OSTs disabled by administrator.
779 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
781 Severity : enhancement
783 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
785 Severity : enhancement
787 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
789 Severity : enhancement
791 Description: Enhancement for lfs(1) command to use numeric uid/gid.
795 Description: Improved cleanup procedure for conf-sanity test_46a.
799 Description: Do not start bulk transfer, if client tried to reconnect.
800 Details : Do not start bulk transfer, if client tried to reconnect,
801 but failed due to active rpcs. Check for exp_abort_active_req
802 has been implemented.
806 Description: Prevent inconsistences between linux and lustre mount structures.
807 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
808 Make the sleep interruptible.
810 Severity : enhancement
812 Description: provide server to client comms path
813 Details : server to client communications path via new LDLM_SET_INFO rpc
817 Description: truncate starts GFP_FS allocation under transaction causing deadlock
818 Details : ldiskfs_truncate calls grab_cache_page which may start page
819 allocation under an open transaction. This may lead to
820 calling prune_icache with consequent lustre reentrance.
824 Description: send correctly lsm on open replay
825 Details : MDS is trust to LSM size on replay open, but client can set wrong size
830 Description: enable client interface failover
831 Details : When a child reconnects from another NID, properly update export
832 nid hash position and ldlm reverse import.
834 Severity : enhancement
836 Description: Communicate OST degraded/readonly state via statfs to MDS
837 Details : Flags in the statfs returned from OSTs indicate whether the
838 OST is in a degraded RAID state, or if the filesystem has
839 turned read-only after a filesystem error is detected.
844 Description: don't panic if EPROTO was hit when reading symlink
845 Details : correctly handling request reference in error cases.
847 Severity : enhancement
849 Description: Add LustreNetLink, a kernel-userspace communcation path.
851 Severity : enhancement
853 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
858 Description: access to llog context before init.
859 Details : move handling CATALOGS file at osc layer and forbid access to llog
863 Frequency : rare, only when client is writing to OST on local node
865 Description: Avoid deadlock for local client writes
866 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
867 memory freeing context. This allows OST threads to set the
868 PF_MEMALLOC flag on task structures in order to allocate memory
869 from reserved pools and complete IO.
870 Use GFP_HIGHUSER for OST allocations for non-local client writes,
871 so that the OST threads generate memory pressure and allow
872 inactive pages to be reclaimed.
874 Severity : enhancement
876 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
881 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
882 Details : move ldlm namespace creation in setup phase to avoid grab
883 _lprocfs_lock with cli_sem held.
887 Description: Temporarily disable grant shrink.
888 Details : Disable the feature for debugging.
891 Frequency : only when formatting test filesystems
893 Description: Unable to run several mkfs.lustre on loop devices at the same time
894 Details : mkfs.lustre returns error 256 on the concurrent loop devices
895 formatting. The solution is to proper handle the error.
897 Severity : enhancement
899 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
901 Severity : enhancement
903 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
908 Description: resolve race between obd_disconnect and class_disconnect_exports
909 Details : if obd_disconnect will be called to already disconnected export he
910 forget release one reference and osc module can't unloaded.
912 Severity : enhancement
914 Description: Allow tuning service thread via /proc
915 Details : For each service a new
916 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
917 created that can be used to set min/max thread counts, and get the
918 current number of running threads.
923 Descriptoin: don't return error if have partially created objects for file.
924 Details : lov_update_create_set uses set->set_success as index for created
925 objects, so if some requests will be failed, they will have hole
926 at end of array and qos_shrink_lsm for allocate correct lsm.
928 Severity : enhancement
930 Description: Update OFED support to 1.4.1
935 Description: Reduce small size read RPC
936 Details : Set read-ahead limit for every file and only do read-ahead when
937 available read-ahead pages are bigger than 1M to avoid small size
940 Severity : enhancement
942 Description: move AT tunable parameters for more consistent usage
943 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
945 Severity : enhancement
947 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
948 Details : allow skip disconnected ost for send statfs request and hide error
954 Description: fix racy locking of mballoc block bitmaps causing BUG
955 Details : The locking of the mballoc buddy bitmap and the in-memory
956 block bitmap was using two different spin locks in some
957 cases. This made it possible to incorrectly access the
958 mballoc bitmap while another process was modifying it,
959 causing a sanity assertion to fail. While no on-disk corruption
960 was reported, there was some risk of this happening.
963 Frequency : rare, on llog test 6
965 Descriptoin: don't allow connect to already connected import
966 Details : allowing connect to already connected import is hide connecting problem.
971 Descriptoin: allow kill process which wait statahead result
972 Details : for some reasons 'ls' can stick in waiting result from statahead,
973 in this case need way for kill this process.
975 Severity : enhancement
977 Description: Add state history info file, enhance import info file
978 Details : Track import connection state changes in a new osc/mdc proc file;
979 add overview-type data to the osc/mdc import proc file.
981 Severity : enhancement
983 Description: MDS create should not wait for statfs RPC while holding DLM lock.
985 Severity : enhancement
987 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
992 Descriptoin: OST grant too much space to client even there are not enough space.
993 Details : Client will shrink its grant cache to OST if there are no write
994 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
995 this grant cache if there are already not enough avaible space
996 (left_space < total_clients * 32M).
999 Frequency : start MDS on uncleanly shutdowned MDS device
1001 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1002 Details : stay in waiting mds<>ost recovery finished produce random bugs
1003 due race between two ll_sync thread for one lov target. send
1004 ACTIVATE event only if connect realy finished and import have
1008 Frequency : rare, connect and disconnect target at same time
1010 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1011 Details : don't call obd_disconnect under lov_lock. this long time
1012 operation and can block ptlrpcd which answer to connect request.
1017 Descriptoin: don't lose wakeup for imp_recovery_waitq
1018 Details : recover_import_no_retry or invalidate_import and import_close can
1019 both sleep on imp_recovery_waitq, but we was send only one wakeup
1023 Frequency : always with long access acl
1025 Descriptoin: mds can't pack reply with long acl.
1026 Details : mds don't control size of acl but they limited by reint/getattr
1029 Severity : enhancement
1031 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1033 Severity : enhancement
1035 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1038 Frequency : start MDS on uncleanly shutdowned MDS device
1040 Descriptoin: aborting recovery hang on MDS
1041 Details : don't throttle destroy RPCs for the MDT.
1044 Frequency : on remount
1046 Description: external journal device not working after the remount
1047 Details : clear dev_rdonly flag for external journal devices in
1053 Description: shutdown vs evict race
1054 Details : client_disconnect_export vs connect request race.
1055 if client will evicted at this time - we start invalidate
1056 thread without referece to import and import can be freed
1060 Frequency : rare, need acl's on inode.
1062 Description: client can't handle ost additional correctly
1063 Details : if ost was added after client connected to mds client can have
1064 hit lnet_try_match_md ... to big messages to wide striped files.
1065 in this case need teach client to handle config events about add
1066 lov target and update client max ea size at that event.
1068 Severity : enhancement
1070 Description: Changelogs
1071 Details : Changelogs are a lightweight mechanism to track filesystem
1072 metadata and namespace changes. The changelog is recorded
1073 permanently on the MDTs, and is periodically "consumed" / purged
1074 when records are no longer needed.
1076 Severity : enhancement
1078 Description: compact fld format with extents
1079 Details : Store range of seq rather than every seq in FLD. Seq
1080 controller update FLD rather than clients. In Case of CMD, mdt0
1081 has FLD, all other metadata server act as non persistent proxy
1082 for FLD queries and cache fld entries in fld cache.
1087 Description: don't skip ost target if they assigned to file
1088 Details : Drop slow OSCs if we can, but not for requested start idx.
1089 This means "if OSC is slow and it is not the requested
1090 start OST, then it can be skipped, otherwise skip it only
1091 if it is inactive/recovering/out-of-space.
1095 Description: more cleanup in mds_lov
1096 Details : not send LOV EA under replay, we can't know about they size at this
1097 time. Don't allow client connect to mds before any ost connected,
1098 for avoid problems with LOV EA size and returning EIO to client.
1100 Severity : enhancement
1102 Description: Interoperability at server side (Disk interoperability)
1104 Severity : enhancement
1106 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1108 Severity : enhancement
1110 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1112 Severity : enhancement
1114 Description: New client IO stack (CLIO).
1116 Severity : enhancement
1118 Description: Commit on sharing. Eliminate inter-client dependencies between
1119 uncommitted transactions by doing transaction commits.
1120 Thereby clients may recovery independently.
1123 Frequency : Create a symlink file with a very long name
1125 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1126 Details : If there is no extra space in the request for early cancels,
1127 ldlm_req_handles_avail() returns 0 instead of a negative value.
1129 Severity : enhancement
1131 Description: Add /proc entry for import status
1132 Details : The mdc, osc, and mgc import directories now have
1133 an import directory that contains useful import data for debugging
1134 connection problems.
1136 Severity : enhancement
1138 Description: Re-disable certain /proc logging
1139 Details : Enable and disable client's offset_stats, extents_stats and
1140 extents_stats_per_process stats logging on the fly.
1143 Frequency : Only on FC kernels 2.6.22+
1145 Description: oops in statahead
1146 Details : Do not drop reference count for the dentry from VFS when lookup,
1147 VFS will do that by itself.
1149 Severity : enhancement
1151 Description: Generic /proc file permissions
1152 Details : Set /Proc file permissions in a more generic way to enable non-
1153 root users operate on some /proc files.
1157 Description: Hitting mdc_commit_close() ASSERTION
1158 Details : Properly handle request reference release in
1159 ll_release_openhandle().
1163 Description: quota recovery deadlock during mds failover
1164 Details : This patch includes att18982, att18236, att18237 in bz14840.
1166 1. fix osts hang when mds does failover with quotaon
1167 2. prevent watchdog storm when osts threads wait for the
1172 Frequency : only patchless client
1173 Description: add workaround for race between add/remove dentry from hash
1175 Severity : enhancement
1177 Description: Allow OST glimpses to return PW locks
1181 Description: LBUG when llog conf file is full
1182 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1186 Description: Prevent import from entering FULL state when server in recovery
1190 Description: service mount cannot take device name with ":"
1191 Details : Only when device name contains ":/" will mount treat it as
1197 Description: replace ptlrpcd with the statahead thread to interpret the async
1198 statahead RPC callback
1202 Frequency : on recovery
1203 Description: I/O failures after umount during fail back
1204 Details : if client reconnected to restarted server we need join to recovery
1205 instead of find server handler is changed and process self
1206 eviction with cancel all locks.
1208 Severity : enhancement
1210 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1212 Severity : enhancement
1214 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1216 Severity : enhancement
1218 Description: Upcall on Lustre log has been dumped
1219 Details : Allow for a user mode script to be called once a Lustre log has
1220 been dumped. It passes the filename of the dumped log to the
1221 script, the location of the script can be specified via
1222 /proc/sys/lnet/debug_log_upcall.
1227 Description: avoid idr_remove called for id which is not allocated.
1228 Details : Move assigment s_dev for clustered nfs to end of initialization,
1229 to avoid problem with error handling.
1234 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1235 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1236 into CONN_USED_HASH and this prodice warning when put connection
1237 again in unused hash.
1240 Severity : enhancement
1242 Description: File striping can now be set to use an arbitrary pool of OSTs.
1244 Severity : enhancement
1246 Description: Export bytes_read/bytes_write count on OSC/OST.
1250 Description: Early reply size mismatch, MGC loses connection
1251 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1252 the connect flags are properly negotiated.
1257 Description: quotas are not honored with O_DIRECT
1258 Details : all writes with the flag O_DIRECT will use grants which leads to
1259 this problem. Now using OBD_BRW_SYNC to guard this.
1263 Description: add quota statistics
1264 Details : 1. sort out quota proc entries and proc code.
1265 2. add quota statistics
1267 Severity : enhancement
1269 Description: enable quota support for HEAD.
1273 Description: Properly propagate oinfo flags from lov to osc for statfs
1274 Details : restore missing copy oi_flags to lov requests.
1276 Severity : enhancement
1278 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1279 : l_getgroups(8), lst(8), routerstat(8)
1280 Details : included man pages for llobdstat(8), llstat(8),
1281 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1283 Severity : enhancement
1285 Description: configure's --enable-quota should check the
1286 : kernel .config for CONFIG_QUOTA
1287 Details : configure is terminated if --enable-quota is passed but
1288 : no quota support is in kernel
1292 Description: Remove portals compatibility
1293 Details : Remove portals compatibility, not interoperable with releases
1298 Description: Resolve device initialization race
1299 Details : Prevent proc handler from accessing devices added to the
1300 obd_devs array but yet be intialized.
1302 Severity : enhancement
1304 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1306 Severity : enhancement
1308 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1313 Description: Using wrong pointer in osc_brw_prep_request
1314 Details : Access to array[-1] can produce panic if kernel compiled with
1315 CONFIG_PAGE_ALLOC enabled
1319 Description: Client runs out of low memory
1320 Details : Consider only lowmem when counting initial number of llap pages
1324 Description: *optional* service tags registration
1325 Details : if the "service tags" package is installed on a Lustre node
1326 When the filesystem is mounted, a local-node service tag will
1327 be created. See http://inventory.sun.com/ for more information
1328 about the Service Tags asset management system.
1332 Description: Kernel BUG tries to release flock
1333 Details : Lustre does not destroy flock lock before last reference goes
1334 away. So always drop flock locks when client is evicted and
1335 perform unlock regardless of successfulness of speaking to MDS.
1339 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1343 Description: sanity test 65a fails if stripecount of -1 is set
1344 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1349 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1350 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1353 Severity : enhancement
1355 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1359 Description: do not process already freed flock
1360 Details : flock can possibly be freed by another thread before it reaches
1361 to ldlm_flock_completion_ast.
1365 Description: LBUG during stress test
1366 Details : Need properly lock accesses the flock deadlock detection list.
1370 Description: oops in page fault handler
1371 Details : kernel page fault handler can return two special 'pages' in error
1372 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1376 Description: timeout with invalidate import.
1377 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1378 should be handled by ptlrpcd. This produce long age waiting and
1379 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1381 Severity : enhancement
1383 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1387 Description: enable MGS and MDT services start separately
1388 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1389 a co-located MGS without starting the MGS, which is a complement
1390 to 'nosvc' mount option.
1394 Frequency : after recovery
1395 Description: precreate to many object's after del orphan.
1396 Details : del orphan st in oscc last_id == next_id and this triger growing
1397 count of precreated objects. Set flag LOW to skip increase count
1398 of precreated objects.
1402 Frequency : rare, on clear nid stats
1403 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1404 Details : when clean nid stats sometimes try destroy live entry,
1405 and this produce panic in free.
1409 Description: Stack overflow during MDS log replay
1410 ease stack pressure by using a thread dealing llog_process.
1414 Description: wait until IO finished before start new when do lock cancel.
1415 Details : VM protocol want old IO finished before start new, in this case
1416 need wait until PG_writeback is cleared until check dirty flag and
1417 call writepages in lock cancel callback.
1419 Severity : enhancement
1421 Description: using special macro for print time and cleanup in includes.
1425 Description: mds_mfd_close() ASSERTION(rc == 0)
1426 Details : In mds_mfd_close(), we need protect inode's writecount change
1427 within its orphan write semaphore to prevent possible races.
1431 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1435 Frequency : rare, on shutdown ost
1436 Description: don't hit live lock with umount ost.
1437 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1438 use shrink_dcache_sb instead.
1442 Description: don't panic with use echo client
1443 Details : echo client pass NULL as client nid pointer and this produce null
1444 pointer dereference.
1448 Description: fix build on ppc32
1449 Details : compile code with -m64 flag produce wrong object file for ppc32.
1453 Description: add message levels for liblustreapi
1457 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1458 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1459 will result in an -EAGAIN and in the exit call path we call
1460 alloc_rr() which will with increasing aggressiveness attempt to
1461 aquire precreated objects on the minimum number of required OSCs.
1465 Description: Use old size assignment to avoid deadlock
1466 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1467 the scheduling while holding a spinlock. We do not need locking
1468 for size in ll_update_inode() because size is only updated from
1469 the MDS for directories or files without objects, so there is no
1470 other place to do the update, and concurrent access to such inodes
1471 are protected by the inode lock.
1475 Description: resolve "_IOWR redefined" build error on SLES10
1479 Description: dump the memory debugging after all modules are unloaded to
1480 suppress false negative in conf_sanity test 39
1482 Severity : enhancement
1484 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1489 Description: fix SLES kernel versioning
1490 Details : the kernel version for our SLES 10 kernel did not include a "-"
1491 before the "smp" at the end. while this was not a problem in
1492 general, it did mean that software trying to use the kernel
1493 version to try to detect a vendor specific kernel would fail.
1494 this was most evident by the OFED build scripts.
1498 Description: Don't update lov_desc members until making sure they are valid
1499 Details : When updating lov_desc members via proc fs, need fix their
1500 validities before doing the real update.
1504 Description: don't put request into delay list while invalidate in flight.
1505 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1506 import in flight. this produce timeout for invalidate and sometimes
1507 can cause stale data.
1511 Frequency : on ppc only
1512 Description: not convert ost objects for directory because it's not exist.
1513 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1517 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1519 Severity : enhancement
1521 Description: Update to support 2.6.22.14 vanilla kernel.
1525 Frequency : rare, on recovery
1526 Description: read procfs can produce deadlock in some situation
1527 Details : Holding lprocfs lock which send rpc can produce block for destroy
1528 obd objects and this also block reconnect with -EALREADY. This
1529 isn't fix all lprocfs bugs - but make it rare.
1531 Severity : enhancement
1533 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1536 Frequency : frequent on X2 node
1538 Description: mdc_set_open_replay_data LBUG
1539 Details : Set replay data for requests that are eligible for replay.
1543 Description: lustre_mgs: operation 101 on unconnected MGS
1544 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1545 MGC, and late on MGC cannot successfully connect to MGS and a lot
1546 of the error messages complaining that MGS is not connected.
1550 Frequency : on network error
1551 Description: panic with double free request if network error
1552 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1553 but it's true only for synchronus enqueue, for async enqueue
1554 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1557 Severity : enhancement
1559 Description: client-side metadata stat-ahead during readdir(directory readahead)
1560 Details : perform client-side metadata stat-ahead when the client detects
1561 readdir and sequential stat of dir entries therein
1564 Frequency : on start mds
1566 Description: Implement get_info(last_id) in obdfilter.
1569 Frequency : occasional
1571 Description: Correctly check stale fid, not start epoch if ost not support SOM
1572 Details : open with flag O_CREATE need set old fid in op_fid3 because
1573 op_fid2 was overwritten with new generated fid, but mds can answer
1574 with one of these two fids and both is not stale. Setattr
1575 incorrectly started an epoch and assume will be called
1576 done_writing, but without SOM done_writing ever being called.
1579 Frequency : rare, depends on device drivers and load
1581 Description: MDS or OSS nodes crash due to stack overflow
1582 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1583 In some cases, in conjunction with device drivers that use a lot
1584 of stack the MDS (or possibly OSS) service threads could overflow
1585 the stack. One change which was identified to consume additional
1586 stack has been reworked to avoid the extra stack usage.
1589 Frequency : occasional
1591 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1592 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1593 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1594 oscc_has_objects and see OSC has no objests and call
1595 oscc_internal_create to resend create request.
1597 Severity : enhancement
1599 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1601 Severity : enhancement
1603 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1606 Frequency : very rare
1608 Description: Fix replay if there is an un-replied request and open
1609 Details : In some cases, older replay request will revert the
1610 mcd->mcd_last_xid on MDS which is used to record the client's
1611 latest sent request.
1613 Severity : enhancement
1615 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1617 Severity : enhancement
1619 Description: Add rhel5 support to HEAD.
1621 Serverity : enhancement
1623 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1628 Description: Don't allow skipping OSTs if index has been specified.
1629 Details : Don't allow skipping OSTs if index has been specified, make
1630 locking in internal create lots better.
1634 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1635 Details : access to bitfield in structure is always rounded to long
1636 and this produce problem with not atomic change any bit.
1640 Description: Lustre make rpms failed.
1641 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1642 builds Lustre rpms from tarball.
1645 Frequency : rare on shutdown ost
1647 Description: If llog cancel was not send before clean_exports phase, this can
1648 produce deadlock in llog code.
1649 Details : If llog thread has last reference to obd and call class_import_put
1650 this produce deadlock because llog_cleanup_commit_master wait when
1651 last llog_commit_thread exited, but this never success because was
1652 called from llog_commit_thread.
1656 Description: allow userland application know is lost one of stripes.
1657 Details : fill lvb_blocks with error code on ost and return it to
1658 application if error flag found.
1662 Description: NULL lov_tgts causing MDS oops
1663 Details : more safe checks for NULL lov_tgts for avoid oops.
1665 Severity : enhancement
1667 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1671 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1672 Details : If an OST is down the MDS will hang indefinitely in
1673 obd_statfs_rqset() waiting for the statfs data. While for
1674 MDS QOS usage of statfs, it should not stuck in waiting.
1676 Severity : enhancement
1678 Description: remote_acl support
1679 Details : Support ACL-based permission check for remote user.
1680 Support setfacl/getfacl for remote user with the utils
1681 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1682 the system "{s,g}etfacl" utils.
1684 Severity : enhancement
1686 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1688 Severity : enhancement
1690 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1694 Description: make mgs_setparam() handle fsname containing dash
1695 Details : fsname containing a dash does not work with lctl conf_param
1697 Severity : enhancement
1699 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1701 Severity : enhancement
1703 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1707 Frequency : rare, at shutdown
1708 Description: access already free / zero obd_namespace.
1709 Details : if client_disconnect_export was called without force flag set,
1710 and exist connect request in flight, this can produce access to
1711 NULL pointer (or already free pointer) when connect_interpret
1712 store ocd flags in obd_namespace.
