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 3.10.0-1062.9.1.el7 (RHEL7.7)
7 * Other server kernels known to build and work at some point (others may also work):
8 3.10.0-862.14.4.el7 (RHEL7.5)
9 3.10.0-957.27.2.el7 (RHEL7.6)
10 4.18.0-80.11.2.el8 (RHEL8.0)
11 4.18.0-147.3.1.el8 (RHEL8.1)
12 4.4.120-92.70 (SLES12 SP2)
13 4.4.180-94.100 (SLES12 SP3)
14 4.4.0-131 (Ubuntu 16.04)
15 4.15.0-32 (Ubuntu 18.04)
16 vanilla linux 4.6.7 (ZFS only)
17 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
18 * Client primary unpatched kernels built and tested during release cycle:
19 3.10.0-1062.9.1.el7 (RHEL7.7)
20 4.12.14-95.45 (SLES12 SP4)
21 4.12.14-197.29 (SLES15 SP1)
22 4.15.0-48 (Ubuntu 18.04)
23 * Other clients known to build on unpatched kernels at some point (others may also work):
24 3.10.0-862.14.4.el7 (RHEL7.5)
25 3.10.0-957.27.2.el7 (RHEL7.6)
26 4.14.0-49.13.1.el7a (RHEL7.5)
27 4.14.0-115.2.2.el7a (RHEL7.6)
28 4.18.0-80.11.2.el8 (RHEL8.0)
29 4.18.0-147.3.1.el8 (RHEL8.1)
30 4.4.120-92.70 (SLES12 SP2)
31 4.4.180-94.100 (SLES12 SP3)
32 4.4.0-131 (Ubuntu 16.04)
33 vanilla linux 4.15.0 (ZFS only)
34 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
35 * Recommended ZFS version: 0.8.3
36 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
37 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
38 stack overflow. For more information, please refer to bugzilla 17630.
39 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
40 the re-exporting nodes to be mounted with "32bitapi" mount option
41 * Note for Ubuntu 14.04.5:
42 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
43 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
45 --------------------------------------------------------------------------------
49 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
50 for currently supported client and server kernel versions.
51 * Server primary kernels built and tested during release cycle:
52 3.10.0-1062.1.1.el7 (RHEL7.7)
53 * Other server kernels known to build and work at some point (others may also work):
54 3.10.0-862.14.4.el7 (RHEL7.5)
55 3.10.0-957.27.2.el7 (RHEL7.6)
56 4.18.0-80.7.1.el8 (RHEL8.0)
57 4.4.120-92.70 (SLES12 SP2)
58 4.4.180-94.100 (SLES12 SP3)
59 4.4.0-131 (Ubuntu 16.04)
60 4.15.0-32 (Ubuntu 18.04)
61 vanilla linux 4.6.7 (ZFS only)
62 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
63 * Client primary unpatched kernels built and tested during release cycle:
64 3.10.0-1062.1.1.el7 (RHEL7.7)
65 4.12.14-95.29 (SLES12 SP4)
66 4.15.0-48 (Ubuntu 18.04)
67 * Other clients known to build on unpatched kernels at some point (others may also work):
68 3.10.0-862.14.4.el7 (RHEL7.5)
69 3.10.0-957.27.2.el7 (RHEL7.6)
70 4.14.0-49.13.1.el7a (RHEL7.5)
71 4.14.0-115.2.2.el7a (RHEL7.6)
72 4.18.0-80.7.1.el8 (RHEL8.0)
73 4.4.120-92.70 (SLES12 SP2)
74 4.4.180-94.100 (SLES12 SP3)
75 4.4.0-131 (Ubuntu 16.04)
76 vanilla linux 4.15.0 (ZFS only)
77 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
78 * Recommended ZFS / SPL version: 0.7.13
79 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
80 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
81 stack overflow. For more information, please refer to bugzilla 17630.
82 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
83 the re-exporting nodes to be mounted with "32bitapi" mount option
84 * Note for Ubuntu 14.04.5:
85 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
86 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
88 --------------------------------------------------------------------------------
92 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
93 for currently supported client and server kernel versions.