1716 Frequency : only at startup
1717 Description: not alloc memory with spinlock held.
1718 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1719 if any spinlock held.
1721 Severity : enhancement
1723 Description: make lustre randomly fail allocating memory
1724 Details : Make lustre randomly failed allocating memory for testing purpose.
1726 Severity : enhancement
1728 Description: lost problems with lov objid file
1729 Details : Fixes some scability and access to not inited memory problems
1730 in work with lov objdid file.
1735 Description: lfs find does not continue on file error
1736 Details : Continue other files processing when a file/dir is absent.
1740 Description: Inconsistent usage of lustre_pack_reply()
1741 Details : Standardize the usage of lustre_pack_reply() such that it
1742 always generate a CERROR on failure.
1745 Frequency : occasional
1747 Description: MDS hang or stay in waiting lock
1748 Details : If client receive lock with CBPENDING flag ldlm need send lock
1749 cancel as separate rpc, to avoid situation when cancel request
1750 can't processed due all i/o threads stay in wait lock.
1754 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1758 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1762 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1766 Description: Update to RHEL4 latest kernel.
1768 Severity : enhancement
1770 Description: Build SLES10 patchless client fails
1771 Details : The configure was broken by run ./configure with
1772 --with-linux-obj=.... argument for patchless client. When the
1773 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1774 can't search header adequately. Use absolute path such as
1775 -I($LINUX)/include instead.
1779 Description: interrupt oig_wait produce painc on resend.
1780 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1781 private set, and this produce situation when rq_ptlrpcd_data not
1782 copyed to new allocated request and triggered LBUG on assert
1783 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1784 ptlrpcd set if request is changed and can be safety changed to use
1789 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1790 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1791 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1792 these macros and replace them with their difinitions instead.
1796 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1797 Details : Update which_patch & target file for SLES10 latest kernel.
1800 Description: Remove client patches from SLES 10 kernel.
1801 Details : This causes SLES 10 clients to behave as patchless clients
1802 even on a Lustre-patched (server) kernel.
1804 Severity : enhancement
1806 Description: self-adjustable client's lru lists
1807 Details : use adaptive algorithm for managing client cached locks lru
1808 lists according to current server load, other client's work
1809 pattern, memory activities, etc. Both, server and client
1810 side namespaces provide number of proc tunables for controlling
1813 Severity : enhancement
1815 Description: light-weight GSS support
1816 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1817 and reduce performance overhead.
1819 Severity : enhancement
1821 Description: Linux keyring support
1822 Details : Support using service of Linux keyring for Lustre GSS internal
1823 context refresh/cache mechanism.
1827 Description: Fix errors in lfs documentation
1828 Details : Fixes man pages
1832 Description: don't use GFP_* in generic Lustre code.
1833 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1837 Description: obdclass is limited by single OBD_ALLOC(idarray)
1838 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1842 Description: warnings with build patchless client with vanila 2.6.19 and up
1843 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1844 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1848 Frequency : Only for SLES
1849 Description: sanity_quota fail test_1
1850 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1851 file, which makes TSTID[2] inunique.
1855 Frequency : Only for SLES9
1856 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1857 Details : Update target/ChangeLog/which_patch .
1861 Frequency : Only for test_10 in sanity-quota.sh
1862 Description: a bug in quota test code
1863 Details : checking a test flag in a wrong place
1867 Frequency : Only for test_13 in sanity-quota.sh
1868 Description: a bug in quota test script
1869 Details : shouldn't check overfull allocated quota
1873 Description: jbd statistics
1874 Details : Port older jbd statistics patch for sles10
1876 Severity : enhancement
1878 Description: Add printing inode info into message about error in writepage.
1883 Description: reply_lock_interpret crash due to race with it and lock cancel.
1884 Details : Do not replay locks that are being cancelled. Do not reference
1885 locks by their address during replay, just by their handle.
1889 Frequency : When flocks are used.
1890 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1891 Details : Flock locks might destroy just granted lock if it could be merged
1892 with another existing flock, this is done in completion handler,
1893 so teach ldlm_cli_enquque_fini that this is a valid case for
1899 Description: Oops in read and write path when failing to allocate lock.
1900 Details : Check if lock allocation failed and return error back.
1904 Description: lstripe command fails for valid OST index
1905 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1906 instead of lov->desc.ld_active_tgt_count.
1910 Description: sanity.sh failed test 103
1911 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1916 Description: df doesn't work properly if diskfs blocksize != 4K
1917 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1921 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1922 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1924 Severity : enhancement
1926 Description: Adaptive timeouts
1927 Details : RPC timeouts adapt to changing server load and network
1928 conditions to reduce resend attempts and improve recovery time.
1932 Description: llapi_file_create() does not allow some changes
1933 Details : add llapi_file_open() that allows specifying the mode and
1934 open flags, and also returns an open file handle.
1938 Description: merge and cleanup kernel patches.
1939 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1943 Description: Add journal checksum support.(Kernel part)
1944 Details : The journal checksum feature adds two new flags i.e
1945 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1946 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1947 indicates that the commit block contains the checksum for
1948 the blocks described by the descriptor blocks. Now commit
1949 record can be sent to disk without waiting for descriptor
1950 blocks to be written to disk. This behavior is controlled
1951 using JBD2_FEATURE_ASYNC_COMMIT flag.
1955 Description: OSS needs mutliple precreate threads
1956 Details : Add ability to start more than one create thread per OSS.
1960 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1961 Details : execute lfs setstripe on client
1965 Description: mds_obd_create error creating tmp object
1966 Details : When the user sets quota on root, llog will be affected and can't
1967 create files and write files.
1970 Frequency : Always on ia64 patchless client, and possibly others.
1972 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1973 Details : This allows the patchless client to be loaded on architectures
1974 without this export.
1978 Description: RedHat Update kernel for RHEL5
1979 Details : Add the kernel config file for RHEL5.
1983 Description: RedHat Update kernel for RHEL5
1984 Details : Modify the kernel config file more closer RHEL5.
1988 Description: Build failure against Centos5 (RHEL5)
1989 Details : Define PAGE_SIZE when it isn't present.
1993 Description: test_7 on sanity-quota.sh can't be used on separate machines
1994 Details : add facet to handle it
1998 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1999 Details : with NFS, the anon dentry's parent was set to itself in
2000 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2001 corresponding dentry other than use rec->ur_name.
2003 Severity : enhancement
2005 Description: lfs setstripe enhancement
2006 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2010 Description: enable data checksumming by default
2011 Details : enable checksum by default, allow --disable-checksum
2012 configure option and "-o nochecksum" mount option. Checksums
2013 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2017 Description: lustre support for RHEL5
2018 Details : Add support for RHEL5.
2022 Description: Client eviction due to failover config
2023 Details : after a connection loss, the lustre client should attempt to
2024 reconnect to the last active server first before trying the
2025 other potential connections.
2029 Description: when mds and osts use different quota unit(32bit and 64bit),
2030 quota will be released repeatly.
2031 Details : void sending multiple quota reqs to mds, which will keep the
2032 status between the reqs.
2036 Description: osts not allocated evenly to files
2037 Details : change the condition to increase offset_idx
2040 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2042 Description: Data corruption for OSTs that are formatted larger than 2TB
2044 Details : When generating the bio request for lustre file writes the
2045 sector number would overflow a temporary variable before being
2046 used for the IO. The data reads correctly from Lustre (which
2047 will overflow in a similar manner) but other file data or
2048 filesystem metadata may be corrupted in some cases.
2052 Description: Tune the kernel for good SCSI performance.
2053 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2054 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2059 Description: rewrite ext2-derived code in obdclass/uuid.c
2060 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2061 from scratch preserving functionality.
2065 Description: "lfs find -obd UUID" prints directories
2066 Details : "lfs find -obd UUID" will return all directory names instead
2067 of just file names. It is incorrect because the directories
2068 do not reside on the OSTs.
2072 Description: use i_size_read and i_size_write in 2.6 port
2073 Details : replace inode->i_size access with i_size_read/write()
2076 Frequency : when removing large files
2078 Description: scheduling issue during removal of large Lustre files
2079 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2080 It causes scheduling issues when removing large files (17TB in the
2084 Frequency : only with liblustre clients on XT3
2086 Description: evictions taking too long
2087 Details : allow llrd to evict clients directly on OSTs
2090 Frequency : only on ppc
2092 Description: /proc/fs/lustre/devices broken on ppc
2093 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2094 We should make sure the type of 'index' is loff_t and then cast
2095 explicitly as needed below. Do not assign an explicitly cast
2099 Frequency : only for rhel5
2101 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2102 Details : Modify the target file & which_kernel.
2104 Severity : enhancement
2106 Description: omit set fsid for export NFS
2107 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2108 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2112 Frequency : Always, for kernels after 2.6.16
2113 Description: Fix warning idr_remove called for id=.. which is not allocated.
2114 Details : Last kernels save old s_dev before kill super and not allow
2115 to restore from callback - restore it before call kill_anon_super.
2119 Description: buffer overruns could theoretically occur
2120 Details : llapi_semantic_traverse() modifies the "path" argument by
2121 appending values to the end of the origin string, and a
2122 overrun may occur. Adding buffer overrun check in liblustreapi.
2126 Description: Fix error on 'ls .' at the top of the Lustre mount.
2127 Details : Don't revalidate dentry if it is a root dentry.
2131 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2132 Details : Modify vm-tunables-rhel4.patch.
2136 Description: Kernel config for 2.6.18-vanilla.
2137 Details : Modify targets/2.6-vanilla.target.in.
2138 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2139 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2140 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2141 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2145 Description: improve handling recoverable errors
2146 Details : if request processig with error which can be recoverable on server
2147 request should be resend, otherwise page released from cache and
2152 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2153 Details : Modify target file & which_patch.
2154 A flaw was found in the IA32 system call emulation provided
2155 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2156 value could be stored in the %RAX register, which could trigger an
2157 out-of-bounds system call table access. An untrusted local user
2158 could exploit this flaw to run code in the kernel
2159 (ie a root privilege escalation). (CVE-2007-4573).
2163 Description: O_DIRECT bypasses client statistics.
2164 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2165 read_ahead_stats were updated. copied the stats section from
2166 osc_send_oap_rpc() into async_internal().
2170 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2174 Description: change order of libsysio includes
2175 Details : '#include sysio.h' should always come before '#include xtio.h'
2179 Description: handle "serious error: objid * already exists" more gracefully
2180 Details : If LAST_ID value on disk is smaller than the objects existing in
2181 the O/0/d* directories, it indicates disk corruption and causes an
2182 LBUG(). If the object is 0-length, then we should use the existing
2183 object. This will help to avoid a full fsck in most cases.
2185 Severity : enhancement
2187 Description: adapt the lustre_config script to support the upgrade case
2188 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2189 server targets to 1.6 in parallel.
2193 Description: To avoid grant space > avaible space when the disk is almost
2194 full. Without this patch you might see the error "grant XXXX >
2195 available" or some LBUG about grant, when the disk is almost
2197 Details : In filter_check_grant, for non_grant cache write, we should
2198 check the left space by if (*left > ungranted + bytes), instead
2199 of (*left > ungranted), because only we are sure the left space
2200 is enough for another "bytes", then the ungrant space should be
2201 increase. In client, we should update cl_avail_grant only there
2202 is OBD_MD_FLGRANT in the reply.
2206 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2207 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2208 described in CVE-2007-4573. This problem could allow untrusted
2209 local users to gain root access.
2212 Frequency : when using O_DIRECT and quotas
2214 Description: Incorrect file ownership on O_DIRECT output files
2215 Details : block usage reported by 'lfs quota' does not take into account
2216 files that have been written with O_DIRECT.
2221 Description: touch file failed when fs is not full
2222 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2223 otherwise we can get ENOSP while fs is not full.
2227 Description: parallel lock callbacks
2228 Details : Instead of sending blocking and completion callbacks as separated
2229 requests, adding them to a set and sending in parallel.
2232 Frequency : only for Cray XT3
2233 Bugzilla : 12829/13455
2234 Description: Changing primary group doesn't change the group lustre assigns to
2236 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2237 overridden with the primary group provided by the group upcall,
2238 whereas the supplied fsgid can be trusted if it is in the list of
2239 supplementary groups returned by the group upcall.
2241 Severity : enhancement
2243 Description: Allow masking D_WARNING, D_ERROR messages from console
2244 Details : Console messages can now be disabled via lnet.printk.
2248 Description: User code with malformed file open parameter crashes client node
2249 Details : Before packing join_file req, all the related reference should be
2250 checked carefully in case some malformed flags cause fake
2251 join_file req on client.
2255 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2256 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2257 mode is equal to the granted mode. However, it should also check
2258 that LDLM_FL_CP_REQD is not set.
2262 Description: Heavy nfs access might result in deadlocks
2263 Details : After ELC code landed, it is now improper to enqueue any mds
2264 locks under och_sem, because enqueue might want to decide to
2265 cancel open locks for same inode we are holding och_sem for.
2269 Description: Client eviction while running blogbench
2270 Details : A lot of unlink operations with concurrent I/O can lead to a
2271 deadlock causing evictions. To address the problem, the number of
2272 oustanding OST_DESTROY requests is now throttled to
2273 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2274 callbacks are processed in priority.
2278 Description: enable ACLs on MDS by default
2279 Details : ACLs must be enabled on MDS by default.
2282 Frequency : PPC/PPC64 only
2284 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2285 Details : fix duplicated definitions between asm-ppc64/types.h and
2286 lustre_types.h on PPC.
2289 Frequency : PPC/PPC64 only
2291 Description: asm-ppc/segment.h does not exist
2292 Details : fix compile issue on PPC.
2296 Description: better handle error messages in extents code
2299 Frequency : RHEL4 only
2301 Description: mkfs is very slow on IA64/RHEL4
2302 Details : A performance regression has been discovered in the MPT Fusion
2303 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2304 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2305 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2307 Severity : enhancement
2309 Description: SNMP support enhancement
2310 Details : Adding total number of sampled request for an MDS node in snmp
2313 Severity : enhancement
2315 Description: Optimize ldlm waiting list processing for PR extent locks
2316 Details : When processing waiting list for read extent lock and meeting read
2317 lock that is same or wider to it that is not contended, skip
2318 processing rest of the list and immediatelly return current
2319 status of conflictness, since we are guaranteed there are no
2320 conflicting locks in the rest of the list.
2324 Description: Time out and refuse to reconnect
2325 Details : When the failover node is the primary node, it is possible
2326 to have two identical connections in imp_conn_list. We must
2327 compare not conn's pointers but NIDs, otherwise we can defeat
2328 connection throttling.
2332 Description: port llog fixes from b1_6 into HEAD
2333 Details : Port llog reference couting and some llog cleanups from b1_6
2334 (bug 10800) into HEAD, for protect from panic and access to already
2335 free llog structures.
2339 Description: Detect stride IO mode in read-ahead
2340 Details : When a client does stride read, read-ahead should detect that and
2341 read-ahead pages according to the detected stride pattern.
2345 Description: data checksumming impacts single node performance
2346 Details : add support for several checksum algorithm. Currently, only CRC32
2347 and Adler-32 are supported. The checksum type can be changed on
2348 the fly via /proc/fs/lustre/osc/*/checksum_type.
2352 Description: use adler32 for page checksums
2353 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2358 Description: build for x2 fails
2359 Details : fix compile issue on Cray systems.
2363 Description: Properly match for duplicate locks
2364 Details : Due to different lock order from skiplists code, we need to
2365 traverse entire list for now
2368 Frequency : only on PPC/SLES10
2370 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2371 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2372 be defined. Add a hack in mkfs_lustre.c to work around this compile
2377 Description: LASSERT on MDS when client holding flock lock dies
2378 Details : ldlm pool logic depends on number of granted locks equal to
2379 number of released locks which is not true for flock locks, so
2380 just exclude such locks from consideration.
2384 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2385 Details : Use fsfilt_sync() for both the whole filesystem sync and
2386 individual file sync to eliminate dangerous inode locking
2387 with I_LOCK that can lead to a deadlock.
2391 Description: performance in 1.6.3
2392 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2393 sure that 1MB requests are not fragmented by the block layer.
2395 Severity : enhancement
2397 Description: organize the server-side client stats on per-nid basis
2398 Details : Change the structure of stats under obdfilter and mds to
2406 The "uuid"s file would list the uuids of _active_ exports.
2407 And the clear entry is to clear all stats and stale nids.
2409 Severity : enhancement
2411 Description: eliminate client locks in face of contention
2412 Details : file contention detection and lockless i/o implementation
2413 for contended files.
2417 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2420 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2422 Description: MDS may lose file striping (and hence file data) in some cases
2423 Details : If there are additional extended attributes stored on the MDS,
2424 in particular ACLs, SELinux, or user attributes (if user_xattr
2425 is specified for the client mount options) then there is a risk
2426 of attribute loss. Additionally, the Lustre file striping
2427 needs to be larger than default (e.g. striped over all OSTs),
2428 and an additional attribute must be stored initially in the
2429 inode and then increase in size enough to be moved to the
2430 external attribute block (e.g. ACL growing in size) for file
2435 Description: skiplist implementation simplification
2436 Details : skiplists are used to group compatible locks on granted list
2437 that was implemented as tracking first and last lock of each
2438 lock group the patch changes that to using doubly linked lists
2442 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2443 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2444 because LASSERT on both the data supplied by a client, and the
2445 data on disk is dangerous and incorrect.
2447 Severity : enhancement
2449 Description: Slow truncate/writes to huge files at high offsets.
2450 Details : Directly associate cached pages to lock that protect those pages,
2451 this allows us to quickly find what pages to write and remove
2452 once lock callback is received.
2456 Description: more ldlm soft lockups
2457 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2458 starve other threads from the resource lock for a long time in
2459 case of long waiting queue, so change the debug level from
2460 D_OTHER to the less frequently used D_INFO.
2462 Severity : enhancement
2464 Description: add -gid, -group, -uid, -user options to lfs find
2468 Description: Hung threads in invalidate_inode_pages2_range
2469 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2470 once one is completed. As a result, some RPCs are stuck in the
2471 queue and are never sent.
2475 Description: filter threads hungs on waiting journal commit
2476 Details : Cleanup filter group llog code, then only filter group llog will
2477 be only created in the MDS/OST syncing process.
2481 Description: Procfs and llog threads access destoryed import sometimes.
2482 Details : Sync the import destoryed process with procfs and llog threads by
2483 the import refcount and semaphore.
2485 Severity : enhancement
2487 Description: openlock cache of b1_6 port to HEAD
2492 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2493 Details : If insertion of an extent fails, then discard the inode
2494 preallocation and free data blocks else it can lead to duplicate
2499 Description: don't always update ctime in ext3_xattr_set_handle()
2500 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2501 In some cases the ctime should not be updated, for example for
2502 2.0->1.8 compatibility it is necessary to delete an xattr and it
2503 should not update the ctime.
2507 Bugzilla : 15713/16362
2508 Description: Assertion in iopen_connect_dentry in 1.6.3
2509 Details : looking up an inode via iopen with the wrong generation number can
2510 populate the dcache with a disconneced dentry while the inode
2511 number is in the process of being reallocated. This causes an
2512 assertion failure in iopen since the inode's dentry list contains
2513 both a connected and disconnected dentry.
2517 Description: assertion failure in ldlm_handle2lock()
2518 Details : fix a race between class_handle_unhash() and class_handle2object()
2519 introduced in lustre 1.6.5 by bug 13622.
2524 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2525 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2530 Description: oops in sd_iostats_seq_show()
2531 Details : unloading/reloading the scsi low level driver triggers a kernel
2532 bug when trying to access the sd iostat file.
2537 Description: Kernel panics during QLogic driver reload
2538 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2539 patch, causing memory corruption.
2544 Description: journal_dev option does not work in b1_6
2545 Details : pass mount option during pre-mount.
2547 Severity : enhancement
2549 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2550 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2551 extent information of a file. It can be used to map logical blocks
2552 in a file to physical blocks in the block device.
2556 Description: LDLM soft lockups - improvement
2557 Details : It is be possible to send the lock handle along with each read
2558 or write request because the client is already doing a lock match
2559 itself so there isn't any reason the OST should have to re-do that
2563 Frequency : only on Cray X2
2565 Description: X2 build failures
2566 Details : fix build failures on Cray X2.
2570 Description: xid & resent requests
2571 Details : Initialize RPC XID from clock at startup (randomly if clock is
2574 Severity : enhancement
2576 Description: Add lustre_start utility to start or stop multiple Lustre servers
2581 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2582 Details : In case of memory pressure, list_del() can be called twice on
2583 req->rq_history_list, causing a kernel oops.
2587 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2588 Details : fix stack overflow in the distributed lock manager by defering
2589 export eviction after a failed AST to the elt thread instead of
2590 handling it in the dlm interpret routine.
2594 Description: Convert some comments to new format.
2595 Details : Update documenting comments to match doxygen conventions.
2599 Description: Grammar fixes.
2600 Details : A couple of trivial spelling fixes.
2604 Description: OSD_COUNTERS-mandatory
2605 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2606 to simplify the code.
2610 Description: simplify cmm_device freeing logic.
2611 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2615 Description: Add lockdep support to dt_object_operations locking interface.
2616 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2617 parameter indicating lock ordering. Update mdd code to use new
2622 Description: Introduce failloc constants for lockless IO tests.
2623 Details : Add two new failloc constants to test lockless IO. Only one of
2624 them in implemented---another is checked in yet to be landed
2629 Description: Add lockdep support for inode mutex.
2630 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2631 in the situations where Lustre has to lock more than one inode
2636 Description: Add optional invariants checking support.
2637 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2638 configure switch. This macro is to be used for consistency and
2639 sanity checks that are too expensive to be left in `production'
2644 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2645 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2650 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2651 Details : Use spin_lock_nested() in (the only) situation where more than
2652 one ldlm_lock is locked simultaneously. Also, fix possible
2653 dead-lock in ldlm_lock_change_resource() by enforcing particular
2658 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2659 Details : Instead of specifying each ldlm_lock call-back through separate
2660 parameter, wrap them into struct ldlm_callback_suite.