94 * Server known to build on the following kernels (others may also work):
95 2.6.32-696.18.7.el6 (RHEL6.9)
96 3.10.0-862.14.4.el7 (RHEL7.5)
97 3.10.0-957.el7 (RHEL7.6)
98 4.4.120-92.70 (SLES12 SP2)
99 4.4.162-94.69 (SLES12 SP3)
100 4.4.0-131 (Ubuntu 16.04)
101 4.15.0-32 (Ubuntu 18.04)
102 vanilla linux 4.6.7 (ZFS only)
103 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
104 * Client known to build on unpatched kernels (others may also work):
105 2.6.32-696.18.7.el6 (RHEL6.9)
106 2.6.32-754.6.3.el6 (RHEL6.10)
107 3.10.0-862.14.4.el7 (RHEL7.5)
108 3.10.0-957.el7 (RHEL7.6)
109 4.14.0-49.13.1.el7a (RHEL7.5)
110 4.14.0-115.2.2.el7a (RHEL7.6)
111 4.4.120-92.70 (SLES12 SP2)
112 4.4.162-94.69 (SLES12 SP3)
113 4.4.0-131 (Ubuntu 16.04)
114 4.15.0-32 (Ubuntu 18.04)
115 vanilla linux 4.15.0 (ZFS only)
116 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
117 * Recommended ZFS / SPL version: 0.7.9
118 * Tested with ZFS / SPL version: 0.7.9
119 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
120 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
121 stack overflow. For more information, please refer to bugzilla 17630.
122 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
123 the re-exporting nodes to be mounted with "32bitapi" mount option
124 * Note for Ubuntu 14.04.5:
125 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
126 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
128 --------------------------------------------------------------------------------
130 2018-04-03 Intel Corporation
132 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
133 for currently supported client and server kernel versions.
134 * Server known to build on patched kernels:
135 2.6.32-431.29.2.el6 (RHEL6.5)
136 2.6.32-504.30.3.el6 (RHEL6.6)
137 2.6.32-573.26.1.el6 (RHEL6.7)
138 2.6.32-642.15.1.el6 (RHEL6.8)
139 2.6.32-696.18.7.el6 (RHEL6.9)
140 3.10.0-693.21.1.el7 (RHEL7.4)
141 3.0.101-0.47.71 (SLES11 SP3)
142 3.0.101-107 (SLES11 SP4)
143 3.12.74-60.64.40 (SLES12 SP1)
144 4.4.114-92.64 (SLES12 SP2)
145 4.4.114-94.11 (SLES12 SP3)
146 3.13.0-101 (Ubuntu 14.04, ZFS only)
147 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
148 4.4.0-116 (Ubuntu 16.04)
149 vanilla linux 4.6.7 (ZFS only)
150 * Client known to build on unpatched kernels:
151 2.6.32-431.29.2.el6 (RHEL6.5)
152 2.6.32-504.30.3.el6 (RHEL6.6)
153 2.6.32-573.26.1.el6 (RHEL6.7)
154 2.6.32-642.15.1.el6 (RHEL6.8)
155 2.6.32-696.18.7.el6 (RHEL6.9)
156 3.10.0-693.21.1.el7 (RHEL7.4)
157 3.0.101-0.47.71 (SLES11 SP3)
158 3.0.101-107 (SLES11 SP4)
159 3.12.74-60.64.40 (SLES12 SP1)
160 4.4.114-92.64 (SLES12 SP2)
161 4.4.114-94.11 (SLES12 SP3)
162 3.13.0-101 (Ubuntu 14.04)
163 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
164 4.4.0-116 (Ubuntu 16.04)
166 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
167 * Recommended ZFS / SPL version: 0.7.6
168 * Tested with ZFS / SPL version: 0.7.6
169 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
170 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
171 stack overflow. For more information, please refer to bugzilla 17630.
172 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
173 the re-exporting nodes to be mounted with "32bitapi" mount option
174 * Note for Ubuntu 14.04.5:
175 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
176 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
178 --------------------------------------------------------------------------------
180 2017-07-05 Intel Corporation
182 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
183 for currently supported client and server kernel versions.