2664 Description: Kill join_lru obd method and its callers.
2665 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2666 in memory---a requirement implemented through ->o_join_lru() obd
2667 method in HEAD. Remove this method and its users.
2671 Description: Add asynchronous ldlm ENQUEUE completion handler.
2672 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2673 case that stock ldlm_completion_ast() cannot handle as it waits
2674 until lock is granted. Introduce new ldlm_completion_ast_async()
2675 for this. Also comment ldlm_completion_ast().
2679 Description: ldlm_error <-> errno conversion.
2680 Details : Add functions to map (rather arbitrary) between LDLM error codes
2681 and standard errno values. CLIO needs this to prevent LDLM specific
2682 constants from escaping ldlm and osc.
2686 Description: Kill unused ldlm_handle2lock_ns() function.
2687 Details : Kill unused ldlm_handle2lock_ns() function.
2691 Description: Add lu_ref support to ldlm_lock
2692 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2693 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2694 and ldlm_resource. LDLM interface has to be changed, because code that
2695 releases a reference on a lock, has to "know" what reference this is.
2696 In the most frequent case
2698 lock = ldlm_handle2lock(handle);
2700 LDLM_LOCK_PUT(lock);
2702 no changes are required. When any other reference (received _not_ from
2703 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2704 instead of LDLM_LOCK_PUT().
2706 Arguably, changes are pervasive, and interface requires some discipline
2707 for proper use. On the other hand, it was very instrumental in finding
2708 a few leaked lock references.
2712 Description: Add ldlm_lock_addref_try().
2713 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2714 attempts to addref a lock that might be being canceled
2719 Description: Add ldlm_weigh_callback().
2720 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2721 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2722 hard-coded `number of pages' logic.
2726 Description: Add lockdep annotations to llog code.
2727 Details : Use appropriately tagged _nested() locking calls in the places
2728 where llog takes more than one ->lgh_lock lock.
2732 Description: Add loi_kms_set().
2733 Details : Wrap kms updates into a helper function.
2737 Description: Constify instances of struct lsm_operations.
2738 Details : Constify instances of struct lsm_operations.
2742 Description: lu_conf support.
2743 Details : On a server, a file system object is uniquely identified
2744 by a fid, which is sufficient to locate and load all object
2745 state (inode). On a client, on the other hand, more data are
2746 necessary instantiate an object. Change lu_object_find() and
2747 friends to take additional `lu_conf' argument describing object.
2748 Typically this includes layout information.
2752 Description: lu_context fixes.
2753 Details : Introduce new lu_context functions that are needed on the client
2754 side, where some system threads (ptlrpcd) are shared by multiple
2755 modules, and so cannot be stopped during module shutdown.
2759 Description: Add start and stop methods to lu_device_type_operations.
2760 Details : Introduce two new methods in lu_device_type_operations, that are
2761 invoked when first instance of a given type is created and last one
2762 is destroyed respectively. This is need by CLIO.
2766 Description: Add lu_ref support to struct lu_device.
2767 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2768 track leaked references.
2772 Description: Introduce lu_kmem_descr.
2773 Details : lu_kmem_descr and its companion interface allow to create
2774 and destroy a number of kmem caches at once.
2778 Description: Fix lu_object finalization race.
2779 Details : Fix a race between lu_object_find() finding an object and its
2780 concurrent finalization. This race is (most likely) not possible
2781 on the server, but might happen on the client.
2785 Description: Introduce lu_ref interface.
2786 Details : lu_ref is a debugging module allowing to track references to
2787 a given object. It is quite cpu expensive, and has to be
2788 explicitly enabled with --enable-lu_ref. See usage description
2793 Description: Factor lu_site procfs stats into a separate function.
2794 Details : Separate lu_site stats printing code into a separate function
2795 that can be reused on a client.
2799 Description: Constify instances of struct {lu,dt,md}_device_operations.
2800 Details : Constify instances of struct {lu,dt,md}_device_operations.
2804 Description: Introduce struct md_site and move meta-data specific parts of
2805 struct lu_site here.
2806 Details : Move md-specific fields out of struct lu_site into special struct
2807 md_site, so that lu_site can be used on a client.
2811 Description: Kill mdd_lov_destroy().
2812 Details : Remove unused mdd code.
2816 Description: Add st_block checking to multistat.c.
2817 Details : Add st_block checking to multistat.c.
2821 Description: Add lu_ref support to struct obd_device.
2822 Details : Add lu_ref tracking to obd_device.
2826 Description: Kill obd_set_fail_loc().
2827 Details : Remove unused code.
2831 Description: Add special type for ptlrpc_request interpret functions.
2832 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2833 there. Actual usage will be in CLIO.
2837 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2838 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2839 doesn't work with lockdep.
2843 Description: Add rwv.c test program.
2844 Details : New testing program exercising readv(2) and writev(2) (Qian).
2848 Description: Add sendfile.c test program.
2849 Details : New testing program exercising sendfile(2) (Jay).
2853 Description: Ratelimit a message that can be very frequent.
2854 Details : Ratelimit a memory allocation failure message that can
2859 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2860 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2864 Description: Liblustre build fixes.
2865 Details : Liblustre build fixes.
2869 Description: libcfs: add cfs_{need,cond}_resched() interface.
2870 Details : libcfs: add cfs_{need,cond}_resched() definition and
2871 implementations for Linux, NT, and liblustre.
2873 Severity : enhancement
2875 Description: More exported tunables for mballoc
2876 Details : Add support for tunable preallocation window and new tunables for
2877 large/small requests
2881 Description: Detect corruption of block bitmap and checking for preallocations
2882 Details : Checks validity of on-disk block bitmap. Also it does better
2883 checking of number of applied preallocations. When corruption is
2884 found, it turns filesystem readonly to prevent further corruptions.
2888 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2889 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2890 window to 0. If it does hit cache miss, and read-pattern is still
2891 stride-io mode, does not reset the stride window, but also does not
2892 increase the stride window length in this case.
2896 Frequency : only for big-endian servers
2897 Description: Check if system is big-endian while mounting fs with extents feature
2898 Details : Mounting a filesystem with extents feature will fail on big-endian
2899 systems since ext3-based ldiskfs is not supported on big-endian
2900 systems. This can be over-riden with "bigendian_extents" mount option.
2902 Severity : enhancement
2904 Description: The root squash functionality
2905 Details : A security feature, which is to prevent users from being able
2906 to mount lustre on their desktop, run as root, and delete
2907 all of the files in the filesystem. The goal is accomplished by
2908 remapping user id (UID) and group id (GID) of the root user to
2909 a UID and GID specified by the system administartor via Lustre
2910 configuration management server (MGS). The functionality also
2911 allows to specify sets of clients for which the remapping does
2916 Description: Excessive recovery window
2917 Details : With AT enabled, the recovery window can be excessively long (6000+
2918 seconds). To address this problem, we no longer use
2919 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2920 timeout no longer depends on the service time, it is set to
2921 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2922 time via pb_service_time.
2926 Description: Watchdog triggered on MDS failover
2927 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2928 the OSTs know that the MDS "UUID" can be reused for the same export
2929 from a different NID, so we do not need to wait for the export to be
2933 Frequency : rare, only if using MMP with Linux RAID
2935 Description: MMP doesn't work with Linux RAID
2936 Details : While using HA for Lustre servers with Linux RAID, it is possible
2937 that MMP will not detect multiple mounts. To make this work we
2938 need to unplug the device queue in RAID when the MMP block is being
2939 written. Also while reading the MMP block, we should read it from
2940 disk and not the cached one.
2942 Severity : enhancement
2944 Description: open file using fid
2945 Details : A file can be opened using just its fid, like
2946 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2949 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2952 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2953 Details : add proper locking for creation of jbd_4k slab cache
2957 Description: MMP check in ext3_remount() fails without displaying any error
2958 Details : When multiple mount protection fails during remount, proper error
2961 Severity : enhancement
2963 Description: Allow stripe size to be up to 4G-64k
2964 Details : Fix math logic to allow large stripe sizes.
2968 Description: add check for >8TB ldiskfs filesystems
2969 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2970 Don't allow >8TB ldiskfs filesystems to be mounted without
2971 force_over_8tb mount option
2976 Descriptoin: OSS complains "too many blocks" and crashes
2977 Details : In filter_commitrw_write():
2978 1. lquota_chkquota();
2979 2. obdo_from_inode();
2980 3. lquota_pending_commit();
2981 "2." has changed uid and gid in this case, then caused
2982 lquota_chkquota() and lquota_pending_commit() to use
2983 different uid and gid. That is the root of the bug.
2987 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2988 Details : Detect and print out the 'stripe count = -1' case properly.
2992 Description: Out or order replies might be lost on replay
2993 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2994 request with tid smaller than one currently being added, add it
2995 to the start, not end of the list.
2997 --------------------------------------------------------------------------------
2999 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3001 * Support for kernels:
3002 2.6.5-7.283 (SLES 9),
3003 2.6.9-55.EL (RHEL 4),
3004 2.6.16.46-0.14 (SLES 10),
3005 2.6.18.8 vanilla (kernel.org)
3006 * Client support for unpatched kernels:
3007 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3008 2.6.16 - 2.6.22 vanilla (kernel.org)
3009 * Due to recently discovered recovery problems, we do not recommend
3010 using patchless RHEL 4 clients with this or any earlier release.
3011 * Recommended e2fsprogs version: 1.39.cfs8
3012 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3013 * Starting with this release, the ldiskfs backing filesystem required
3014 by Lustre is now in its own package, lustre-ldiskfs. This package
3015 should be installed. It is versioned separately from Lustre and
3016 may be released separately in future.
3020 Description: block reactivating mgc import until all deactivates complete
3021 Details : Fix race when failing back MDT/MGS to itself (testing)
3024 Frequency : at statup only
3026 Description: mds_lov_synchronize race leads to various problems
3027 Details : simultaneous MDT->OST connections at startup can cause the
3028 sync to abort, leaving the OSC in a bad state.
3030 Severity : enhancement
3032 Description: add optional extra BUILD_VERSION info
3033 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3034 to override the lustre version.
3037 Frequency : 2.6.18 servers only
3039 Description: ll_kern_mount() doesn't release the module reference
3040 Details : The ldiskfs module reference count never drops down to 0
3041 because ll_kern_mount() doesn't release the module reference.
3046 Description: server LBUG when using old ost_num_threads parameter
3047 Details : Accept the old ost_num_threads parameter but warn that it
3048 is deprecated, and fix an off-by-one error that caused an LBUG.
3053 Description: Transient SCSI error results in persistent IO issue
3054 Details : iobuf->dr_error is not reinitialized to 0 between two
3058 Frequency : sometimes when underlying device returns I/O errors
3060 Description: OSTs not going read-only during write failures
3061 Details : OSTs are not remounted read-only when the journal commit threads
3062 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3063 instead of the ext3 wrappers.
3067 Description: poor connect scaling with increasing client count
3068 Details : Don't run filter_grant_sanity_check for more than 100 exports
3069 to improve scaling for large numbers of clients.
3072 Frequency : SLES10 only
3074 Description: sanity-quota.sh quotacheck failed: rc = -22
3075 Details : Quotas cannot be enabled on SLES10.
3078 Frequency : liblustre clients only
3080 Description: getdirentries does not give error when run on compute nodes
3081 Details : getdirentries does not fail when the size specified as an argument
3082 is too small to contain at least one entry
3084 Severity : enhancement
3086 Description: Add LNET router traceability for debug purposes
3087 Details : If a checksum failure occurs with a router as part of the
3088 IO path, the NID of the last router that forwarded the bulk data
3089 is printed so it can be identified.
3094 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3095 Details : Due to a race condition, liblustre clients were occasionally
3096 evicted incorrectly.
3098 Severity : enhancement
3100 Description: lfs setstripe use optional parameters instead of postional
3103 Severity : enhancement
3105 Description: Nanosecond timestamp support for ldiskfs
3106 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3107 resolution timestamps. There is not yet support for this at
3108 the Lustre filesystem level.
3111 Frequency : during server recovery
3113 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3114 Details : request with rq_no_resend flag not awake l_wait_event if they get
3118 Frequency : nfs export on patchless client
3120 Description: connectathon hang when test nfs export over patchless client
3121 Details : Disconnected dentry cannot be found with lookup, so we do not need
3122 to unhash it or make it invalid
3125 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3126 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3127 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3132 Description: open req refcounting wrong on reconnect
3133 Details : If reconnect happened between getting open reply from server and
3134 call to mdc_set_replay_data in ll_file_open, we will schedule
3135 replay for unreferenced request that we are about to free.
3136 Subsequent close will crash in variety of ways.
3137 Check that request is still eligible for replay in
3138 mdc_set_replay_data().
3143 Description: disable writes to filesystem when reading health_check file
3144 Details : the default for reading the health_check proc file has changed
3145 to NOT do a journal transaction and write to disk, because this
3146 can cause reads of the /proc file to hang and block HA state
3147 checking on a healthy but otherwise heavily loaded system. It
3148 is possible to return to the previous behaviour during configure
3149 with --enable-health-write.
3151 Severity : enhancement
3153 Description: 64-bit inode version
3154 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3155 to the inode. This will be required for version-based recovery.
3160 Description: MDS fails to start if a duplicate client export is detected
3161 Details : in some rare cases it was possible for a client to connect to
3162 an MDS multiple times. Upon recovery the MDS would detect this
3163 and fail during startup. Handle this more gracefully.
3165 Severity : enhancement
3167 Description: Add -o localflock option to simulate old noflock
3169 Details : This will achieve local-only flock/fcntl locks
3175 Description: log_commit_thread vs filter_destroy race leads to crash
3176 Details : Take import reference before releasing llog record semaphore
3181 Description: Wrong request locking in request set processing
3182 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3183 imp_delayed_list, in this place should be used imp_lock.
3186 Frequency : when reconnection
3188 Description: Grant Leak when osc reconnect to OST
3189 Details : When osc reconnect ost, OST(filter) should check whether it
3190 should grant more space to client by comparing fed_grant and
3191 cl_avail_grant, and return the granted space to client instead
3192 of "new granted" space, because client will call osc_init_grant
3193 to update the client grant space info.
3196 Frequency : when client reconnect to OST
3198 Description: Grant Leak when osc do resend and replay bulk write
3199 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3200 bulk write request, because the grant info while be sync between
3201 OSC and OST when reconnect, and we should ignore the grant info
3202 these of resend/replay write req.
3207 Description: Grant space more than available left space sometimes.
3208 Details : When then OST is about to be full, if two bulk writing from
3209 different clients came to OST. Accord the avaliable space of the
3210 OST, the first req should be permitted, and the second one
3211 should be denied by ENOSPC. But if the seconde arrived before
3212 the first one is committed. The OST might wrongly permit second
3213 writing, which will cause grant space > available space.
3216 Frequency : when client is evicted
3218 Description: Grant might be wrongly erased when osc is evicted by OST
3219 Details : when the import is evicted by server, it will fork another
3220 thread ptlrpc_invalidate_import_thread to invalidate the
3221 import, where the grant will be set to 0. While the original
3222 thread will update the grant it got when connecting. So if
3223 the former happened latter, the grant will be wrongly errased
3224 because of this race.
3229 Description: Checking Stale with correct fid
3230 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3231 to check whether it is stale, because sometimes, we want the
3232 enqueue happened anyway, and op_data.fid2 will not be initialized.
3234 Severity : enhancement
3236 Description: update patchless client
3237 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3240 Frequency : only with 2.4 kernel
3242 Description: random memory corruption
3243 Details : size of struct ll_inode_info is to big for union inode.u and this
3244 can be cause of random memory corruption.
3249 Description: Memory leak in recovery
3250 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3251 It should also check obd_fail before fsfilt_start, otherwise if
3252 fsfilt_start return -EROFS,(failover mds during mds recovery).
3253 then the req will return with repmsg->transno = 0 and rc = EROFS.
3254 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3255 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3256 be freed no matter whether fsfilt_commit success or not.
3259 Frequency : only with huge count clients
3261 Description: Prevents from taking the superblock lock in llap_from_page for
3263 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3264 for llap_from_page call in ll_removepage prevents from taking the
3265 superblock lock for a soon died page.
3270 Description: Not check open intent error before release open handle
3271 Details : in some rare cases, the open intent error is not checked before
3272 release open handle, which may cause
3273 ASSERTION(open_req->rq_transno != 0), because it tries to release
3274 the failed open handle.
3279 Description: Set cat log bitmap only after create log success.
3280 Details : in some rare cases, the cat log bitmap is set too early. and it
3281 should be set only after create log success.
3285 Description: Accessing a block bevice can re-enable I/O when Lustre is
3286 tearing down a device.
3287 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3292 Description: service threads may hog cpus when there are a lot of requests
3293 Details : Insert cond_resched to give other threads a chance to use some CPU
3298 Description: the cat log was not initialized in recovery
3299 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3300 unlink log on mds will not be initialized until mds post
3301 recovery. And also in mds post recovery, the unlink log will
3302 initialization will be done asynchronausly, so there will be race
3303 between add unlink log and unlink log initialization.
3307 Description: brw_stats were being printed incorrectly
3308 Details : brw_stats were being printed as log2 but all of them were not
3309 recorded as log2. Also remove some code duplication arising from
3310 filter_tally_{read,write}.
3314 Frequency : rare, only in recovery.
3315 Description: ASSERTION(req->rq_type != LI_POISON) failed
3316 Details : imp_lock should be held while iterating over imp_sending_list for
3317 prevent destroy request after get timeout in ptlrpc_queue_wait.
3321 Description: replay-single.sh test 52 fails
3322 Details : A lock's skiplist need to be cleanup when it being unlinked
3323 from its resource list.
3327 Description: Short directio read returns full requested size rather than
3329 Details : Direct I/O operations should return actual amount of bytes
3330 transferred rather than requested size.
3332 Severity : enhancement
3334 Description: metadata RPC reduction (e.g. for rm performance)
3335 Details : decrease the amount of synchronous RPC between clients and servers
3336 by canceling conflicing lock before the operation on the client
3337 and packing thier handles into the main operation RPC to server.
3339 Severity : enhancement
3341 Description: Async OSC create to avoid the blocking unnecessarily.
3342 Details : If an OST has no remaining object, system will block on the
3343 creation when it needs to create a new object on this OST. Now,
3344 ways use pre-created objects when available, instead of blocking on
3345 an empty osc while others are not empty. If we must block, we
3346 block for the shortest possible period of time.
3350 Description: improve handling recoverable errors
3351 Details : if request processig with error which can be recoverable on server
3352 request should be resend, otherwise page released from cache and
3355 Severity : enhancement
3357 Description: refine locking for avoid write wrong info into lov_objid file
3358 Details : fix possible races with add new target and write/update data in
3361 --------------------------------------------------------------------------------
3363 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3368 Frequency : on some architectures
3370 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3371 Details : Uninitialized flags sometimes cause configuration commands to
3375 Frequency : patchless clients only
3377 Description: missing __iget() symbol export
3378 Details : The __iget() symbol export is missing. To avoid the need for
3379 this on patchless clients the deathrow inode reaper is turned
3380 off, and we depend on the VM to clean up old inodes. This
3381 dependency was during via the fix for bug 12181.
3383 --------------------------------------------------------------------------------
3385 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3387 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3388 INTEROPERATE with older versions automatically. In many cases a
3389 special upgrade step is needed. Please read the
3390 user documentation before upgrading any part of a live system.
3391 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3392 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3393 * WARNING: Lustre configuration and startup changes are required with
3394 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3398 Severity : enhancement
3400 Description: MountConf
3401 Details : Lustre configuration is now managed via mkfs and mount
3402 commands instead of lmc and lconf. New obd types (MGS, MGC)
3403 are added for dynamic configuration management. See
3404 https://mail.clusterfs.com/wikis/lustre/MountConf for
3407 Severity : enhancement
3409 Description: dynamic OST addition
3410 Details : OSTs can now be added to a live filesystem
3412 Severity : enhancement
3414 Description: startup order invariance
3415 Details : MDTs and OSTs can be started in any order. Clients only
3416 require the MDT to complete startup.
3418 Severity : enhancement
3420 Description: parallel, asynchronous orphan cleanup
3421 Details : orphan cleanup is now performed in separate threads for each
3422 OST, allowing parallel non-blocking operation.
3424 Severity : enhancement
3426 Description: optimized stripe assignment
3427 Details : stripe assignments are now made based on ost space available,
3428 ost previous usage, and OSS previous usage, in order to try
3429 to optimize storage space and networking resources.
3431 Severity : enhancement
3433 Description: Permanently set tunables
3434 Details : All writable /proc/fs/lustre tunables can now be permanently
3435 set on a per-server basis, at mkfs time or on a live system.
3437 Severity : enhancement
3439 Description: Lustre message v2
3440 Details : Add lustre message format v2.
3442 Severity : enhancement
3444 Description: client OST exclusion list
3445 Details : Clients can be started with a list of OSTs that should be
3446 declared "inactive" for known non-responsive OSTs.
3450 Description: SPEC SFS validation failure on NFS v2 over lustre.
3451 Details : Changes the blocksize for regular files to be 2x RPC size,
3452 and not depend on stripe size.
3454 Severity : enhancement
3456 Description: Multiple MD RPCs in flight.
3457 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3458 To avoid overly-overloading MDS, introduce a limit on number of
3459 MDS RPCs in flight for a single client and add /proc controls
3460 to adjust this limit.
3462 Severity : enhancement
3464 Description: client read/write statistics
3465 Details : Add client read/write call usage stats for performance
3466 analysis of user processes.
3467 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3468 file access. extents_stats shows chunk size distribution.