184 * Server known to build on patched kernels:
185 2.6.32-431.29.2.el6 (RHEL6.5)
186 2.6.32-504.30.3.el6 (RHEL6.6)
187 2.6.32-573.26.1.el6 (RHEL6.7)
188 2.6.32-642.15.1.el6 (RHEL6.8)
189 2.6.32-696.3.1.el6 (RHEL6.9)
190 3.10.0-514.21.1.el7 (RHEL7.3)
191 3.0.101-0.47.71 (SLES11 SP3)
192 3.0.101-100 (SLES11 SP4)
193 3.12.74-60.64.40 (SLES12 SP1)
194 4.4.59-92.17 (SLES12 SP2)
195 vanilla linux 4.6.7 (ZFS only)
196 * Client known to build on unpatched kernels:
197 2.6.32-431.29.2.el6 (RHEL6.5)
198 2.6.32-504.30.3.el6 (RHEL6.6)
199 2.6.32-573.26.1.el6 (RHEL6.7)
200 2.6.32-642.15.1.el6 (RHEL6.8)
201 2.6.32-696.3.1.el6 (RHEL6.9)
202 3.10.0-514.21.1.el7 (RHEL7.3)
203 3.0.101-0.47.71 (SLES11 SP3)
204 3.0.101-100 (SLES11 SP4)
205 3.12.74-60.64.40 (SLES12 SP1)
206 4.4.59-92.17 (SLES12 SP2)
208 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
209 * Recommended ZFS / SPL version: 0.6.5.9
210 * Tested with ZFS / SPL version: 0.6.5.9
211 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
212 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
213 stack overflow. For more information, please refer to bugzilla 17630.
214 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
215 the re-exporting nodes to be mounted with "32bitapi" mount option
217 --------------------------------------------------------------------------------
219 12-07-2016 Intel Corporation
221 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
222 for currently supported client and server kernel versions.
223 * Server known to build on patched kernels:
224 2.6.32-431.29.2.el6 (RHEL6.5)
225 2.6.32-504.30.3.el6 (RHEL6.6)
226 2.6.32-573.26.1.el6 (RHEL6.7)
227 2.6.32-642.11.1.el6 (RHEL6.8)
228 3.10.0-514.el7 (RHEL7.3)
229 3.0.101-0.47.71 (SLES11 SP3)
230 3.0.101-80 (SLES11 SP4)
231 3.12.62-60.64.8 (SLES12 SP1)
232 vanilla linux 4.5.7 (ZFS only)
233 * Client known to build on unpatched kernels:
234 2.6.32-431.29.2.el6 (RHEL6.5)
235 2.6.32-504.30.3.el6 (RHEL6.6)
236 2.6.32-573.26.1.el6 (RHEL6.7)
237 2.6.32-642.11.1.el6 (RHEL6.8)
238 3.10.0-514.el7 (RHEL7.3)
239 3.0.101-0.47.71 (SLES11 SP3)
240 3.0.101-80 (SLES11 SP4)
241 3.12.62-60.64.8 (SLES12 SP1)
243 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
244 * Recommended ZFS / SPL version: 0.6.5.8
245 * Tested with ZFS / SPL version: 0.6.5.8
246 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
247 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
248 stack overflow. For more information, please refer to bugzilla 17630.
249 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
250 the re-exporting nodes to be mounted with "32bitapi" mount option
252 --------------------------------------------------------------------------------
254 02-29-2016 Intel Corporation
256 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
257 for currently supported client and server kernel versions.
258 * Server known to build on patched kernels:
259 2.6.32-431.29.2.el6 (RHEL6.5)
260 2.6.32-504.30.3.el6 (RHEL6.6)
261 2.6.32-573.12.1.el6 (RHEL6.7)
262 3.10.0-327.3.1.el7 (RHEL7.2)
263 3.0.101-0.47.71 (SLES11 SP3)
264 3.0.101-68 (SLES11 SP4)
265 vanilla linux 4.2.1 (ZFS only)
266 * Client known to build on unpatched kernels:
267 2.6.32-431.29.2.el6 (RHEL6.5)
268 2.6.32-504.30.3.el6 (RHEL6.6)
269 2.6.32-573.12.1.el6 (RHEL6.7)
270 3.10.0-327.3.1.el7 (RHEL7.2)
271 3.0.101-0.47.71 (SLES11 SP3)
272 3.0.101-68 (SLES11 SP4)
275 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
276 * Recommended ZFS / SPL version: 0.6.4.2
277 * Tested with ZFS / SPL version: 0.6.4.2
278 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
279 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
280 stack overflow. For more information, please refer to bugzilla 17630.
281 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
282 the re-exporting nodes to be mounted with "32bitapi" mount option
284 --------------------------------------------------------------------------------
286 03-10-2015 Intel Corporation
288 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
289 for currently supported client and server kernel versions.
290 * Server known to build on patched kernels:
291 2.6.32-431.29.2.el6 (RHEL6.5)
292 2.6.32-504.8.1.el6 (RHEL6.6)
293 3.0.101-0.46 (SLES11 SP3)
294 * Client known to build on unpatched kernels:
295 2.6.32-431.29.2.el6 (RHEL6.5)
296 2.6.32-504.8.1.el6 (RHEL6.6)
297 3.10.0-123.20.1.el7 (RHEL7)
298 3.0.101-0.46 (SLES11 SP3)
299 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
300 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
301 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
302 stack overflow. For more information, please refer to bugzilla 17630.
303 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
304 the re-exporting nodes to be mounted with "32bitapi" mount option
306 Severity : enhancement
308 Description: control OST-index in IDIF via ROCOMPAT flag.
309 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
310 last_rcvd file. For new formatted OST device, it will be auto set;
311 for the case of upgrading from old OST device, you can enable it
312 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
313 enabled, for new created OST-object, its IDIF-in-LMA will contain
314 the OST-index; for the existing OST-object, the OSD will convert
315 old format IDIF as new format IDIF with OST-index stored in the
316 LMA EA when accessing such OST-object or via OI scrub. Once such
317 flag is enabled, it cannot be reverted back, so the system cannot
318 be downgraded to the orignal incompatible version.
320 --------------------------------------------------------------------------------
322 07-30-2014 Intel Corporation
324 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
325 for currently supported client and server kernel versions.
326 * Server known to build on patched kernels:
327 2.6.32-431.20.3.el6 (RHEL6.5)
328 3.0.101-0.31 (SLES11 SP3)
329 * Client known to build on unpatched kernels:
330 2.6.32-431.20.3.el6 (RHEL6.5)
331 3.0.101-0.31 (SLES11 SP3)
334 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
335 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
336 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
337 stack overflow. For more information, please refer to bugzilla 17630.
338 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
339 the re-exporting nodes to be mounted with "32bitapi" mount option
341 --------------------------------------------------------------------------------
343 10-31-2013 Intel Corporation
345 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
346 * Server builds for kernels:
347 2.6.32-358.18.1.el6 (RHEL6.4)
348 * Client builds for unpatched kernels:
349 2.6.32-358.18.1.el6 (RHEL6.4)
350 3.0.92-0.8 (SLES11 SP3)
352 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
353 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
354 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
355 stack overflow. For more information, please refer to bugzilla 17630.
356 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
357 the re-exporting nodes to be mounted with "32bitapi" mount option
359 --------------------------------------------------------------------------------
361 05-30-2013 Intel Corporation
363 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
364 * Server builds for kernels:
365 2.6.32-358.6.2.el6 (RHEL6.4)
366 * Client builds for unpatched kernels:
367 2.6.32-358.6.2.el6 (RHEL6.4)
368 3.0.74-0.6.6 (SLES11 SP2)
370 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
371 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
372 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
373 stack overflow. For more information, please refer to bugzilla 17630.
374 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
375 the re-exporting nodes to be mounted with "32bitapi" mount option
377 --------------------------------------------------------------------------------
379 9-30-2012 Whamcloud, Inc.
381 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
382 * Server support for kernels:
383 2.6.32-279.5.1.el6 (RHEL6)
384 * Client support for unpatched kernels:
385 2.6.32-279.5.1.el6 (RHEL6)
387 2.6.18-238.19.1.el5 (RHEL5)
388 2.6.16.60-0.69.1 (SLES10)
389 * Recommended e2fsprogs version: 1.41.90.wc4
390 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
391 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
392 stack overflow. For more information, please refer to bugzilla 17630.
393 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
394 the re-exporting nodes to be mounted with "32bitapi" mount option
396 Severity : enhancement
398 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
400 Severity : enhancement
402 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
404 Severity : enhancement
406 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
408 Severity : enhancement
410 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
412 Severity : enhancement
414 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
416 --------------------------------------------------------------------------------
418 04-30-2012 Whamcloud, Inc.
420 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
421 * Server support for kernels:
422 2.6.32-220.4.2.el6 (RHEL6)
423 * Client support for unpatched kernels:
424 2.6.32-220.4.2.el6 (RHEL6)
426 2.6.18-238.19.1.el5 (RHEL5)
427 2.6.16.60-0.69.1 (SLES10)
428 * Recommended e2fsprogs version: 1.41.90.wc4
429 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
430 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
431 stack overflow. For more information, please refer to bugzilla 17630.
432 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
433 the re-exporting nodes to be mounted with "32bitapi" mount option
435 Severity : enhancement
437 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
439 Severity : enhancement
441 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
443 Severity : enhancement
445 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
447 Severity : enhancement
449 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
451 --------------------------------------------------------------------------------
453 09-30-2011 Whamcloud, Inc.