3469 extents_stats_per_process show chunk size distribution per
3472 Severity : enhancement
3474 Description: mds statistics
3475 Details : Add detailed mds operations statistics in
3476 /proc/fs/lustre/mds/*/stats.
3480 Description: Failure of copying files with lustre special EAs.
3481 Details : Client side always return success for setxattr call for lustre
3482 special xattr (currently only "trusted.lov").
3487 Description: Refcount LNET uuids
3488 Details : The global LNET uuid list grew linearly with every startup;
3489 refcount repeated list entries instead of always adding to
3492 Severity : enhancement
3494 Description: Dynamic service threads
3495 Details : Within a small range, start extra service threads
3496 automatically when the request queue builds up.
3499 Frequency : mixed-endian client/server environments
3501 Description: mixed-endian crashes
3502 Details : The new msg_v2 system had some failures in mixed-endian
3505 Severity : enhancement
3507 Description: Easy OST removal
3508 Details : OSTs can be permanently deactivated with e.g. 'lctl
3509 conf_param lustre-OST0001.osc.active=0'
3511 Severity : enhancement
3513 Description: MGS proc entries
3514 Details : Added basic proc entries for the MGS showing what filesystems
3517 Severity : enhancement
3519 Description: provide MGS failover
3520 Details : Added config lock reacquisition after MGS server failover.
3522 Severity : enhancement
3524 Description: add Linux 2.4 support
3525 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3529 Description: a large application tries to do I/O to the same resource and dies
3530 in the middle of it.
3531 Details : Check the req->rq_arrival time after the call to
3532 ost_brw_lock_get(), but before we do anything about
3533 processing it & sending the BULK transfer request. This
3534 should help move old stale pending locks off the queue as
3535 quickly as obd_timeout.
3538 Frequency : when an incorrect nid is specified during startup
3540 Description: ptlrpc connect to non-existant node causes kernel crash
3541 Details : LNET can't be re-entered from an event callback, which
3542 happened when we expire a message after the export has been
3543 cleaned up. Instead, hand the zombie cleanup off to another
3546 Severity : enhancement
3548 Description: plain/inodebits lock performance improvement
3549 Details : Grouping plain/inodebits in granted list by their request modes
3550 and bits policy, thus improving the performance of search through
3554 Frequency : only if OST filesystem is corrupted
3556 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3557 Details : for a short time RPCs with bulk IO are in the replay list,
3558 but replay of bulk IOs is unimplemented. If the OST filesystem
3559 is corrupted due to disk cache incoherency and then replay is
3560 started it is possible to trip an assertion. Avoid putting
3561 committed RPCs into the replay list at all to avoid this issue.
3564 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3566 Description: System hang on startup
3567 Details : This bug allowed the liblustre (e.g. catamount) client to
3568 return to the app before handling all startup RPCs. This
3569 could leave the node unresponsive to lustre network traffic
3570 and manifested as a server ptllnd timeout.
3572 Severity : enhancement
3574 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3575 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3576 boolean to control whether to print peer debug info when a
3577 client's RPC times out.
3580 Frequency : only for kernels with patches from Lustre below 1.4.3
3582 Description: Remove old rdonly API
3583 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3586 Frequency : only for devices with external journals
3588 Description: Set external device read-only also
3589 Details : During a commanded failover stop, we set the disk device
3590 read-only while the server shuts down. We now also set any
3591 external journal device read-only at the same time.
3594 Frequency : when upgrading from 1.4 while trying to change parameters
3596 Description: The wrong (new) MDC name was used when setting parameters for
3597 upgraded MDT's. Also allows changing of OSC (and MDC)
3598 parameters if --writeconf is specified at tunefs upgrade time.
3601 Frequency : when setting specific ost indicies
3603 Description: QOS code breaks on skipped indicies
3604 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3605 created with --index need not be sequential.
3607 Severity : enhancement
3609 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3610 Details : The uninit_groups feature works in conjunction with the kernel
3611 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3612 up the pass1 processing of e2fsck. This is a read-only feature
3613 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3614 mount filesystems that have had this feature enabled.
3616 Severity : enhancement
3618 Description: Improve multi-block allocation algorithm to avoid fragmentation
3619 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3620 allocation locality and avoid filesystem fragmentation.
3622 ------------------------------------------------------------------------------
3624 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3626 * Support for kernels:
3627 2.4.21-47.0.1.EL (RHEL 3)
3628 2.6.5-7.283 (SLES 9)
3629 2.6.9-42.0.10.EL (RHEL 4)
3630 2.6.12.6 vanilla (kernel.org)
3631 2.6.16.27-0.9 (SLES 10)
3632 * Recommended e2fsprogs version: 1.39.cfs5
3634 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3638 Frequency : occasional, depends on client load and configuration
3639 Bugzilla : 12181, 12203
3640 Description: data loss for recently-modified files
3642 Details : In some cases it is possible that recently written or created
3643 files may not be written to disk in a timely manner (this should
3644 normally be within 30s unless client IO load is very high).
3645 The problem appears as zero-length files or files that are a
3646 multiple of 1MB in size after a client crash or client eviction
3647 that are missing data at the end of the file.
3649 This problem is more likely to be hit on clients where files are
3650 repeatedly created and unlinked in the same directory, clients
3651 have a large amount of RAM, have many CPUs, the filesystem has
3652 many OSTs, the clients are rebooted frequently, and/or the files
3653 are not accessed by other nodes after being written.
3655 The presence of the problem can be detected by looking at
3656 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3657 smaller than the second (nr_unused) then dirty files will not
3658 be flushed automatically to disk. "sync; sleep 10" should be
3659 run several times on the node before unmounting it to update
3660 Lustre (this is also safe to run on nodes without this problem).
3662 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3663 that can cause this same problem, so customers using that kernel
3664 also need to update the kernel in addition to Lustre. In order
3665 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3667 It is normal that files written just before a client crash (less
3668 than 30s) may not yet have been flushed to disk, even for local
3672 Frequency : frequent on thin XT3 nodes
3674 Description: UUID collision on thin XT3 Linux nodes
3675 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3676 randomly, since we relied on an insufficiently-seeded PRNG.
3681 Description: OSS hangs after "All ost request buffers busy"
3682 Details : A deadlock between quota and journal operations caused OSS
3683 hangs after printing "All ost request buffers busy."
3686 Frequency : always on liblustre builds
3688 Description: Cleanup compiler warnings on liblustre
3691 Frequency : always on liblustre builds on XT3
3693 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3697 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3698 > 32000 subdirectories
3699 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3700 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3701 subdirectory count crosses 32000. This will aid e2fsck to
3702 correctly handle more than 32000 subdirectories.
3705 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3707 Description: System hang on startup
3708 Details : This bug allowed the liblustre (e.g. catamount) client to
3709 return to the app before handling all startup RPCs. This
3710 could leave the node unresponsive to lustre network traffic
3711 and manifested as a server ptllnd timeout.
3713 Severity : enhancement
3715 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3716 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3717 boolean to control whether to print peer debug info when a
3718 client's RPC times out.
3723 Description: make O_SYNC working on 2.6 kernels
3724 Details : 2.6 kernels use different method for mark pages for write,
3725 so need add a code to lustre for O_SYNC work.
3730 Description: Failure to close file and release space on NFS
3731 Details : Put inode details into lock acquired in ll_intent_file_open.
3732 Use mdc_intent_lock in ll_intent_open to properly
3733 detect all kind of errors unhandled by mdc_enqueue.
3738 Description: proc file read during shutdown sometimes raced obd removal,
3740 Details : Add lock to prevent obd access after proc file removal.
3743 Frequency : Only for files larger than 4GB on 32-bit clients.
3745 Description: improperly doing page alignment of locks
3746 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3747 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3752 Description: RPCs being resent when they shouldn't be
3753 Details : Some RPCs that should not be resent are being resent. This
3754 can cause inconsistencies in the RPC state machine. Do not
3755 resend such requests.
3758 Frequency : rare, only with NFS export
3760 Description: Crash on NFS re-export node
3761 Details : under very unusual load conditions an assertion is hit in
3762 ll_intent_file_open()
3765 Frequency : only if OST filesystem is corrupted
3767 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3768 Details : for a short time RPCs with bulk IO are in the replay list,
3769 but replay of bulk IOs is unimplemented. If the OST filesystem
3770 is corrupted due to disk cache incoherency and then replay is
3771 started it is possible to trip an assertion. Avoid putting
3772 committed RPCs into the replay list at all to avoid this issue.
3777 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3778 Details : Large single O_DIRECT read and write calls can fail to allocate
3779 a sufficiently large buffer to process the request. In case of
3780 allocation failure the allocation is retried with a smaller
3781 buffer and broken into smaller requests.
3783 Severity : enhancement
3785 Description: Add -o localflock option to simulate old noflock behaviour.
3786 Details : This will achieve local-only flock/fcntl locks coherentness.
3791 Description: versioning check is incomplete
3792 Details : Checking the version difference of client vs. server, report
3793 error if the gap is too big.
3798 Description: add support PG_writeback bit
3799 Details : add support for PG_writeback bit for Lustre, for more carefull
3800 work with page cache in 2.6 kernel. This also fix some deadlocks
3801 and remove hack for work O_SYNC with 2.6 kernel.
3803 Severity : enhancement
3805 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3806 Details : The uninit_groups feature works in conjunction with the kernel
3807 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3808 up the pass1 processing of e2fsck. This is a read-only feature
3809 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3810 mount filesystems that have had this feature enabled.
3812 Severity : enhancement
3814 Description: Improve multi-block allocation algorithm to avoid fragmentation
3815 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3816 allocation locality and avoid filesystem fragmentation.
3818 ------------------------------------------------------------------------------
3820 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3822 * Support for kernels:
3823 2.6.9-42.0.3.EL (RHEL 4)
3824 2.6.5-7.276 (SLES 9)
3825 2.4.21-47.0.1.EL (RHEL 3)
3826 2.6.12.6 vanilla (kernel.org)
3827 2.6.16.21-0.8 (SLES10)
3828 * Recommended e2fsprogs version: 1.39.cfs2-0
3830 * The backwards-compatible /proc/sys/portals symlink has been removed
3831 in this release. Before upgrading, please ensure that you change
3832 any configuration scripts or /etc/sysctl.conf files that access
3833 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3834 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3835 in advance of the upgrade on any system running Lustre 1.4.6 or
3836 newer, since /proc/sys/lnet was added in that version.
3837 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3841 Frequency : only when quota is used
3843 Description: avoid scanning export list for quota master
3844 Details : Change the algorithms to avoid scanning export list in order
3845 to improve the efficiency.
3848 Frequency : MDS failover only, very rarely
3850 Description: "went back in time" messages on mds failover
3851 Details : The greatest transno may be lost when the current operation
3852 finishes with an error (transno==0) and the client's last_rcvd
3853 record is over-written. Save the greatest transno in the
3854 mds_last_transno for this case.
3857 Frequency : always for specific kernels and striping counts
3859 Description: client may get "Matching packet too big" without ACL support
3860 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3861 when trying to access files in certain configurations. The
3862 clients should in fact be denied when mounting because they do
3863 not understand ACLs.
3866 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3868 Description: many clients connecting with IO in progress causes connect timeouts
3869 Details : Avoid synchronous journal commits to avoid delays caused by many
3870 clients connecting/disconnecting when bulk IO is in progress.
3871 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3872 OST_IO_PORTAL to avoid delays behind potentially many pending
3876 Frequency : occasionally with multiple writers to a single file
3878 Description: shared writes to file may result in wrong size reported by stat()
3879 Details : Allow growing of kms when extent lock is cancelled
3882 Frequency : always with random mmap IO to multi-striped file
3884 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3885 Details : Only if the hole is at the end of OST object so that kms is too
3886 small. Fix is to increase kms accordingly in ll_nopage.
3889 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3891 Description: writes to a missing object would leak memory on the OST
3892 Details : If there is an inconsistency between the MDS and OST filesystems,
3893 such that the MDS references an object that doesn't exist, writes
3894 to that object will leak memory due to incorrect cleanup in the
3895 error handling path, eventually running out of memory on the OST.
3900 Description: Creating too long symlink causes lustre errors
3901 Details : Check symlink and name lengths before sending requests to MDS.
3904 Frequency : only if flock is enabled (not on by default)
3906 Description: posix locks not released on fd closure on 2.6.9+
3907 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3908 kernels, this caused such locks not to be released on fd close and
3909 then assertions on fs unmount about still used locks.
3912 Frequency : MDS failover only, very rarely
3914 Description: clients may get ASSERTION(granted_lock != NULL)
3915 Details : When request was taking a long time, and a client was resending
3916 a getattr by name lock request. The were multiple lock requests
3917 with the same client lock handle and
3918 mds_getattr_name->fixup_handle_for_resent_request found one of the
3919 lock handles but later failed with ASSERTION(granted_lock != NULL).
3924 Description: handle->h_buffer_credits > 0, assertion failure
3925 Details : h_buffer_credits is zero after truncate, causing assertion
3926 failure. This patch extends the transaction or creates a new
3930 Frequency : NFS re-export or patchless client
3931 Bugzilla : 11179, 10796
3932 Description: Crash on NFS re-export node (__d_move)
3933 Details : We do not want to hash the dentry if we don't have a lock.
3934 But if this dentry is later used in d_move, we'd hit uninitialised
3935 list head d_hash, so we just do this to init d_hash field but
3936 leave dentry unhashed.
3939 Frequency : NFS re-export or patchless client
3941 Description: NFS exports has problem with symbolic link
3942 Details : lustre client didn't properly install dentry when re-exported
3943 to NFS or running patchless client.
3946 Frequency : NFS re-export or patchless client
3948 Description: Various nfs/patchless fixes.
3949 Details : fixes reuse disconected alias for lookup process - this fixes
3950 warning "find_exported_dentry: npd != pd",
3951 fix permission error with open files at nfs.
3952 fix apply umask when do revalidate.
3955 Frequency : occasional
3957 Description: Crash on NFS re-export node
3958 Details : calling clear_page() on the wrong pointer triggered oops in
3959 generic_mapping_read().
3962 Frequency : rarely, using O_DIRECT IO
3964 Description: unaligned directio crashes client with LASSERT
3965 Details : check for unaligned buffers before trying any requests.
3968 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3970 Description: stale data returned from RAID cache
3971 Details : If only a small amount of IO is done to the RAID device before
3972 reading it again it is possible to get stale data from the RAID
3973 cache instead of reading it from disk.
3976 Frequency : always for sles10 kernel
3978 Description: sles10 support
3979 Details : ll_follow_link: compile fixes and using of nd_set_link
3980 under newer kernels.
3983 Frequency : depends on arch, kernel and compiler version, always on sles10
3986 Description: recursive or deep enough symlinks cause stack overflow
3987 Details : getting rid of large stack-allocated variable in
3991 Frequency : depends on hardware
3993 Description: lustre write performance loss in the SLES10 kernel
3994 Details : the performance loss is caused by using of write barriers in the
3995 ext3 code. The SLES10 kernel turns barrier support on by
3996 default. The fix is to undo that change for ldiskfs.
3998 ------------------------------------------------------------------------------
4000 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4002 * Support for kernels:
4003 2.6.9-42.0.3EL (RHEL 4)
4004 2.6.5-7.276 (SLES 9)
4005 2.4.21-47.0.1.EL (RHEL 3)
4006 2.6.12.6 vanilla (kernel.org)
4010 Frequency : quota enabled and large files being deleted
4012 Description: releasing more than 4GB of quota at once hangs OST
4013 Details : If a user deletes more than 4GB of files on a single OST it
4014 will cause the OST to spin in an infinite loop. Release
4015 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4020 Description: statfs data retrieved from /proc may be stale or zero
4021 Details : When reading per-device statfs data from /proc, in the
4022 {kbytes,files}_{total,free,avail} files, it may appear
4023 as zero or be out of date.
4026 Frequency : systems with MD RAID1 external journal devices
4028 Description: lconf's call to blkid is confused by RAID1 journal devices
4029 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4030 of returning all block devices that match the journal UUID.
4033 Frequency : always, for aggregate stripe size over 4GB
4035 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4036 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4037 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4040 Frequency : always if "lfs find" used on a local file/directory
4042 Description: "lfs find" segfaults if used on a local file/directory
4043 Details : The case where a directory component was not specified wasn't
4044 handled correctly. Handle this properly.
4047 Frequency : always on ppc64
4049 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4050 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4051 perform endianness conversion.
4054 Frequency : rarely (truncate to non-zero file size after write under load)
4055 Bugzilla : 10730, 10687
4056 Description: Files padded with zeros to next 4K multiple
4057 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4058 it is possible that files that are truncated to a non-zero size
4059 immediately after being written are filled with zero bytes beyond
4060 the truncated size. No file data is lost.
4062 Severity : enhancement
4064 Description: Allow recovery/failover for liblustre clients.
4065 Details : liblustre clients were unaware of failover configurations until
4068 Severity : enhancement
4070 Description: user file locks should fail when not mounting with flock option
4071 Details : Set up an error-returning stub in ll_file_operations.lock field
4072 to prevent incorrect behaviour when client is mounted without
4073 flock option. Also, set up properly f_op->flock field for
4077 Frequency : always on ia64
4079 Description: "lfs df" loops on printing out MDS statfs information
4080 Details : The obd_ioctl_data was not initialized and in some systems
4081 this caused a failure during the ioctl that did not return
4082 an error. Initialize the struct and return an error on failure.
4085 Frequency : SLES 9 only
4087 Description: Error of copying files with lustre special EAs as root
4088 Details : Client side always return success for setxattr call for lustre
4089 special xattr (currently only "trusted.lov").
4092 Frequency : rarely on clusters with both ia64+i386 clients
4094 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4095 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4096 clients are doing concurrent writes to the same file it is
4097 possible that the ia64 clients may LASSERT because the OST
4098 extent locks are not PAGE_SIZE aligned. Ensure that grown
4099 locks are always aligned on the request boundary.
4102 Frequency : specific use, occasional
4104 Description: Overwriting in use executable truncates on-disk binary image
4105 Details : If one node attempts to overwrite an executable in use by
4106 another node, we now correctly return ETXTBSY instead of
4107 truncating the file.
4109 Severity : enhancement
4111 Description: Async OSC create to avoid the blocking unnecessarily.
4112 Details : If an OST has no remaining object, system will block on the creating
4113 when need to create a new object on this OST. Now, ways use
4114 pre-created objects when available, instead of blocking on an
4115 empty osc while others are not empty. If we must block, we block
4116 for the shortest possible period of time.
4121 Description: chmod on Lustre root is propagated to other clients
4122 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4123 invalid by the follow_mount time.
4126 Frequency : liblustre clients only
4128 Description: Race in 'instant cancel' lock handling could lead to such locks
4129 never to be granted in case of SMP MDS
4130 Details : Do not destroy not yet granted but cbpending locks in
4134 Frequency : replay/resend of open
4136 Description: non null lock assetion failure in mds_intent_policy
4137 Details : Trying to replay/resend lockless open requests resulted in
4138 mds_open() returning 0 with no lock. Now it sets a flag if
4139 a lock is going to be returned.
4141 Severity : enhancement
4143 Description: Checksum enhancements
4144 Details : New checksum enhancements allow for resending RPCs that failed
4147 Severity : enhancement
4149 Description: Tunables on number of dirty pages in cacche
4150 Details : Allow to set limit on number of dirty pages cached.
4155 Description: client crash on unmount - lock still has references
4156 Details : In some error handling cases it was possible to leak a lock
4157 reference on a client while accessing a file. This was not
4158 harmful to the client during operation, but would cause the
4159 client to crash when the filesystem is unmounted.
4162 Frequency : specific case, rare
4164 Description: ETXTBSY on mds though file not in use by client
4165 Details : ETXTBSY is no longer incorrectly returned when attempting to
4166 chmod or chown a directory that the user previously tried to
4167 execute or a currently-executing binary.
4170 Frequency : extremely rare except on liblustre-based clients
4172 Description: Lustre space not freed when files are deleted
4173 Details : Clean up open-unlinked files after client eviction. Previously
4174 the unlink was skipped and the files remained as orphans.
4179 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4180 Details : In some cases it was possible to send a blocking callback to
4181 a client doing a glimpse, even though that client didn't get
4182 a lock granted. When the glimpse lock is cancelled on the OST
4183 the freed lock is left on the waiting list and corrupted the list.
4186 Frequency : all core dumps
4188 Description: Broke core dumps to lustre
4189 Details : Negative dentry may be unhashed if parent does not have UPDATE
4190 lock, but some callers, e.g. do_coredump, expect dentry to be
4191 hashed after successful create, hash it in ll_create_it.
4193 ------------------------------------------------------------------------------
4195 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4197 * Support for kernels:
4198 2.6.9-42.0.2.EL (RHEL 4)
4199 2.6.5-7.276 (SLES 9)
4200 2.4.21-40.EL (RHEL 3)
4201 2.6.12.6 vanilla (kernel.org)
4205 Frequency : always on RHEL 3
4207 Description: Number of open files grows over time
4208 Details : The number of open files grows over time, whether or not
4209 Lustre is started. This was due to a filp leak introduced
4210 by one of our kernel patches.
4212 ------------------------------------------------------------------------------
4214 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4216 * Support for kernels:
4217 2.6.9-42.EL (RHEL 4)
4218 2.6.5-7.276 (SLES 9)
4219 2.4.21-40.EL (RHEL 3)
4220 2.6.12.6 vanilla (kernel.org)
4225 Bugzilla : 5719, 9635, 9792, 9684
4226 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4227 Details : If a server is under heavy load and cannot reply to new
4228 connection requests before the client resends the (re)connect,
4229 the connection handling code can behave badly if two service
4230 threads are concurrently handing separate (re)connections from
4231 the same client. Add better locking to the connection handling
4232 code, and ensure that only a single connection will be processed
4233 for a given client UUID, even if the lock is dropped.
4235 Severity : enhancement
4237 Description: add TCP zero-copy support to kernel
4238 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4239 sends if the hardware supports scatter-gather and checksumming.
4240 This allows socklnd to do client-write and server-read more
4241 efficiently and reduce CPU utilization from skbuf copying.
4244 Frequency : only if NFS exporting from client
4246 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4247 Details : If setting attributes on a file created under NFS that had
4248 never been opened it would be possible to oops the client
4249 if the file had no objects.
4252 Frequency : always for liblustre
4254 Description: liblustre client does MDS+OSTs setattr RPC for each write
4255 Details : When doing a write from a liblustre client, the client
4256 incorrectly issued an RPC to the MDS and each OST the file was
4257 striped over in order to update the timestamps. When writing
4258 with small chunks and many clients this could overwhelm the MDS
4259 with RPCs. In all cases it would slow down the write because
4260 these RPCs are unnecessary.
4262 Severity : enhancement
4264 Description: allow number of MDS service threads to be changed at module load
4265 Details : It is now possible to change the number of MDS service threads
4266 running. Adding "options mds mds_num_threads={N}" to the MDS's
4267 /etc/modprobe.conf will set the number of threads for the next
4268 time Lustre is restarted (assuming the "mds" module is also
4269 reloaded at that time). The default number of threads will
4270 stay the same, 32 for most systems.
4275 Description: OST crash if filesystem is unformatted or corrupt
4276 Details : If an OST is started on a device that has never been formatted
4277 or if the filesystem is corrupt and cannot even mount then the
4278 error handling cleanup routines would dereference a NULL pointer.
4283 Description: NULL pointer deref in llap_from_page.
4284 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4285 incorrect) mapping if the page was truncated/reclaimed while it was
4286 searched for. Check for this condition and skip such pages when
4287 doing readahead. Introduce extra check to llap_from_page() to
4288 verify page->mapping->host is non-NULL (so page is not anonymous).
4291 Frequency : Sometimes when using sys_sendfile
4293 Description: "page not covered by a lock" warnings from ll_readpage
4294 Details : sendfile called ll_readpage without right page locks present.
4295 Now we introduced ll_file_sendfile that does necessary locking
4296 around call to generic_file_sendfile() much like we do in
4300 Frequency : with certain MDS communication failures at client mount time
4302 Description: NULL pointer deref after failed client mount
4303 Details : a client connection request may delayed by the network layer
4304 and not be sent until after the PTLRPC layer has timed out the
4305 request. If the client fails the mount immediately it will try
4306 to clean up before the network times out the request. Add a
4307 reference from the request import to the obd device and delay
4308 the cleanup until the network drops the request.
4311 Frequency : occasionally during client (re)connect
4313 Description: assertion failure during client (re)connect
4314 Details : processing a client connection request may be delayed by the
4315 client or server longer than the client connect timeout. This
4316 causes the client to resend the connection request. If the
4317 original connection request is replied in this interval, the
4318 client may trip an assertion failure in ptlrpc_connect_interpret()
4319 which thought it would be the only running connect process.
4322 Frequency : only with obd_echo servers and clients that are rebooted
4324 Description: kernel BUG accessing uninitialized data structure
4325 Details : When running an obd_echo server it did not start the ping_evictor
4326 thread, and when a client was evicted an uninitialized data
4327 structure was accessed. Start the ping_evictor in the RPC
4328 service startup instead of the OBD startup.
4330 Severity : enhancement
4331 Bugzilla : 10193 (patchless)
4332 Description: Remove dependency on various unexported kernel interfaces.
4333 Details : No longer need reparent_to_init, exit_mm, exit_files,
4334 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4337 Frequency : rare (only users of deprecated and unsupported LDAP config)
4339 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4340 Details : LDAP apparently contains 'lustreName' attributes instead of
4341 'name'. A simple remapping of the name is sufficient.
4344 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4346 Description: waiting_locks_callback trips kernel BUG if client is evicted
4347 Details : Running with the dump_on_timeout debug flag turned on makes
4348 it possible that the waiting_locks_callback() can try to dump
4349 the Lustre kernel debug logs from an interrupt handler. Defer
4350 this log dumping to the expired_lock_main() thread.
4352 Severity : enhancement
4354 Description: Support NFS exporting on 2.6 kernels.
4355 Details : Implement non-rawops metadata methods for NFS server to use without
4356 changing NFS server code.
4359 Frequency : very rare (synthetic metadata workload only)
4361 Description: two racing renames might cause an MDS thread to deadlock
4362 Details : Running the "racer" program may cause one MDS thread to rename
4363 a file from being the source of a rename to being the target of
4364 a rename at exactly the same time that another thread is doing
4365 so, and the second thread has already enqueued these locks after
4366 doing a lookup of the target and is trying to relock them in
4367 order. Ensure that we don't try to re-lock the same resource.
4370 Frequency : only very large systems with liblustre clients
4372 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4373 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4374 all OSTs in parallel. This allows the eviction of stale liblustre
4375 clients to proceed much faster than if they were done in series,
4376 and also offers similar improvements for other set_info RPCs.
4381 Description: excessive CPU usage during initial read phase on client
4382 Details : During the initial read phase on a client, it would agressively
4383 retry readahead on the file, consuming too much CPU and impacting
4384 performance (since 1.4.5.8). Improve the readahead algorithm
4385 to avoid this, and also improve some other common cases (read
4386 of small files in particular, where "small" is files smaller than
4387 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4392 Description: MDS crash when receiving packet with unknown intent.
4393 Details : Do not LBUG in unknown intent case, just return -EFAULT
4395 Severity : enhancement
4396 Bugzilla : 9293, 9385
4397 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4398 Details : Do not serialize getattr (non-intent version) and statfs.
4401 Frequency : occasional, when OST network is overloaded/intermittent
4403 Description: client evicted by OST after bulk IO timeout
4404 Details : If a client sends a bulk IO request (read or write) the OST
4405 may evict the client if it is unresposive to its data GET/PUT
4406 request. This is incorrect if the network is overloaded (takes
4407 too long to transfer the RPC data) or dropped the OST GET/PUT
4408 request. There is no need to evict the client at all, since
4409 the pinger and/or lock callbacks will handle this, and the
4410 client can restart the bulk request.
4413 Frequency : Always when mmapping file with no objects
4415 Description: client crashes when mmapping file with no objects
4416 Details : Check that we actually have objects in a file before doing any
4417 operations on objects in ll_vm_open, ll_vm_close and
4423 Description: Request leak when working with deleted CWD
4424 Details : Introduce advanced request refcount tracking for requests
4425 referenced from lustre intent.
4427 Severity : Enhancement
4429 Description: Cache open file handles on client.
4430 Details : MDS now will return special lock along with openhandle, if
4431 requested and client is allowed to hold openhandle, even if unused,
4432 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4433 closing files for every read/write openration.
4435 Severity : Enhancement
4437 Description: Cache open negative dentries on client when possible.
4438 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4439 negative dentries on lock revocation.
4444 Description: Remounting a client read-only wasn't possible with a zconf mount
4445 Details : It wasn't possible to remount a client read-only with llmount.
4447 Severity : enhancement
4448 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4449 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4450 MPICH is included to add Lustre-specific ADIO interfaces.
4451 This is based closely on the UFS ADIO layer and only differs
4452 in file creation, in order to allow the OST striping to be set.
4453 This is user-contributed code and not supported by CFS.
4458 Description: extended inode attributes (immutable, append-only) work improperly
4459 when 2.4 and 2.6 kernels are used on client/server or vice versa
4460 Details : Introduce kernel-independent values for these flags.
4462 Severity : enhancement
4465 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4466 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4467 granularity. Now, Fractional MB granularity is be supported,
4468 this is very useful for low memory system.
4470 Severity : enhancement
4472 Description: Getattr by fid
4473 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4474 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4475 have locks and avoids getattr rpc requests for it on every stat.
4478 Frequency : Always, for filesystems larger than 2TB
4480 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4481 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4482 filesystems larger than 2TB to be mounted with mballoc enabled.
4485 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4487 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4488 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4489 filesystems larger than 2TB is mounted then the kernel will
4490 silently corrupt the start of the filesystem. CONFIG_LBD is
4491 enabled for all CFS-supported kernels, but the possibility of
4492 this happening with a modified kernel config exists.
4494 Severity : enhancement
4496 Description: add client O_DIRECT support for 2.6 kernels
4497 Details : It is now possible to do O_DIRECT reads and writes to files
4498 in the Lustre client mountpoint on 2.6 kernel clients.
4500 Severity : enhancement
4502 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4503 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4504 parallel, not waiting for response from every OST before sending
4505 a rpc to the next OST.
4510 Description: setattr vs write race when updating file timestamps
4511 Details : Client processes that update a file timestamp into the past
4512 right after writing to the file (e.g. tar) it is possible that
4513 the updated file modification time can be reset to the current
4514 time due to a race between processing the setattr and write RPC.
4516 Severity : enhancement
4518 Description: Bring 'lfs find' closer in line with regular Linux find.
4519 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4520 -print0 options and obtains all the needed info through the lustre
4523 Severity : enhancement
4525 Description: support up to 1024 configured devices on one node
4526 Details : change obd_dev array from statically allocated to dynamically
4527 allocated structs as they are first used to reduce memory usage
4532 Description: Flush dirty partially truncated pages during truncate
4533 Details : Immediatelly flush partially truncated pages in filter_setattr,
4534 this way we completely avoid having any pages in page cache on OST
4535 and can retire ugly workarounds during writes to flush such pages.
4540 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4541 Details : protect inode from truncation within vfs_unlink() context
4542 just take a reference before calling vfs_unlink() and release it
4543 when parent's i_sem is free.
4548 Description: last_id value checked outside lock on OST caused LASSERT failure
4549 Details : If there were multiple MDS->OST object precreate requests in
4550 flight, it was possible that the OST's last object id was checked
4551 outside a lock and incorrectly tripped an assertion. Move checks
4552 inside locks, and discard old precreate requests.
4555 Frequency : always, if extents are used on OSTs
4557 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4558 Details : OSTs using the extents format would not zero the high 16 bits of
4559 the index physical block number. This is not a problem for any
4560 OST filesystems smaller than 16TB, and no kernels support ext3
4561 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4562 new/modified files) and can be fixed for existing filesystems
4563 with e2fsprogs-1.39-cfs1.
4568 Description: import connection selection may be incorrect if timer wraps
4569 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4570 import connections to be ignored if the 32-bit jiffies counter
4571 wraps. Use a 64-bit jiffies counter.
4574 Frequency : very large clusters immediately after boot
4576 Description: LNET request buffers exhausted under heavy short-term load
4577 Details : If a large number of client requests are generated on a service
4578 that has previously never seen so many requests it is possible
4579 that the request buffer growth cannot keep up with the spike in
4580 demand. Instead of dropping incoming requests, they are held in
4581 the LND until the RPC service can accept more requests.
4584 Frequency : Sometimes during replay
4586 Description: Assertion failure in ll_local_open after replay.
4587 Details : If replay happened on an open request reply before we were able
4588 to set replay handler, reply will become not swabbed tripping the
4589 assertion in ll_local_open. Now we set the handler right after
4590 recognising of open request
4593 Frequency : very rare
4595 Description: kernel reports "badness in vsnprintf"
4596 Details : Reading from the "recovery_status" /proc file in small chunks
4597 may cause a negative length in lprocfs_obd_rd_recovery_status()
4598 call to vsnprintf() (which is otherwise harmless). Exit early
4599 if there is no more space in the output buffer.
4601 Severity : enhancement
4603 Description: clear OBD RPC statistics by writing to them
4604 Details : It is now possible to clear the OBD RPC statistics by writing
4605 to the "stats" file.
4610 Description: Client mtime is not the same on different clients after utimes
4611 Details : In some cases, the client was using the utimes() syscall on
4612 a file cached on another node. The clients now validate the
4613 ctime from the MDS + OSTs to determine which one is right.
4618 Description: Inability to activate failout mode
4619 Details : lconf script incorrectly assumed that in pythong string's numeric
4620 value is used in comparisons.
4623 Frequency : always with multiple stripes per file
4625 Description: Inefficient object allocation for mutli-stripe files
4626 Details : When selecting which OSTs to stripe files over, for files with
4627 a stripe count that divides evenly into the number of OSTs,
4628 the MDS is always picking the same starting OST for each file.
4629 Return the OST selection heuristic to the original design.
4634 Description: mount failures may take full timeout to return an error
4635 Details : Under some heavy load conditions it is possible that a
4636 failed mount can wait for the full obd_timeout interval,
4637 possibly several minutes, before reporting an error.
4638 Instead return an error as soon as the status is known.
4640 Frequency : quota enabled and large files being deleted
4642 Description: releasing more than 4GB of quota at once hangs OST
4643 Details : If a user deletes more than 4GB of files on a single OST it
4644 will cause the OST to spin in an infinite loop. Release
4645 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4650 Description: statfs data retrieved from /proc may be stale or zero
4651 Details : When reading per-device statfs data from /proc, in the
4652 {kbytes,files}_{total,free,avail} files, it may appear
4653 as zero or be out of date.
4656 Frequency : systems with MD RAID1 external journal devices
4658 Description: lconf's call to blkid is confused by RAID1 journal devices
4659 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4660 of returning all block devices that match the journal UUID.
4663 Frequency : always, for aggregate stripe size over 4GB
4665 Description: assertion fails when trying to use 4GB stripe size
4666 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4667 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4670 Frequency : always on ppc64
4672 Description: the first write on an ext3 filesystem with mballoc got stuck
4673 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4674 perform endianness conversion.
4676 ------------------------------------------------------------------------------
4678 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4680 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4681 INTEROPERATE with older versions automatically. Please read the
4682 user documentation before upgrading any part of a live system.
4683 * WARNING: Lustre networking configuration changes are required with
4684 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4687 * Support for kernels:
4688 2.6.9-22.0.2.EL (RHEL 4)
4689 2.6.5-7.244 (SLES 9)
4690 2.6.12.6 vanilla (kernel.org)
4693 Severity : enhancement
4694 Bugzilla : 7981/8208
4695 Description: Introduced Lustre Networking (LNET)
4696 Details : LNET is new networking infrastructure for Lustre, it includes
4697 a reorganized network configuration mode (see the user
4698 documentation for full details) as well as support for routing
4699 between different network fabrics. Lustre Networking Devices
4700 (LNDs) for the supported network fabrics have also been
4701 created for this new infrastructure.
4703 Severity : enhancement
4704 Description: Introduced Access control lists
4705 Details : clients can set ACLs on files and directories in order to have
4706 more fine-grained permissions than the standard Unix UGO+RWX.
4707 The MDS must be started with the "-o acl" mount option.
4709 Severity : enhancement
4710 Description: Introduced filesystem quotas
4711 Details : Administrators may now establish per-user quotas on the
4714 Severity : enhancement
4716 Description: Configuration change for the XT3
4717 The PTLLND is now used to run Lustre over Portals on the XT3
4718 The configure option(s) --with-cray-portals are no longer used.
4719 Rather --with-portals=<path-to-portals-includes> is used to
4720 enable building on the XT3. In addition to enable XT3 specific
4721 features the option --enable-cray-xt3 must be used.
4726 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4727 Details : A race between allocating a new llcd and re-getting the llcd_lock
4728 allowed another thread to grab newly-allocated llcd.
4730 Severity : enhancement
4732 Description: 2.6 OST async journal commit and locking fix to improve performance
4733 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4734 2.6 kernels are now async as they already were in 2.4 kernels so
4735 that they can commit concurrently with the network bulk transfer.
4736 For block-allocated files the filter allocation semaphore is held
4737 to avoid filesystem fragmentation during allocation. BKL lock
4738 removed for 2.6 xattr operations where it is no longer needed.
4743 Description: lconf incorrectly determined whether two IP networks could talk
4744 Details : In some more complicated routing and multiple-network
4745 configurations, lconf will avoid trying to make a network
4746 connection to a disjoint part of the IP space. It was doing the
4747 math incorrectly for one set of cases.
4752 Description: Fix for potential infinite loop processing records in an llog.
4753 Details : If an llog record is corrupted/zeroed, it is possible to loop
4754 forever in llog_process(). Validate the llog record length
4755 and skip the remainder of the block on error.
4758 Frequency : occasional (liblustre only)
4760 Description: liblustre could not open files whose last component is a symlink
4761 Details : sysio_path_walk() would incorrectly pass the open intent to
4762 intermediate path components.
4765 Frequency : rare (liblustre only with non-standard tuning)
4766 Bugzilla : 7201 (7350)
4767 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4768 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4769 after it has been released, only do this for referenced locks
4771 Severity : enhancement
4773 Description: specifying an (invalid) directory default stripe_size of -1
4774 would reset the directory default striping
4775 Details : stripe_size -1 was used internally to signal directory stripe
4776 removal, now use "all default" to signal dir stripe removal
4777 as a directory striping of "all default" is not useful
4780 Frequency : common for large clusters running liblustre clients
4782 Description: doing an ls when liblustre clients are running is slow
4783 Details : sending a glimpse AST to a liblustre client waits for every AST
4784 to time out, as liblustre clients will not respond. Since they
4785 cannot cache data we refresh the OST lock LVB from disk instead.
4787 Severity : enhancement
4789 Description: doing an ls at the same time as file IO can be slow
4790 Details : enqueue and other "small" requests can be blocked behind many
4791 large IO requests. Create a new OST IO portal for non-IO
4792 requests so they can be processed faster.
4795 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4797 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4798 created file with mode 000
4799 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4800 allow an ftruncate() as a non-root user to a file with mode 000.
4801 Lustre now respects this flag to disable mode checks when
4802 truncating a file owned by the user
4805 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4808 Description: Revoking locks from clients that went dead or catatonic might take
4810 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4811 cancellation of such locks instant on servers without waiting for
4812 any reply from clients. Clients drops these locks when cancel
4813 notification from server is received without replying.
4816 Frequency : liblustre-only, when liblustre client dies or becomes busy
4818 Description: Doing ls on Linux clients can take a long time with active
4820 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4821 granting such locks to it in the first place if possible. Locks
4822 are taken by proxy on the OST during the read or write and
4823 dropped immediately afterward. Add connect flags handling, do
4824 not grant locks to liblustre clients for glimpse ASTs.
4826 Severity : enhancement
4828 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4829 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4830 sequential reads and behaves badly with applications doing random
4831 reads. Improve it by reading ahead at least read region, and
4832 avoiding excessive large RPC for small reads.
4834 Severity : enhancement
4836 Description: Creating more than 1000 files for a single job may cause a load
4837 imbalance on the OSTs if there are also a large number of OSTs.
4838 Details : qos_prep_create() uses an OST index reseed value that is an
4839 even multiple of the number of available OSTs so that if the
4840 reseed happens in the middle of the object allocation it will
4841 still utilize the OSTs as uniformly as possible.
4846 Description: OST or MDS may oops in ping_evictor_main()
4847 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4848 but doesn't restart at beginning of obd_exports_timed list
4851 Severity : enhancement
4853 Description: improve by-nid export eviction on the MDS and OST
4854 Details : allow multiple exports with the same NID to be evicted at one
4855 time without re-searching the exports list.
4858 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4860 Description: MDS may oops in groups_free()
4861 Details : in rare race conditions a newly allocated group_info struct is
4862 freed again, and this can be NULL. The 2.4 compatibility code
4863 for groups_free() checked for a NULL pointer, but 2.6 did not.
4866 Frequency : common for liblustre clients doing little filesystem IO
4867 Bugzilla : 9352, 7313
4868 Description: server may evict liblustre clients accessing contended locks
4869 Details : if a client is granted a lock or receives a completion AST
4870 with a blocking AST already set it would not reply to the AST
4871 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4874 Frequency : lfs setstripe, only systems with more than 160 OSTs
4876 Description: unable to set striping with a starting offset beyond OST 160
4877 Details : llapi_create_file() incorrectly limited the starting stripe
4878 index to the maximum single-file stripe count.
4881 Frequency : LDAP users only
4883 Description: lconf did not handle in-kernel recovery with LDAP properly
4884 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4886 Severity : enhancement
4888 Description: bind OST threads to NUMA nodes to improve performance
4889 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4890 node and do their allocations there to localize memory access
4892 Severity : enhancement
4894 Description: llmount can determine client NID directly from Myrinet (GM)
4895 Details : the client NID code from gmnalnid was moved directly into
4896 llmount, removing the need to use this or specifying the
4897 client NID explicitly when mounting GM clients with zeroconf
4900 Frequency : if client is started with down MDS
4902 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4903 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4904 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4905 ptlrpc_import_delayed_req() because we didn't check if the
4906 request was interrupted, and we also didn't break out of the
4907 event loop if there was no timeout
4912 Description: data loss during non-page-aligned writes to a single file from
4913 both multiple nodes and multiple threads on one node at same time
4914 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4915 inconsistency led to false short-reads, that were cached and later
4916 used by ->prepare_write() to fill in partially written page,
4917 leading to data loss.
4920 Frequency : always, if lconf --abort_recovery used
4922 Description: lconf --abort_recovery fails with 'Operation not supported'
4923 Details : lconf was attempting to abort recovery on the MDT device and not
4926 Severity : enhancement
4928 Description: remove cleanup logs
4929 Details : replace lconf-generated cleanup logs with lustre internal
4930 cleanup routines. Eliminates the need for client-cleanup and
4933 Severity : enhancement
4935 Description: add support for EAs (user and system) on lustre filesystems
4936 Details : it is now possible to store extended attributes in the Lustre
4937 client filesystem, and with the user_xattr mount option it
4938 is possible to allow users to store EAs on their files also
4940 Severity : enhancement
4942 Description: Add possibility (config option) to show minimal available OST free
4944 Details : When compiled with --enable-mindf configure option, statfs(2)
4945 (and so, df) will return least minimal free space available from
4946 all OSTs as amount of free space on FS, instead of summary of
4947 free spaces of all OSTs.