455 * Server support for kernels:
456 2.6.18-238.19.1.el5 (RHEL5)
457 2.6.32-131.6.1.el6 (RHEL6)
458 * Client support for unpatched kernels:
459 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
460 2.6.18-238.19.1.el5 (RHEL5)
461 2.6.32-131.6.1.el6 (RHEL6)
462 2.6.29.4-167.fc11 (FC11)
463 2.6.16.60-0.69.1 (SLES10)
465 * Recommended e2fsprogs version: 1.41.90.wc3
466 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
467 Lustre filesystem with 4K stack may cause a stack overflow. For more
468 information, please refer to bugzilla 17630.
469 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
470 mounted with "32bitapi" mount option
472 Severity : enhancement
473 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
475 Severity : enhancement
476 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
478 Severity : enhancement
479 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
482 Frequency : only with SLES10
484 Description: Use OFED "KMP" provided by Novell
485 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
486 Lustre is now built against this package. That means you need to
487 install the ofed-kmp package from Novell for the patchless client.
488 Note that the ofed-kmp that Novell ships may not exactly match the
489 kernel version but should still be compatible.
491 Severity : enhancement
493 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
494 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
496 --------------------------------------------------------------------------------
498 2010-07-15 Oracle, Inc.
500 * Server support for kernels:
501 2.6.18-164.11.1.el5 (RHEL 5)
502 2.6.18-164.11.1.0.1.el5 (OEL 5)
503 * Client support for unpatched kernels:
504 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
505 2.6.18-164.11.1.el5 (RHEL 5),
506 2.6.18-164.11.1.0.1.el5 (OEL 5)
507 2.6.16.60-0.42.8 (SLES 10),
509 2.6.29.4-167.fc11 (FC11)
510 * Recommended e2fsprogs version: 1.41.10-sun2
512 Severity : enhancement
514 Description: Distribute lfs_migrate script with lustre
515 Details : The lfs_migrate mentioned in the manual is now included in the
516 lustre tree in lustre/scripts and will distributed in lustre RPMs.
518 Severity : enhancement
520 Description: Failover nid list added to import proc file.
522 Severity : enhancement
524 Description: "som_preview" mount option required to use SOM
525 Details : The clients will need to use som_preview mount option
529 Description: Lustre does not do 1MB IOs to HW RAID
530 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
532 Severity : enhancement
534 Description: lfs quota output is not very convenient for awk/sed-parsing
535 Details : Some positions in lfs quota output table could be empty or
536 non-empty which made it hard to parse it with scripts, now
537 a dash is put instead of space where there is not supposed
542 Description: Update to OFED 1.5.1
543 Details : OFED 1.5.1 will be used in environments that don't use
544 vendor supplied OFED.
548 Description: Remove lnet. prefix from lctl params display
549 Details : Also renamed lnet.memused to lnet_memused to
554 Description: Use conf_param -d to remove permanant settings.
558 Description: Add list_param -R
559 Details : lctl list_param now supports recursive mode.
562 Bugzilla : 21829, 22312
563 Description: Fix llobdstat.
564 Details : 1. Now updating the write_bytes counter
565 2. Port script fixes from 1.8.
569 Description: decrease the usage of memory on clients.
570 Details : 1. On clients, recycle dentries and inodes unused.
571 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
576 Description: llobdstat fix and enhancement
577 Details : add a counter to set a limit to how many samples will be returned
578 fix a wildcard in the path to limit to obdfilter stats only
582 Description: lfs check only allowed for root
583 Details : Code cleanups and added a regresion test of non-root use of
588 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
589 Details : Fix hr->hr_index overload.
593 Description: Fix FIEMAP functionality.
597 Description: FC11 patchless client support added
601 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
602 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
606 Description: New acc-sm script and various test-framework enhancments.
610 Description: Previously landed patch was reverted.
614 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
618 Description: Add ldiskfs maxdirsize mount option.
622 Description: Fix mkfs.lustre support for 16TB LUNs.
626 Description: Print mmp_check_interval
627 Details : print mmp_check_interval and make it possible to abort
628 mount operation in case it takes too long.