4949 Severity : enhancement
4951 Description: do not expand extent locks acquired on OST-side
4952 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4953 by server: they are not cached anyway.
4956 Frequency : when mmap is used/binaries executed from Lustre
4958 Description: Unmmap pages before throwing them away from read cache.
4959 Details : llap_shrink cache now attempts to unmap pages before discarding
4960 them (if unmapping failed - do not discard). SLES9 kernel has
4961 extra checks that trigger if this unmapping is not done first.
4966 Description: lconf didn't resolve symlinks before checking to see whether a
4967 given mountpoint was already in use
4970 Frequency : when migrating failover services
4971 Bugzilla : 6395, 9514
4972 Description: When migrating a subset of services from a node (e.g. failback
4973 from a failover service node) the remaining services would
4974 time out and evict clients.
4975 Details : lconf --force (implied by --failover) sets the global obd_timeout
4976 to 5 seconds in order to quickly disconnect, but this caused
4977 other RPCs to time out too quickly. Do not change the global
4978 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4980 Severity : enhancement
4981 Frequency : if MDS is started with down OST
4982 Bugzilla : 9439,5706
4983 Description: Allow startup/shutdown of an MDS without depending on the
4984 availability of the OSTs.
4985 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4986 Add appropriate locking and lov-osc refcounts for safe
4987 cleaning. Add osc abort_inflight calls in case the
4988 synchronize never started.
4991 Frequency : occasional (Cray XT3 only)
4993 Description: root not authorized to access files in CRAY_PORTALS environment
4994 Details : The client process capabilities were not honoured on the MDS in
4995 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4996 been accessed by an authorized user then root was able to access
4997 the file on the local client also. The root user capabilities
4998 are now allowed on the MDS, as this environment has secure UID.
5001 Frequency : occasional
5003 Description: ldiskfs "too long searching" message happens too often
5004 Details : A debugging message (otherwise harmless) prints too often on
5005 the OST console. This has been reduced to only happen when
5006 there are fragmentation problems on the filesystem.
5011 Description: Division by zero in statfs when all OSCs are inactive
5012 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5013 lov_statfs divides by value returned by lov_get_stripecnt().
5017 Bugzilla : 9489, 3273
5018 Description: First write from each client to each OST was only 4kB in size,
5019 to initialize client writeback cache, which caused sub-optimal
5020 RPCs and poor layout on disk for the first written file.
5021 Details : Clients now request an initial cache grant at (re)connect time
5022 and so that they can start streaming writes to the cache right
5023 away and always do full-sized RPCs if there is enough data.
5024 If the OST is rebooted the client also re-establishes its grant
5025 so that client cached writes will be honoured under the grant.
5030 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5031 Details : Now I/O RPCs go to different portal number and (presumably) fast
5032 lock requests (and glimses) and other RPCs get their own service
5033 threads pool that should be able to service those RPCs
5036 Severity : enhancement
5038 Description: Ability to exchange lustre version between client and servers and
5039 issue warnings at client side if client is too old. Also for
5040 liblustre clients there is ability to refuse connection of too old
5042 Details : New 'version' field is added to connect data structure that is
5043 filled with version info. That info is later checked by server and
5047 Frequency : rare, liblustre only.
5048 Bugzilla : 9296, 9581
5049 Description: Two simultaneous writes from liblustre at offset within same page
5050 might proceed at the same time overwriting eachother with stale
5052 Details : I/O lock withing llu_file_prwv was released too early, before data
5053 actually was hitting the wire. Extended lock-holding time until
5054 server acknowledges receiving data.
5057 Frequency : extremely rare. Never observed in practice.
5059 Description: avoid generating lustre_handle cookie of 0.
5060 Details : class_handle_hash() generates handle cookies by incrementing
5061 global counter, and can hit 0 occasionaly (this is unlikely, but
5062 not impossible, because initial value of cookie counter is
5063 selected randonly). Value of 0 is used as a sentinel meaning
5064 "unassigned handle" --- avoid it. Also coalesce two critical
5065 sections in this function into one.
5067 Severity : enhancement
5069 Description: allow liblustre clients to delegate truncate locking to OST
5070 Details : To avoid overhead of locking, liblustre client instructs OST to
5071 take extent lock in ost_punch() on client's behalf. New connection
5072 flag is added to handle backward compatibility.
5074 Severity : enhancement
5075 Bugzilla : 4928, 7341, 9758
5076 Description: allow number of OST service threads to be specified
5077 Details : a module parameter allows the number of OST service threads
5078 to be specified via "options ost ost_num_threads={N}" in the
5079 OSS's /etc/modules.conf or /etc/modprobe.conf.
5083 Bugzilla : 6146, 9635, 9895
5084 Description: servers crash with bad pointer in target_handle_connect()
5085 Details : In rare cases when a client is reconnecting it was possible that
5086 the connection request was the last reference for that export.
5087 We would temporarily drop the export reference and get a new
5088 one, but this may have been the last reference and the export
5089 was just destroyed. Get new reference before dropping old one.
5091 Severity : enhancement
5092 Frequency : if client is started with failover MDS
5094 Description: Allow multiple MDS hostnames in the mount command
5095 Details : Try to read the configuration from all specified MDS
5096 hostnames during a client mount in case the "primary"
5099 Severity : enhancement
5101 Description: Stop sending data to evicted clients as soon as possible.
5102 Details : Check if the client we are about to send or are sending data to
5103 was evicted already. (Check is done every second of waiting,
5104 for which l_wait_event interface was extended to allow checking
5105 of exit condition at specified intervals).
5108 Frequency : rare, normally only when NFS exporting is done from client
5110 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5112 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5115 Frequency : always, if srand() is called before liblustre initialization
5117 Description: Liblustre uses system PRNG disturbing its usage by user application
5118 Details : Introduce internal to lustre fast and high-quality PRNG for
5119 lustre usage and make liblustre and some other places in generic
5120 lustre code to use it.
5122 Severity : enhancement
5123 Bugzilla : 9477, 9557, 9870
5124 Description: Verify that the MDS configuration logs are updated when xml is
5125 Details : Check if the .xml configuration logs are newer than the config
5126 logs stored on the MDS and report an error if this is the case.
5127 Request --write-conf, or allow starting with --old_conf.
5129 Severity : enhancement
5131 Description: Handle symlinks in the path when checking if Lustre is mounted.
5132 Details : Resolve intermediate symlinks when checking if a client has
5133 mounted a filesystem to avoid duplicate client mounts.
5138 Description: lconf can hit an error exception but still return success.
5139 Details : The lconf command catches the Command error exception at the top
5140 level script context and will exit with the associated exit
5141 status, but doesn't ensure that this exit status is non-zero.
5146 Description: failure of ptlrpc thread startup can cause oops
5147 Details : Starting a ptlrpc service thread can fail if there are a large
5148 number of threads or the server memory is very fragmented.
5149 Handle this without oopsing.
5152 Frequency : always, only if liblustre and non-default acceptor port was used
5154 Description: liblustre cannot connect to servers with non-default acceptor port
5155 Details : tcpnal_set_default_params() was not called and was therefore
5156 ignoring the environment varaible TCPNAL_PORT, as well as other
5157 TCPNAL_ environment variables
5162 Description: two objects could be created on the same OST for a single file
5163 Details : If an OST is down, in some cases it was possible to create two
5164 objects on a single OST for a single file. No problems other
5165 than potential performance impact and spurious error messages.
5169 Bugzilla : 5681, 9562
5170 Description: Client may oops in ll_unhash_aliases
5171 Details : Client dcache may become inconsistent in race condition.
5172 In some cases "getcwd" can fail if the current directory is
5178 Description: Inode refcounting problems in NFS export code
5179 Details : link_raw functions used to call d_instantiate without obtaining
5180 extra inode reference first.
5184 Bugzilla : 9942, 9903
5185 Description: Referencing freed requests leading to crash, memleaks with NFS.
5186 Details : We used to require that call to ll_revalidate_it was always
5187 followed by ll_lookup_it. Also with revalidate_special() it is
5188 possible to call ll_revalidate_it() twice for the same dentry
5189 even if first occurence returned success. This fix changes semantic
5190 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5191 reference on a request referred from the intent.
5192 ll_intent_release() then releases such a request.
5195 Frequency : rare, normally benchmark loads only
5197 Description: unlinked inodes were kept in memory on the client
5198 Details : If a client is repeatedly creating and unlinking files it
5199 can accumulate a lot of stale inodes in the inode slab cache.
5200 If there is no other client load running this can cause the
5201 client node to run out of memory. Instead flush old inodes
5202 from client cache that have the same inode number as a new inode.
5205 Frequency : SLES9 2.6.5 kernel and long filenames only
5206 Bugzilla : 9969, 10379
5207 Description: utime reports stale NFS file handle
5208 Details : SLES9 uses out-of-dentry names in some cases, which confused
5209 the lustre dentry revalidation. Change it to always use the
5213 Frequency : rare, unless heavy write-truncate concurrency is continuous
5214 Bugzilla : 4180, 6984, 7171, 9963, 9331
5215 Description: OST becomes very slow and/or deadlocked during object unlink
5216 Details : filter_destroy() was holding onto the parent directory lock
5217 while truncating+unlinking objects. For very large objects this
5218 may block other threads for a long time and slow overall OST
5219 responsiveness. It may also be possible to get a lock ordering
5220 deadlock in this case, or run out of journal credits because of
5221 the combined truncate+unlink. Solution is to do object truncate
5222 first in one transaction without parent lock, and then do the
5223 final unlink in a new transaction with the parent lock. This
5224 reduces the lock hold time dramatically.
5227 Frequency : rare, 2.4 kernels only
5229 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5230 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5231 before unmounting their filesystem in order to avoid deadlock.
5232 The kernel_locked() function in 2.4 kernels only checks whether
5233 the kernel lock is held, not whether it is this process that is
5234 holding it as 2.6 kernels do.
5239 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5240 Details : The client ptlrpc code may be trying to reconnect to a down
5241 server before a previous connection attempt has timed out.
5242 Increase the reconnect interval to be longer than the connection
5243 timeout interval to avoid sending duplicate connections to
5247 Frequency : echo_client brw_test command
5249 Description: fix echo_client to work with OST preallocated code
5250 Details : OST preallocation code (5137) didn't take echo_client IO path
5251 into account: echo_client calls filter methods outside of any
5252 OST thread and, hence, there is no per-thread preallocated
5253 pages and buffers to use. Solution: hijack pga pages for IO. As
5254 a byproduct, this avoids unnecessary data copying.
5258 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5259 Description: Client can oops in mdc_commit_close() after open replay
5260 Details : It was possible for the MDS to return an open request with no
5261 transaction number in mds_finish_transno() if the client was
5262 evicted, but without actually returning an error. Clients
5263 would later try to replay that open and may trip an assertion
5264 Simplify the client close codepath, and always return an error
5265 from the MDS in case the open is not successful.
5268 Frequency : rare, 2.6 OSTs only
5270 Description: OST may deadlock under high load on fragmented files
5271 Details : If there was a heavy load and highly-fragmented OST filesystems
5272 it was possible to have all the OST threads deadlock waiting on
5273 allocation of biovecs, because the biovecs were not released
5274 until the entire RPC IO was completed. Instead, release biovecs
5275 as soon as they are complete to ensure forward IO progress.
5277 Severity : enhancement
5279 Description: Support for specifying external journal device at mount
5280 Details : If an OST or MDS device is formatted with an external journal
5281 device, this device major/minor is stored in the ext3 superblock
5282 and may not be valid for failover. Allow detecting and
5283 specifying the external journal at mount time.
5288 Description: Mounting an MDS with pending unlinked files may cause oops
5289 Details : target_finish_recovery() calls mds_postrecov() which returned
5290 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5291 considers this an error and immediately begins cleaning up the
5292 lov, just after starting the mds_lov process
5294 Severity : enhancement
5296 Description: Implement 'lfs df' to report actual free space on per-OST basis
5297 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5298 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5299 usage of objects; '-h' to report in human readable format.
5301 ------------------------------------------------------------------------------
5303 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5310 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5311 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5312 at startup during orphan unlinking. Instead, skip buddy update
5313 before setup, as it will be regenerated after recovery is complete.
5318 Description: If an OST is inactive, its locks might reference stale inodes.
5319 Details : lov_change_cbdata() must iterate over all namespaces, even if
5320 they are inactive to clear inode references from the lock.
5322 Severity : enhancement
5323 Frequency : occasional, if non-standard max_dirty_mb used
5325 Description: Client will block write RPCs if not enough grant
5326 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5327 then the client will block writes while waiting for another RPC
5328 to complete instead of consuming its dirty limit. With change
5329 we get improved performance when max_dirty_mb is small.
5331 Severity : enhancement
5332 Bugzilla : 3389, 6253
5333 Description: Add support for supplementary groups on the MDS.
5334 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5335 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5336 lookups for user supplementary groups into a cache.
5340 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5341 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5342 for liblustre O_EXCL is enforced by MDS.
5346 Description: readdir never returns NULL in liblustre.
5347 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5348 dentry in struct dirent.
5352 Frequency : liblustre only, depends on application IO pattern
5353 Description: liblustre clients evicted if not contacting servers
5354 Details : Don't put liblustre clients into the ping_evictor list, so
5355 they will not be evicted by the pinger ever.
5357 Severity : enhancement
5359 Description: Add ability to evict clients by NID from MDS.
5360 Details : By echoing "nid:$NID" string into
5361 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5362 $NID would be instantly evicted from this MDS and from all active
5363 OSTs connected to it.
5367 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5368 Details : lookup_it_finish() used to query file size from OSTs that was not
5373 Description: service threads change working directory to that of init
5374 Details : Starting lustre service threads may pin the working directory
5375 of the parent thread, making that filesystem busy. Threads
5376 now change to the working directory of init to avoid this.
5380 Frequency : during shutdown only
5381 Description: shutdown with a failed MDS or OST can cause unmount to hang
5382 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5387 Frequency : 2.6 only
5388 Description: chmod/chown may include an extra supplementary group
5389 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5390 supplementary group and if none is specified this is used.
5393 Bugzilla : 5479 (6816)
5395 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5396 Details : If lookup is in progress on a file that is unlinked we might try
5397 to revalidate the inode and fail in revalidate after lookup is
5398 complete and ll_file_open() enqueues the open again but
5399 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5402 Frequency : always, if lconf --abort_recovery used
5404 Description: lconf --abort_recovery fails with 'Operation not supported'
5405 Details : lconf was attempting to abort recovery on the MDT device and not
5408 ------------------------------------------------------------------------------
5410 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5415 Frequency : rare (only unsupported configurations with a node running as an
5417 Bugzilla : 6514, 5137
5418 Description: Mounting a Lustre file system on a node running as an OST could
5420 Details : OSTs now preallocates memory needed to write out data at
5421 startup, instead of when needed, to avoid having to
5422 allocate memory in possibly low memory situations.
5423 Specifically, if the file system is mounted on on OST,
5424 memory pressure could force it to try to write out data,
5425 which it needed to allocate memory to do. Due to the low
5426 memory, it would be unable to do so and the node would
5427 become unresponsive.
5429 Severity : enhancement
5431 Description: Addition of lconf --service command line option
5432 Details : lconf now accepts a '--service <arg>' option, which is
5433 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5435 Severity : enhancement
5437 Description: Failover mode is now the default for OSTs.
5438 Details : By default, OSTs will now run in failover mode. To return to
5439 the old behaviour, add '--failout' to the lmc line for OSTs.
5441 Severity : enhancement
5443 Description: Health checks are now provided for MDS and OSTs
5444 Details : Additional detailed health check information on MSD and OSTs
5445 is now provided through the procfs health_check value.
5448 Frequency : occasional, depends on IO load
5450 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5451 after numerous create/delete cycles
5452 Details : The ext3 inode allocation policy would not allocate new inodes
5453 very well on the OSTs because there are no new directories
5454 being created. Instead we look for groups with free space if
5455 the parent directories are nearly full.
5460 Description: Network or server problems during mount may cause partially
5461 mounted clients instead of returning an error.
5462 Details : The config llog parsing code may overwrite the error return
5463 code during mount error handling, returning success instead
5469 Description: MDS can fail to allocate large reply buffers
5470 Details : After long uptimes the MDS can fail to allocate large reply
5471 buffers (e.g. zconf client mount config records) due to memory
5472 fragmentation or consumption by the buffer cache. Preallocate
5473 some large reply buffers so that these replies can be sent even
5474 under memory pressure.
5478 Frequency : rare (liblustre)
5479 Description: fsx running with liblustre complained that using truncate() to
5480 extend the file doesn't work. This patch corrects that issue.
5481 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5482 ATTR_SIZE and lsm use in llu_setattr_raw.
5486 Frequency : rare, only 2.6 kernels
5487 Description: Unusual file access patterns on the MDS may result in inode
5488 data being lost in very rare circumstances.
5489 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5490 code in the 2.6 kernel caused the inode and/or EA data not to
5491 be read from disk, causing single-file corruption.
5495 Frequency : rare, only 2.6 filesystems using extents
5496 Description: Heavy concurrent write and delete load may cause data corruption.
5497 Details : It was possible under high-load situations to have an extent
5498 metadata block in the block device cache from a just-unlinked
5499 file overwrite a newly-allocated data block. We now unmap any
5500 metadata buffers that alias just-allocated data blocks.
5504 Frequency : filesystems with default stripe_count larger than 77
5505 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5506 Details : lconf specifies an inode size of 4096 bytes when the default
5507 stripe_count is larger than 77. This conflicts with the default
5508 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5509 this case to avoid pinning too much memory for large EAs.
5511 ------------------------------------------------------------------------------
5513 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5518 Frequency : rare (extremely heavy IO load with hundreds of clients)
5520 Description: Client is evicted, gets IO error writing to file
5521 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5522 caused clients to be evicted for AST timeouts. The fixes in
5523 bug 5192 mean we no longer need to have such short AST timeouts
5524 so ldlm_timeout has been increased.
5527 Frequency : occasional during --force or --failover shutdown under load
5528 Bugzilla : 5949, 4834
5529 Description: Server oops/LBUG if stopped with --force or --failover under load
5530 Details : a collection of import/export refcount and cleanup ordering
5531 issues fixed for safer force cleanup
5534 Frequency : only filesystems larger than 120 OSTs
5535 Bugzilla : 5990, 6223
5536 Description: lfs getstripe would oops on a very large filesystem
5537 Details : lov_getconfig used kfree on vmalloc'd memory
5540 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5541 Bugzilla : 6242, 6243
5542 Description: reading from files that had been truncated to a non-zero size
5543 but never opened returned no data
5544 Details : ll_file_read() reads zeros from no-object files to EOF
5549 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5550 Details : There's little harm in aborting MDS/OSS recovery and letting it
5551 try again, so I removed the LASSERT and return an error instead.
5553 Severity : enhancement
5555 Description: New debugging infrastructure for tracking down data corruption
5556 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5557 (b) cover more of the client-side code path, and (c) try to narrow
5558 down where problems occurred
5562 Bugzilla : 3819, 4364, 4397, 6313
5563 Description: Racing close and eviction MDS could cause assertion in mds_close
5564 Details : It was possible to get multiple mfd references during close and
5565 client eviction, leading to one thread referencing a freed mfd.
5567 Severity: : enhancement
5568 Bugzilla : 3262, 6359
5569 Description: Attempts to reconnect to servers are now more aggressive.
5570 Details : This builds on the enhanced upcall-less recovery that was added
5571 in 1.4.2. When trying to reconnect to servers, clients will
5572 now try each server in the failover group every 10 seconds. By
5573 default, clients would previously try one server every 25 seconds.
5578 Description: After recovery, certain operations trigger a failed
5579 assertion on a client.
5580 Details : Failing over an mds, using lconf -d --failover, while a
5581 client was doing a readdir() call would cause the client to
5582 LBUG after recovery completed and the readdir() was resent.
5584 Severity : enhancement
5586 Description: Default groups are now added by lconf
5587 Details : You can now run lconf --group <servicename> without having to
5588 manually add groups with lmc.
5591 Frequency : occasional
5593 Description: Nodes with an elan id of 0 trigger a failed assertion
5596 Frequency : always when accessing e.g. tty/console device nodes
5598 Description: tty and some other devices nodes cannot be used on lustre
5599 Details : file's private_data field is used by device data and lustre
5600 values in there got lost. New field was added to struct file to
5601 store fs-specific private data.
5604 Frequency : when exporting Lustre via NFS
5606 Description: NFSD failed occasionally when looking up a path component
5607 Details : NFSD is looking up ".." which was broken in ext3 directories
5608 that had grown large enough to become hashed.
5611 Frequency : Clusters with multiple interfaces not on the same subnet
5613 Description: Nodes will repeatedly try to reconnect to an interface which it
5614 cannot reach and report an error to the log.
5615 Details : Extra peer list entries will be created by lconf with some peers
5616 unreachable. lconf now validates the peer before adding it.
5619 Frequency : Only if a default stripe is set on the filesystem root.
5621 Description: Setting a default stripe on the filesystem root prevented the
5622 filesystem from being remounted.
5623 Details : The client was sending extra request flags in the root getattr
5624 request and did not allocate a reply buffer for the dir EA.
5627 Frequency : occasional, higher if lots of files are accessed by one client
5628 Bugzilla : 6159, 6097
5629 Description: Client trips assertion regarding lsm mismatch/magic
5630 Details : While revalidating inodes the VFS looks up inodes with ifind()
5631 and in rare cases can find an inode that is being freed.
5632 The ll_test_inode() code will free the lsm during ifind()
5633 when it finds an existing inode and then the VFS later attaches
5634 this free lsm to a new inode.
5638 Bugzilla : 6422, 7030
5639 Description: MDS deadlock between mkdir and client eviction
5640 Details : Creating a new file via mkdir or mknod (starting a transaction
5641 and getting the ns lock) can deadlock with client eviction
5642 (gets ns lock and trying to finish a synchronous transaction).
5645 Frequency : occasional
5646 Description: While starting a server, the fsfilt_ext3 module could not be
5648 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5649 kernels. Previously, lconf would still use the ext3 name
5650 when trying to load modules. Now, it will correctly use
5651 ext3 on 2.4 and ldiskfs on 2.6.
5653 Severity : enhancement
5654 Description: The default stripe count has been changed to 1
5655 Details : The interpretation of the default stripe count (0, to lfs
5656 or lmc) has been changed to mean striping across a single
5657 OST, rather than all available. For general usage we have
5658 found a stripe count of 1 or 2 works best.
5660 Severity : enhancement
5661 Description: Add support for compiling against Cray portals.
5662 Details : Conditional compiling for some areas that are different
5666 Frequency : occasional
5667 Bugzilla : 6409, 6834
5668 Description: Creating files with an explicit stripe count may lead to
5669 a failed assertion on the MDS
5670 Details : If some OSTs are full or unavailable, creating files may
5671 trigger a failed assertion on the MDS. Now, Lustre will
5672 try to use other servers or return an error to the
5676 Frequency : occasional
5678 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5679 caused lock timeout during readahead (since 1.4.2).
5680 Details : Processes doing readahead might match a lock that hasn't been
5681 granted yet if there are overlapping and conflicting lock
5682 requests. The readahead process waits on ungranted lock
5683 (original lock is CBPENDING), while OST waits for that process
5684 to cancel CBPENDING read lock and eventually evicts client.
5686 Severity : enhancement
5688 Description: Initial enabling of flock support for clients
5689 Details : Implements fcntl advisory locking and file status functions.
5690 This feature is provided as an optional mount flag (default
5691 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5692 locking are implemented yet, and those that are are not guaranteed
5693 to be completely correct in production environments.
5694 mount -t lustre -o [flock|noflock] ...
5697 Frequency : occasional
5699 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5700 assertion in the ext3 JBD (journaling) layer.
5701 Details : The b_committed_data struct is protected by the big kernel lock
5702 in 2.4 kernels, serializing journal_commit_transaction() and
5703 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5704 access to this struct. In 2.6 kernels there is finer grained
5705 locking to improve SMP performance of the JBD layer.
5709 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5714 Description: merge and cleanup kernel patches.
5717 -----------------------------------------------------------------------------
5719 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5721 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5722 versions of Lustre. Please update all servers and clients to
5723 version 1.4.2 or later at the same time. You must also run
5724 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5725 to update the configuration logs.
5727 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5728 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5729 - swab llog records of type '0' so we get proper header size/idx (5861)
5730 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5731 - fix rename of one directory over another leaking an inode (5953)
5732 - avoid SetPageDirty on 2.6 (5981)
5733 - don't re-add just-being-destroyed locks to the waiting list (5653)
5734 - when creating new directories, inherit the parent's custom
5735 striping settings if present parent (3048)
5736 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5737 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5738 - don't hold client locks on temporary worklist from l_lru (5666)
5739 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5740 - automatically evict dead clients (5921)
5741 - Update file size properly in create+truncate+fstat case (6196)
5742 - Do not unhash mountpoint dentries, do not allow removal of
5744 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5745 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5746 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5747 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5748 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5749 - Correctly handle reads of files with no objects (6243)
5750 - lctl recover will also mark a device active if deactivate used (5933)
5752 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5753 - allow --write-conf on an MDS with different nettype than client (5619)
5754 - don't write config llogs to MDS for mounts not from that MDS (5617)
5755 - lconf should create multiple TCP connections from a client (5201)
5756 - init scripts are now turned off by default; run chkconfig --on
5757 lustre and chkconfig --on lustrefs to use them
5758 - upcalls are no longer needed for clients to recover to failover
5760 - add --abort-recovery option to lconf to abort recovery on device
5762 - add support for an arbitrary number of OSTs (3026)
5763 - Quota support protocol changes.
5764 - forward compatibility changes to wire structs (6007)
5765 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5766 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5767 - improved "lustre status" script
5768 - initialize blocksize for non-regular files (6062)
5769 - added --disable-server and --disable-client configure options (5782)
5770 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5771 - Vanilla 2.4.29 support
5772 - increase maximum number of obd devices to 520 (6242)
5773 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5774 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5775 SLES 9 2.6.5 kernels
5776 - limit stripes per file to 160 (the maximum EA size) (6093)
5778 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5781 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5782 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5783 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5784 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5785 - small ext3 extents cleanups and fixes (5733)
5786 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5787 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5788 - avoid needless client->OST connect, fix handle mismatch (5317)
5789 - fix DLM error path that led to out-of-sync client, long delays (5779)
5790 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5791 - fix several locking issues related to i_size (5492,5624,5654,5672)
5792 - don't move pending lock onto export if it is already evicted (5683)
5793 - fix kernel oops when creating .foo in unlinked directory (5548)
5794 - fix deadlock in obdfilter statistics vs. object create (5811)
5795 - use time_{before,after} to avoid timer jiffies wrap (5882)
5796 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5797 - Do not leak request if server was not able to process it (5154)
5798 - If mds_open unable to find parent dir, make that negative lookup(5154)
5799 - don't create new directories with extent-mapping (5909, 5936)
5801 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5802 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5803 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5804 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5805 - support for RHEL4 kernel on i686 (5773)
5806 - provide error messages when incompatible logs are encountered (5898)
5808 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5809 * version 1.4.0.10 (1.4.1 release candidate 1)
5811 - don't keep a lock reference when lock is not granted (4238)
5812 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5813 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5814 - drop import inflight refcount on signal_completed_replay error (5255)
5815 - unlock page after async write error during send (3677)
5816 - handle missing objects in filter_preprw_read properly (5265)
5817 - no transno return for symlink open, don't save no-trasno open (3440)
5818 - don't try to complete elan receive that already failed (4012)
5819 - free RPC server reply state on error (5406)
5820 - clean up thread from ptlrpc_start_thread() on error (5160)
5821 - readahead could read extra page into cache that wasn't ejected (5388)
5822 - prevent races in class_attach/setup/cleanup/detach (5260)
5823 - don't dereference de->d_inode after l_dput of de (5458)
5824 - use "int" for stripe value returned from lock_to_stripe (5544)
5825 - mballoc allocation and error-checking fixes in 2.6 (5504)
5826 - block device patches to fix I/O request sizes in 2.6 (5482)
5827 - look up hostnames for IB nals (5602)
5828 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5829 - don't start multiple acceptors for the same port (5277)
5830 - fix incorrect LASSERT in mds_getattr_name (5635)
5831 - export a proc file for general "ping" checking (5628)
5832 - fix "lfs check" to not block when the MDS is down (5628)
5834 - service request history (4965)
5835 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5836 - create an "evict_client" /proc entry on OSTs, like the MDS has
5837 - fix mount usage message, return errors per mount(8) (5168)
5838 - change grep [] to grep "[]" in tests so they work in more UMLs
5839 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5840 - remove ancient LOV_MAGIC_V0 EA support (5047)
5841 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5842 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5843 - allow readahead allocations to fail when low on memory (5383)
5844 - mmap locking landed again, after considerable improvement (2828)
5845 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5847 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5850 - send OST transaction number in read/write reply to free req (4966)
5851 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5852 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5853 - the watchdog thread now runs as interruptible (5246)
5854 - flock/lockf fixes (but it's still disabled, pending 5135)
5855 - don't use EXT3 constants in llite code (5094)
5856 - memory shortage at startup could cause assertion (5176)
5858 - reorganization of lov code
5859 - single portals codebase
5861 - add extents/mballoc support (5025)
5862 - direct I/O reads in the obdfilter (4048)
5863 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5865 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5868 - send OST transaction number in read/write reply to free req (4966)
5869 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5870 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5871 - the watchdog thread now runs as interruptible (5246)
5872 - handle missing objects in filter_preprw_read properly (5265)
5873 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5874 - drop import inflight refcount on signal_completed_replay error (5255)
5875 - unlock page after async write error during send (3677)
5876 - return original error code on reconstructed replies (3761)
5877 - no transno return for symlink open, don't save no-trasno open (3440)
5879 - add pid to ldlm debugging output (4922)
5880 - bump the watchdog timeouts -- we can't handle 30sec yet
5881 - extra debugging for orphan dentry/inode bug (5259)
5883 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5886 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5887 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5888 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5889 - fix NULL dereference in /proc/sys/portals/routes (4827)
5890 - allow failed mdc_close() operations to be interrupted (4561)
5891 - stop precreate on OST before MDS would time out on it (4778)
5892 - don't send partial-page writes before EOF from client (4410)
5893 - discard client grant for sub-page writes on large-page clients (4520)
5894 - don't free dentries not owned by NFS code, check generation (4806)
5895 - fix lsm leak if mds_create_objects() fails (4801)
5896 - limit debug_daemon file size, always print CERROR messages (4789)
5897 - use transno after validating reply (3892)
5898 - process timed out requests if import state changes (3754)
5899 - update mtime on OST during writes, return in glimpse (4829)
5900 - add mkfsoptions to LDAP (4679)
5901 - use ->max_readahead method instead of zapping global ra (5039)
5902 - don't interrupt __l_wait_event() during strace
5904 - add software watchdogs to catch hung threads quickly (4941)
5905 - make lustrefs init script start after nfs is mounted
5906 - fix CWARN/ERROR duplication (4930)
5907 - return async write errors to application if possible (2248)
5908 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5909 - print NAL number in %x format (4645)
5910 - update barely-supported suse-2.4.21-171 series (4842)
5911 - support for sles 9 %post scripts
5912 - support for building 2.6 kernel-source packages
5913 - support for sles km_* packages
5915 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5918 - ignore -ENOENT errors in osc_destroy (3639)
5919 - notify osc create thread that OSC is being cleaned up (4600)
5920 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5921 - reconstruct ost_handle() like mds_handle() (4657)
5922 - create a new thread to do import eviction to avoid deadlock (3969)
5923 - let lconf resolve symlinked-to devices (4629)
5924 - don't unlink "objects" from directory with default EA (4554)
5925 - hold socknal file ref over connect in case target is down (4394)
5926 - allow more than 32000 subdirectories in a single directory (3244)
5927 - fix blocks count for O_DIRECT writes (3751)
5928 - OST returns ENOSPC from object create when no space left (4539)
5929 - don't send truncate RPC if file size isn't changing (4410)
5930 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5931 - bind to privileged port in socknal and tcpnal (3689)
5933 - rate limit CERROR/CWARN console message to avoid overload (4519)
5934 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5935 - basic mmap support (3918)
5936 - kernel patch series update from b1_4 (4711)
5938 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5941 - avoid crash during MDS cleanup with OST shut down (2775)
5942 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5943 - don't use bad inodes on the MDS (3744)
5944 - dynamic object preallocation to improve recovery speed (4236)
5945 - don't hold spinlock over lock dumping or change debug flags (4401)
5946 - don't zero obd_dev when it is force cleaned (3651)
5947 - print grants to console if they go negative (4431)
5948 - "lctl deactivate" will stop automatic recovery attempts (3406)
5949 - look for existing locks in ldlm_handle_enqueue() (3764)
5950 - don't resolve lock handle twice in recovery avoiding race (4401)
5951 - revalidate should check working dir is a directory (4134)
5953 - don't always mark "slow" obdfilter messages as errors (4418)
5955 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5958 - don't close LustreDB during write_conf until it is done (3860)
5959 - fix typo in lconf for_each_profile (3821)
5960 - allow dumping logs from multiple threads at one time (3820)
5961 - don't allow multiple threads in OSC recovery (3812)
5962 - fix debug_size parameters (3864)
5963 - fix mds_postrecov to initialize import for llog ctxt (3121)
5964 - replace config semaphore with spinlock (3306)
5965 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5966 - don't allow enqueue to complete on a destroyed export (3822)
5967 - down write_lock before checking llog header bitmap (3825)
5968 - recover from lock replay timeout (3764)
5969 - up llog sem before sending rpc (3652)
5970 - reduce ns lock hold times when setting kms (3267)
5971 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5972 - fix NULL deref and obd_dev leak on setup error (3312)
5973 - replace some LBUG about llog ops with error handling (3841)
5974 - don't match INVALID dentries from d_lookup and spin (3784)
5975 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5976 - fix invalid assertion in ptlrpc_set_wait (3880)
5978 - add libwrap support for the TCP acceptor (3996)
5979 - add /proc/sys/portals/routes for non-root route listing (3994)
5980 - allow setting MDS UUID in .xml (2580)
5981 - print the stack of a process that LBUGs (4228)
5983 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5986 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5987 - make sure to unset replay flag from failed open requests (3440)
5988 - if default stripe count is 0, use OST count for inode size (3636)
5989 - update parent mtime/ctime on client for create/unlink (2611)
5990 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5991 - free recovery state on server during a forced cleanup (3571)
5992 - unregister_reply for resent reqs (3063)
5993 - loop back devices mounting and status check on 2.6 (3563)
5994 - fix resource-creation race that can provoke i_size == 0 (3513)
5995 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5996 - more debugging for page-accounting assertion (3746)
5997 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5998 - avoid deadlock after precreation failure (3758)
5999 - fix race and lock order deadlock in orphan handling (3450, 3750)
6000 - add validity checks when grabbing inodes from l_ast_data (3599)
6002 - add /proc/.../recovery_status to obdfilter (3428)
6003 - lightweight CDEBUG infrastructure, debug daemon (3668)
6004 - change default OSC RPC parameters to be better on small clusters
6005 - turn off OST read cache for files smaller than 32MB
6006 - install man pages and include them in rpms (3100)
6007 - add new init script for (un)mounting lustre filesystems (2593)
6008 - run chkconfig in %post for init scripts (3701)
6009 - drop scimac NAL (unmaintained)
6011 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6014 - clean kiobufs before and after use (3485)
6015 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6016 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6017 - change init script to fail more clearly if not run as root (1528)
6018 - allow clients to reconnect during replay (1742)
6019 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6020 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6021 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6022 - fix cond_resched (3554)
6023 - only evict unfinished clients after recovery (3515)
6024 - allow bulk resend, prevent data loss (3570)
6025 - dynamic ptlrpc request buffer allocation (2102)
6026 - don't allow unlinking open directory if it isn't empty (2904)
6027 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6028 - remove extraneous obd dereference causing LASSERT failure (3334)
6029 - don't use get_cycles() when creating temp. files on the mds (3156)
6030 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6031 - handle EEXIST for set-stripe, set proper directory name (3336)
6033 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6034 - fix ksocknal_fmb_callback() error messages (2918)
6036 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6039 - don't copy lvb into (possibly NULL) reply on error (2983)
6040 - don't deref dentry after dput, don't free lvb on error (2922)
6041 - use the kms to determine writeback rpc length (2947)
6042 - increment oti_logcookies when osc is inactive (2948)
6043 - update client's i_blocks count via lvb messages (2543)
6044 - handle intent open/close of special files properly (1557)
6045 - mount MDS with errors=remount-ro, like obdfilter (2009)
6046 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6047 - don't use cancelling-locks' kms values (2947)
6048 - use highest lock extent for kms, not last one (2925)
6049 - don't dereference ERR_PTR() dentry in error handling path (3107)
6050 - fix thread race in portals_debug_dumplog() (3122)
6051 - create lprocfs device entries at setup instead of at attach (1519)
6052 - common AST error handler, don't evict client on completion race (3145)
6053 - zero nameidata in detach_mnt in 2.6 (3118)
6054 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6055 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6056 - initialize RPC timeout timer earlier for 2.6 (3219)
6057 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6058 - print nal/nid for unknown nid (3258)
6059 - additional checks for oscc recovery before doing precreate (3284)
6060 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6061 - don't crash in mdc_close for bad permissions on open (3285)
6062 - zero i_rdev for non-device files (3147)
6063 - clear page->private before handing to FS, better assertion (3119)
6064 - tune the read pipeline (3236)
6065 - fix incorrect decref of invalidated dentry (2350)
6066 - provide read-ahead stats and refine rpc in flight stats (3328)
6067 - don't hold journal transaction open across create RPC (3313)
6068 - update atime on MDS at close time (3265)
6069 - close LDAP connection when recovering to avoid server load (3315)
6070 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6071 - don't leak open file on MDS after open resend (3325)
6072 - serialize filter_precreate and filter_destroy_precreated (3329)
6073 - loop device shouldn't call sync_dev() for nul device (3092)
6074 - clear page cache after eviction (2766)
6075 - resynchronize MDS->OST in background (2824)
6076 - refuse to mount the same filesystem twice on same mountpoint (3394)
6077 - allow llmount to create routes for mounting behind routers (3320)
6078 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6079 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6080 - fix rare problem with rename on htree directories (3417)
6082 - allow default OST striping configuration per directory (1414)
6083 - fix compilation for qswnal for 2.6 kernels (3125)
6084 - increase maximum number of MDS request buffers for large systems
6085 - change liblustreapi to be useful for external progs like lfsck (3098)
6086 - increase local configuration timeout for slow disks (3353)
6087 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6089 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6092 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6093 - don't overwrite extent policy data in reply if lock was blocked (2901)
6094 - drop filter export grants atomically with removal from device (2663)
6095 - del obd_self_export from work_list in class_disconnect_exports (2908)
6096 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6097 - swab reply message in mdc_close, other PPC fixes (2464)
6098 - fix destroying of named logs (2325)
6099 - overwrite old logs when running lconf --write_conf (2264)
6100 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6101 - fix race in target_handle_connect (2898)
6102 - mds_reint_create() should take same inode create lock (2926)
6103 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6104 - don't close files for self_export to avoid uninitialized obd (2936)
6105 - allow MDS with the same name as client node (2939)
6106 - hold dentry reference for closed log files for unlink (2325)
6107 - reserve space for all logs during transactions (2059)
6108 - don't evict page beyond end of stripe extent (2925)
6109 - don't oops on a deleted current working directory (2399)
6110 - handle hard links to targets without a parent properly (2517)
6111 - don't dereference NULL lock when racing during eviction (2867)
6112 - don't grow lock extents when lots of conflicting locks (2919)
6114 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6117 - account for cache space usage on clients to avoid data loss (974)
6118 - lfsck support in lustre kernel code (2349)
6119 - reduce journal credits needed for BRW writes (2370)
6120 - orphan handling to avoid losing space on client/server crashes
6121 - ptlrpcd can be blocked, stopping ALL progress (2477)
6122 - use lock value blocks to assist in proper KMS, faster stat (1021)
6123 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6124 - recovery for initial connections (2355)
6125 - fixes for mds_cleanup_orphans (1934)
6126 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6127 - block all file creations until orphan recovery completes (1901)
6128 - client remove rq_connection from request struct (2423)
6129 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6130 - recovery timer race (2670)
6131 - mdc_close recovey bug (2532)
6132 - ptlrpc cleanup bug (2710)
6133 - mds timeout on local locks (2588)
6134 - namespace lock held during RPCs (2431)
6135 - handle interrupted sync write properly (2503)
6136 - don't try to handle a message that hasn't been replied to (2699)
6137 - client assert failure during cleanup after abort recovery (2701)
6138 - leak mdc device after failed mount (2712)
6139 - ptlrpc_check_set allows timedout requests to complete (2714)
6140 - wait for inflight reqs when ptlrpcd finishes (2710)
6141 - make sure unregistered services are removed from the srv_list
6142 - reset bulk XID's when resending them (caught by 1138 test)
6143 - unregister_bulk after timeout
6144 - fix lconf error (2694)
6145 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6146 - readahead locks pages, leaves pending causing memory pressure (2673)
6147 - increase OST request buffers to 4096 on large machines (2729)
6148 - fix up permission of existing directories in simple_mkdir (2661)
6149 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6150 - don't assign transno to errored transactions (2742)
6151 - don't delete objects on OST if given a bogus objid from MDS (2751)
6152 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6153 - if rq_no_resend, then timeout request after recovery (2432)
6154 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6155 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6156 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6157 - MDS recovery shouldn't delete objects that it has given out (2730)
6158 - if enqueue arrives after completion, don't clobber LVB (2819)
6159 - don't unlock pages twice when trigger_group_io returns error (2814)
6160 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6161 - don't write pages to disk if there was an error (1450)
6162 - don't ping imports that have recovery disabled (2676)
6163 - take buffered bytes into account when balancing socknal conn (2817)
6164 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6165 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6166 - remove little-endian swabbing of llog records (1987)
6167 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6168 - retry reposting request buffers if they fail (1191)
6169 - grow extent at grant time to avoid granting a revoked lock (2809)
6170 - lock revoke doesn't evict page if covered by a second lock (2765)
6171 - disable VM readahead to avoid reading outside lock extents (2805)
6173 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6174 - updated kernel patches for hp-2.4.20 kernel (2681)
6176 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6179 - fix truncated write corruption (2366)
6180 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6182 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6183 - ensure len(uuid) < 37 in lmc (1171)
6184 - fix ia64 OOPS in llog_test (2255)
6185 - zero end of page at obdfilter for partial page writes (2648)
6186 - don't leave stale dentries around after renames (bug 2428)
6187 - fix timeouts when evicting a client with a single lock held (2642)
6188 - set deadline for the initial HELLO message to drain (2634)
6189 - print out dotted-quad IP addresses in the socknal (2302)
6191 - additional debugging for MDS client eviction problem (2443)
6192 - fix mkfsoptions support for osts (2603, 2604)
6194 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6197 - add series for the vanilla 2.6.0 kernel
6198 - add series for the vanilla 2.4.24 kernel
6199 - add series for a cray x86/64 UL kernel drop
6200 - fix xattr patches for the vanilla 2.4.19 series
6202 - generate true UUIDs in lmc (1171)
6203 - have portals stack dumping break in UML (2466)
6204 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6205 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6206 - wake callback waiting threads on client eviction (2460)
6207 - Add --ptldebug and --subsystem to lmc (1719)
6208 - update assertion to allow safe interrupt allocation
6209 - set rq_no_resend for cancel requests (2432)
6210 - recalculate ptlrpcd timeout after resend (2494)
6211 - call vfs_rmdir when removing pending directories (2368)
6212 - fix renaming a file to itself (2429)
6213 - lmc creates a default one-stripe lov (2454)
6214 - expand procfs space to handle large clusters (2326)
6215 - increase UML stack to avoid overflow
6216 - update lconf's list of debug and subsystem masks
6217 - fix lfs find --obd (2510)
6218 - /proc tunable for disabling filter read caching (2591)
6219 - stop rpm packages from altering slapd.conf (2301)
6220 - disable nagle in the socknal under 0conf (2578)
6221 - choose mds inode size based on stripe count (2572)
6222 - fix kernel-source rpm problems (2516)
6224 - add --disable-doc to avoid pdf generation (2421)
6225 - update documentation, tests, type-os, comments
6226 - avoid format warnings on ia64
6227 - remove the TOE NAL
6228 - tiny code cleanups by removing unused fields
6230 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6233 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6234 - remove the most blatant lies from BUILDING file (2371)
6235 - change default debug level to reasonable production setting
6236 - reduce client side cache size to reduce cache flush time
6237 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6238 - make TCP zerocopy and pinger support enabled by default (2476)
6239 - sync writes completed after process exits caused crashes (2319)
6240 - maintain correct mount count on the MDS (2356)
6241 - backout 1557, because 2316 wasn't really fixed
6242 - better file I/O statistics gathering in /proc
6243 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6244 - another kernel patch to fix zero-copy TCP function export
6245 - don't take duplicate lock when processing re-sent getattr (2420)
6246 - lctl uses obd_self_export instead of creating new conn (2353)
6247 - MDS/OST recovery case which requires object creation asserted (2425)
6248 - move lfs from /usr/sbin to /usr/bin in packages
6249 - fix race between mds_client_add and mds_client_free (2417)
6250 - use kmalloc instead of slabs in portals (2430)
6251 - don't create duplicate records when a failover MDS is present (2442)
6252 - remove unnecessary mount age check (2332)
6253 - don't remove directory inodes from locks prematurely (2451)
6254 - don't break if MDS service name is the same as hostname (2103)
6255 - fix races in client write RPC generation when cache full (2482)
6257 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6260 - remove now-unused request->rq_obd (278)
6261 - if an allocation fails, print out how much memory we've used (1933)
6262 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6263 - add the "configurable stack size" patch to most series files (1256)
6264 - ability to write large log records, for 100+ OST configs (2306)
6265 - fix NULL deref when filter_prep fails (2314)
6266 - fix operator precedence error in filter_sync
6267 - dynamic allocation of socknal TX descriptors (2315)
6268 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6269 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6270 - ensure that CWARN messages in Portals always get to the syslog
6271 - __init/__exit are not for prototype decls (ldlm_init/exit)
6272 - x86-64 compile warning fixes
6273 - fix gateway LMC keyword conflict (2318)
6274 - fix MDS lock inversions in getattr/reint paths (1844)
6275 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6276 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6277 - DLM race condition prevented some lock evictions (2328)
6278 - ENOMEM detection and retry on socknal sends (2230)
6279 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6280 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6281 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6282 - better error messages when a client is rejected during recovery (1505)
6283 - avoid cancelling locks which were never granted, after failure (2330)
6284 - fix i_sem/journal inversion in mds_client_add (2333)
6285 - fix truncate/getattr lock cycle deadlock (2334)
6286 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6287 - fix two rare exit paths which could leak an l_lock() ref (2321)
6288 - fencepost error in MDS/OST orphan recovery (2226)
6289 - make log record alignment 8 bytes (1988)
6290 - lstripe now fails when requested offset > ost_count (2237)
6291 - ensure that all kernel series have a complete list.h (1607)
6292 - fix crashes in special-file operations (2316)
6293 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6295 - allow configurable automake binary, for testing new versions
6296 - small update to the lfs documentation
6298 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6300 * fix negative export reference count in fsfilt_sync (2312)
6302 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6303 * release candidate 0.9.1
6305 - orphans are moved into the PENDING directory for possible recovery
6306 - replayed opens now open by fid for orphan/rename safety (1042)
6307 - last close of an orphan inode generates a transno (683)
6308 - chdir() and mount() now pin the directory entry (1020)
6309 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6310 - discard very old requests without processing them (1502)
6311 - remove some common, well-understood CERRORs (1505)
6312 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6313 - clear "grant" flags in OST replies until OST grant code lands (1644)
6314 - fix read performance by not clobbering i_blksize on client (1598)
6315 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6316 - make LRU size a /proc tunable, clears locks when reduced (707)
6317 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6318 - server locks take references on exports now (1558)
6319 - build fixes for 2.4.20-rh trees (1663)
6320 - return an error from lov_create if all OSCs are inactive (1751)
6321 - fix import levels when a reconnect happens without a timeout (1597)
6322 - exit early from mds_open if we get a lookup error (1749)
6323 - partial page read at EOF wouldn't wait for disk before sending (1642)
6324 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6325 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6326 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6327 - write/truncate lock inversion (1639)
6328 - Don't auto-load obdclass, portals modules during cleanup (1495)
6329 - fix timestamps from jumping to "now" (1763)
6330 - extra journal assertions (1648)
6331 - add an extra multiunlink test (1771)
6332 - fix read_record/write_record API (1776)
6333 - fix leak of offset_extent, possible incorrect i_size later (1772)
6334 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6335 - Debugging for the kqswnal_get_idle_tx problems (1820)
6336 - Allow recovery to be attempted multiple times (1536)
6337 - Write out MDS last_rcvd file after it is first created (1600)
6338 - Fix tx_descriptor leak in failed transmit situations (1827)
6339 - ext3 journaling fixes for assertion failure after IO error (1871)
6340 - class_export_put() on freed export after completion AST error (1896)
6341 - Fix revalidate looping in VFS (1322)
6342 - Don't access a freed export during MDS_REINT timeout (1521)
6343 - Add open-unlink recovery support on the MDS (1673,1764)
6344 - Return an error if no MDS data was read from last_rcvd (1946)
6345 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6346 - Don't setup a disk device that doesn't match exported UUID (317)
6347 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6348 - avoid committing NULL handle in force close
6349 - local.sh is now a one-stripe LOV configuration
6350 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6351 - fix ext3 htree duplicate directory entry corruption (1516)
6352 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6353 - update to new LOV EA format (2097)
6354 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6356 2003-06-15 Phil Schwan <phil@clusterfs.com>
6359 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6360 - per-import/export recovery handling (958, 931, 959)
6361 - multiple last-rcvd slots, for serving multiple FSes (949)
6362 - connections are again shared between multiple imp/exports (963, 964)
6363 - "umount -f" would hang if any requests needed to be sent (393, 978)
6364 - avoid pinning large req buffer by copying for queued messages (989)
6365 - add "uuid" to "lctl device" command to help upcalls (991)
6366 - "open" RPCs with transnos would confuse recovery counters (1037)
6367 - do proper endian conversion of all wire messages (288, 340, 891)
6368 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6369 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6370 - fix ptlrpc_connection leak in target_handle_connect (1174)
6371 - fix import refcounting bug in OST and MDS cleanup (1134)
6372 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6373 - fix very unlikely obd_types race condition (501)
6374 - remove osc_open hack for echo_client (1187)
6375 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6376 - a failure in read_inode2 leads to deadlock (1139)
6377 - cancel ack-locks as soon as transaction is committed (1072)
6378 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6379 - make sure to commitrw after any preprw to avoid deadlock (1162)
6380 - failing to execute a file in a lustre FS would lock inode (1203)
6381 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6382 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6383 - obd_uuid2tgt would walk off the end of the list (1255)
6384 - on IA64 the osc would give portals incorrect bulk size (1258)
6385 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6386 - fix lock inversion caused by new llite matching code (1282)
6387 - limit the number of dirty pages on a client to 10MB (1286)
6388 - timed out locks were not being corrected cancelled (1289)
6389 - fix O_DIRECT above 4GB on IA-32 (1292)
6390 * major user-visible changes
6391 - fail out/fail over policy now controlled by the upcall (993)
6393 - add OBD_PING to check server availability and failure (954)
6394 - lustre messages are now sent in sending host order (288, 340, 891)
6395 - add eadatalen to MDS getattr reply (340)
6396 - OST read replies may contain second buffer, with per-page status (593)
6398 2003-03-11 Phil Schwan <phil@clusterfs.com>
6401 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6402 - failed open()s could cause deadlock; fixed (867, 869)
6403 - stop cancelling OST locks when files are closed (481)
6404 - overlapping XID spaces caused network corruption (851, 853)
6405 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6406 - setattr_raw added, to do single-RPC, server-side setattrs
6407 - lmc/lconf syntax change for OST UUIDs
6408 - fix crashy race condition between ptlrpc_free_req and osc_close
6409 - don't use request in mdc_enqueue if we hit a timeout (889)
6410 - don't set the inode i_size for regular files from the MDS (896)
6411 - handle out of order completion AST (842)
6412 - don't LBUG if a lock request times out after receiving AST (913)
6413 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6414 - if a bad lock AST arrives, send an error instead of dropping entirely
6415 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6416 - fix leak in bulk IO when only partially completed (899, 900, 926)
6417 - fix O_DIRECT for ia64 (55)
6418 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6419 - C-z after timeout could hang a process forever; fixed (977)
6421 - client-side I/O cache (678, 924, 929, 941, 970)
6423 - READPAGE and SETATTRs which don't take server-side locks get
6426 2003-02-11 Phil Schwan <phil@clusterfs.com>
6429 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6430 - Fix fsfilt_extN_readpage() to read a full page of directory
6431 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6432 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6433 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6434 - Most intent operations were removed, in favour of a new RPC mode
6435 that does a single RPC to the server and bypasses most of the VFS
6436 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6437 - Aggressively cancel local locks on DLM servers
6438 - mds_reint_unlink sends EA to the client if it's the last nlink.
6439 client uses that EA to unlink OST objects.
6440 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6441 - recursive symlinks were fixed (439)
6442 - fixed NULL deref in DEBUG_REQ
6443 - filter_update_lastobjid no longer calls sync, which annoyed extN
6444 - fixed multi-client small-writes to a single file problem (445)
6445 - fixed mtime updates during file writes (607)
6446 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6447 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6448 - fixed Linux OST size reporting problem (444, 656)
6449 - OST now updates object mtime with writes or setattr (607, 619)
6450 - client verifies file size before zeroing page past EOF (445)
6451 - OST now writes last allocated objid to disk with allocation (108)
6452 - LOV on echo now works (409)
6454 - mds_reint_unlink sends a new buffer, with the EA included. this
6455 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6456 set if a regular file was being unlinked, and it was the last link
6457 - use PtlGet from the target for bulk writes (315)
6458 - OST now updates object mtime with writes or setattr (607, 619)
6459 - LDLM now has a grant-time callback to revalidate locked items, if
6461 - Many MDS operations were reorganized to combat race conditions
6463 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6464 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6466 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6469 - Fully reactivate OST imports after reconnection (512, others)
6470 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6471 - More graceful error handling for truncating on dead OST (515)
6472 - Don't error out unless we're actually accessing dead stripes (474)
6473 - Fix garbage sizes when stripes are missing (410)
6474 - LRU counters were broken, causing constant lock purge (433, 432)
6475 - garbage on read from stripes with failed OSTs (441)
6476 - mark OSCs as active before reconnecting during recovery (438)
6477 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6478 - lfind did not preserve OST order in output (443)
6479 - symlinks cause hung clients, incorrect data (439)
6480 - stop dereferencing request after dropping refcount (457)
6481 - don't LASSERT(spin_is_locked) on non-SMP (455)
6482 - fixes for many rename() bugs
6483 - fstat didn't correctly synchronize attributes (399)
6484 - server must handle lock cancellation during blocking AST prep (487)
6485 - bulk descriptors were free()d too soon (511)
6486 - fix paths in lconf, which would load incorrect modules (451, 507)
6487 - fix confusing lconf 'host not found' error message (386)
6488 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6489 - fix race condition in mdc_blocking_ast() for inode access (526)
6490 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6491 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6492 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6493 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6494 - fix lstripe in several ways (kernel side) (527)
6495 - fix request leak in ldlm_cli_enqueue (262)
6496 - incorrect OSC was marked inactive after OST failure
6497 - call mds_fs_cleanup before unmounting filesystem (524)
6498 - fix races between taking ns_lock and ldlm_lock_change_resource
6499 - fix races updating LOV export open file list
6500 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6501 - fix recovery NULL deref in ldlm_cli_cancel_unused
6502 - fix some DLM races by using new hash table for lock handles (419)
6503 - permit the client to specify desired inodes, at replay
6504 - duplicate requests when we queue them for replay reintegration
6505 - fix last_rcvd offset calculation
6506 - sync after each recovered transaction, so we always make progress
6507 - never, not always, ERESTART requests without transnos
6508 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6510 - skip replay if the MDS says that the client is already connected
6511 - don't check for a recovery-enabled export to match lctl's UUID
6512 - don't INC_USE_COUNT for phantom exports
6513 - don't crash when cleaning up phantom exports (567)
6514 - don't double-finish or set replay data for errored mdc_open requests
6515 - abort requests when they time out, so we don't get old replies
6516 - send/receive replies for AST messages again
6517 - if the client says that it doesn't have the lock, cancel it on the
6519 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6520 mark it as destroyed, it will all work out when decref is called
6521 - fix module use counts (22, 581)
6523 - ASTs now expect a reply (server cancels lock on error reply)
6525 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6528 - fix many simultaneous client startup (392)
6529 - fix dentry->d_it clobbering
6530 - credentials weren't being shipped for readdir/getattr operations
6531 - remove invalid assertions triggered during some concurrent MD
6533 - proper Lustre versions added (336, 389)
6534 - fix memory leak for create error case (398)
6535 - fix LOV locking bug that would get cli/srv out of sync
6536 - fix echo client over LOV (409)
6537 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6538 - fix double-O_EXCL intent crash (424)
6539 - avoid sending multiple lock CANCELs (352)
6541 - MDS can do multi-client recovery (modulo bugs in new code)
6543 - many updates, edits, cleanups
6545 2002-11-18 Phil Schwan <phil@clusterfs.com>
6548 - fix null d_it dereference (346)
6549 - fix full OST/dbench hang (333)
6550 - fix permission problem with file removal (286)
6551 - fix removal of OSCs from LOV when they fail
6552 - fix NULL deref during bulk timeout (214)
6553 - fix problems related to multiple filesystems on one MDS (241)
6554 - fixed serious subtle metadata locking bugs
6555 - free locks on clients when inodes are removed due to memory
6557 - fix inode pointer in lock data (285)
6558 - partial support for multiple MDS on a single host (241)
6559 - data locks weren't cancelled at clear_inode time (290, 311)
6560 - intent locks could lead to unbounded lock growth (205)
6561 - added a maximum lock count, an LRU list, and a flusher
6562 - fix multiple rename (365)
6563 - properly abstracted the echo client
6564 - OSC locked 1 byte too many; fixed
6565 - rewrote brw callback code:
6566 - fixed recovery bugs related to LOVs (306)
6567 - fixed too-many-pages-in-one-write crash (191)
6568 - fixed (again) crash in sync_io_timeout (214)
6569 - probably fixed callback-related race (385)
6571 - Add capability to MDS protocol
6572 - LDLM cancellations and callbacks on different portals
6574 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6577 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6578 - defer open object destruction to close time (601981, 138)
6579 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6580 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6581 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6582 - don't call obd_setattr in ll_file_release for destroyed objects
6583 * protocol change to lustre_msg: move |version| and add |flags|
6584 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6585 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6586 * added replay of create, unlink, link and rename operations during
6587 MDS failover; recovery should be much more robust now
6588 * remove failed OSCs from LOVs (only lov_create uses this so far)
6589 * the lustre-HOWTO was brought (more) up to date (582544)
6591 2002-10-23 Phil Schwan <phil@clusterfs.com>
6594 - in-use dentries weren't being reused properly (617851)
6595 - prevent multiple LDLM setup (599178)
6596 - fix LOV size calculations for truncate (617853)
6597 - fix client handling of MDS intent errors (POSIX)
6598 - fix permission bug in lovstripe.c test (624321)
6599 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6600 - truncate past end of file could corrupt data
6601 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6602 - a race in recovery could return ETIMEDOUT to apps (623947)
6603 - building outside the source directory was fixed
6604 * the lustre-HOWTO was brought (more) up to date (582544)
6605 * major progress was made on recovery functionality
6607 2002-10-10 Phil Schwan <phil@clusterfs.com>
6610 - recovery deadlock fix
6611 - rm -rf causes LBUG fix (617817)
6612 - file open by multiple tasks fix (618962)
6613 - directory permissions bugs (602707 and 620007)
6614 - journal_stop fixed with locking (611313)
6615 - O_APPEND failures resolved (618273, perhaps 614459)
6616 - lconf PATH fix (619770)
6617 - IA64 build fix (621450)
6618 - RPC buffer sizes scale with amount of memory
6620 2002-10-01 Phil Schwan <phil@clusterfs.com>
6623 - locks would be cancelled without throwing away data pages,
6624 resulting in inconsistent data (605627)
6625 - inode attributes were not always being refreshed (605627, 612449)
6626 - lconf now continues to cleanup after lctl reports an error
6627 - MDS now enforces user permissions (602707)
6628 - lprocfs cleanup fixed, but not yet enabled (614157)
6629 - fixed infinite server hang, should a client not respond to an AST
6630 - avoid going into recovery if user calls readlink() with a buffer
6631 that's too small (613941)
6632 - AST RPCs no longer require replies (614867) -- this may be changed
6633 - don't crash server if client sends an IOV that's too big (611336)
6634 - fixed lock conversion deadlock (611892)
6635 - fixed the following of symlinks (614622)
6636 * recovery: the server can remove locks from a client that dies, other
6637 clients can make progress
6638 * more extN patch fixes
6639 * compile-time configurable ptlrpc buffer allocations
6641 - collaborative read cache document
6642 - Lustre Lite Performance CDR document-in-progress
6644 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6647 - fix typo in patch-2.4.18
6649 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6652 - clear ptlrpc request each time in handle_incoming_request()
6653 - unlink of files now destroys the object on the OST
6655 2002-09-19 Peter Braam <braam@clusterfs.com>
6657 * add hard link support
6658 * change obdfile creation method
6659 * kernel patch changed
6661 2002-09-19 Peter Braam <braam@clusterfs.com>
6664 - stack overflow bug in extN fixed
6666 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6668 * documentation updates
6669 - add man pages for config tools
6670 - update tests/README to describe testing with new config tools
6671 - finish metadata API descriptions
6672 * bug fixes and cleanups
6673 - statfs workaround for 16TB limit
6674 - LOV stripe allocation improved, can stripe on subset of OSTs
6675 - LOV file size/IO offset was wrong for files > 4GB in size
6676 - object EA data was being dropped, caused files to be unreadable
6677 - memory overflow with non-LOV OST caused memory corruption
6678 - fixed regression tests to work with new config tools, obdfilter
6679 - fixed bug when directory size became larger than 1 block
6680 - fixed bug (for single client case) when PWD was deleted
6681 - invalidate local directory pages when doing intent-based ops
6682 - avoid LDLM oops when lock callback contained bad data
6684 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6686 * documentation updates
6687 * bug fixes and cleanups
6688 - configuration tools
6691 - 64-bit compile warnings
6692 - 64-bit internal statfs data
6694 * test_brw on persistent OST devices
6696 * lprocfs (disabled)
6698 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6700 * documentation updates
6701 * bug fixes and cleanups
6702 * configuration tools
6704 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6707 * many small fixes to 0_5_4
6708 * io/network handling
6709 * thinkos in MDS operations
6711 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6714 * crucial basic fixes to 0.5.3
6715 * IOR, Iozone work over Elan
6718 2002-08-07 Phil Schwan <phil@clusterfs.com>
6719 * version 0_5_3, our first alpha
6720 * we use the new Portals iovs
6721 * documentation updates
6722 * bug fixes and cleanups
6723 * small changes in the DLM wire protocol
6725 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6726 * version 0_5_1 with some initial stability,
6727 * locking on MD and file I/O.
6728 * documentation updates
6729 * several bug fixes since 0.5.0
6730 * small changes in wire protocol
6732 2002-07-18 Phil Schwan <phil@clusterfs.com>
6734 * delivered as Lustre Light Alpha
6735 * fixed a crash after handling invalid MDS requests
6736 * fixed directory pages for architectures with non-4k pages sizes
6738 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6739 * release version v0_4_4
6740 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6742 * Fixes a number of interruption problems with OST operations.
6743 * Update documentation for portals header changes
6744 * Move all wire protocol structs/defines to lustre_idl.h
6745 * Fixes symlink length bug.
6746 * Add tcpdump to repository.
6748 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6749 * release version v0_4_3
6750 * Fixes statfs for inodes on extN.
6751 * Fixes bug in runtests which would delete /etc/hosts.
6752 * Use 64-bit object IDs wherever possible (not into VFS though)
6753 Remove ost_get_info, which is unused by lustre, and out of date.
6755 2002-07-03 Peter Braam <braam@clusterfs.com>
6756 * release version v0_4_2 Fixes a lookup error (type not passed)
6757 * move forward to head of Portals
6758 * move forward to latest Lustre kernel
6760 2002-06-25 Peter Braam <braam@clusterfs.com>
6761 * release version v0_4_1. Hopefully stable on single node use.