632 Description: Add cascading_rw to lustre mpi tests.
636 Description: Store SOM data in EA.
637 Details : Also several SOM related cleanups.
641 Description: Add refcount to md_open_data
645 Description: Use refcount on lov_request_set to prevent memory corruption.
649 Description: Statahead: use parent dir's i_mutex for synchronization.
653 Description: loadgen improvements.
657 Description: Enable peer health detection in ksocklnd and ko2iblnd.
661 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
665 Description: Fix race on obd_devcie:md_stats
669 Description: Fix SLES 11 client builds.
670 Details : Resolved uninitalized pointer gcc warnings.
674 Description: Allow non-root access for "lfs check".
675 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
679 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
681 Severity : enhancement
683 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
684 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
688 Description: Sanity test_27p fails.
689 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
690 Sanity tests 27 were cleaned up and improved.
694 Description: Unify req format on client/servers
695 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
696 buffers and swabbers approach..
701 Description: drop join file stuff.
703 Severity : enhancement
705 Description: Add 2.6.27 and SLES11 (patchless client) support
709 Description: Fix in ptlrpc_expire_one_request() to print the signed time
712 Severity : enhancement
714 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
717 Severity : enhancement
719 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
721 Severity : enhancement
723 Description: correctly handle big reply message.
724 Details : send LNet event if reply is bigger than buffer and adjust this buffer
729 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
731 Severity : enhancement
733 Description: Add support for OEL5.
737 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
739 Severity : enhancement
741 Description: "weak-modules" support
742 Details : Implement "weak-modules" support which enables kernel modules
743 to be used with any kernel that implements the same kABI. In
744 order to achieve this modules are now installed in
745 /lib/modules/$(uname -r)/updates/kernel on all distributions.
749 Description: client could not reconnect to OST because of an active request.
750 Details : The client now retries to reconnect to the same server, if a
751 connect request failed with EBUSY or -EAGAIN. Test case 35b
752 conf-sanity.sh has been created.
756 Description: Don't allow make backward step on assignin osc next id.
757 Details : race between allocation next id and ll_sync thread can be cause
758 of set wrong osc next id and can be kill valid ost objects.
762 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
766 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
770 Description: Error handling in osc_statfs_interpret() has been improved.
771 Details : Check in osc_statfs_interpret() for EBADR.
775 Description: Do not send statfs() requests to OSTs disabled by administrator.
776 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
778 Severity : enhancement
780 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
782 Severity : enhancement
784 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
786 Severity : enhancement
788 Description: Enhancement for lfs(1) command to use numeric uid/gid.
792 Description: Improved cleanup procedure for conf-sanity test_46a.
796 Description: Do not start bulk transfer, if client tried to reconnect.
797 Details : Do not start bulk transfer, if client tried to reconnect,
798 but failed due to active rpcs. Check for exp_abort_active_req
799 has been implemented.
803 Description: Prevent inconsistences between linux and lustre mount structures.
804 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
805 Make the sleep interruptible.
807 Severity : enhancement
809 Description: provide server to client comms path
810 Details : server to client communications path via new LDLM_SET_INFO rpc
814 Description: truncate starts GFP_FS allocation under transaction causing deadlock
815 Details : ldiskfs_truncate calls grab_cache_page which may start page
816 allocation under an open transaction. This may lead to
817 calling prune_icache with consequent lustre reentrance.
821 Description: send correctly lsm on open replay
822 Details : MDS is trust to LSM size on replay open, but client can set wrong size
827 Description: enable client interface failover
828 Details : When a child reconnects from another NID, properly update export
829 nid hash position and ldlm reverse import.
831 Severity : enhancement
833 Description: Communicate OST degraded/readonly state via statfs to MDS
834 Details : Flags in the statfs returned from OSTs indicate whether the
835 OST is in a degraded RAID state, or if the filesystem has
836 turned read-only after a filesystem error is detected.
841 Description: don't panic if EPROTO was hit when reading symlink
842 Details : correctly handling request reference in error cases.
844 Severity : enhancement
846 Description: Add LustreNetLink, a kernel-userspace communcation path.
848 Severity : enhancement
850 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
855 Description: access to llog context before init.
856 Details : move handling CATALOGS file at osc layer and forbid access to llog
860 Frequency : rare, only when client is writing to OST on local node
862 Description: Avoid deadlock for local client writes
863 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
864 memory freeing context. This allows OST threads to set the
865 PF_MEMALLOC flag on task structures in order to allocate memory
866 from reserved pools and complete IO.
867 Use GFP_HIGHUSER for OST allocations for non-local client writes,
868 so that the OST threads generate memory pressure and allow
869 inactive pages to be reclaimed.
871 Severity : enhancement
873 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
878 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
879 Details : move ldlm namespace creation in setup phase to avoid grab
880 _lprocfs_lock with cli_sem held.
884 Description: Temporarily disable grant shrink.
885 Details : Disable the feature for debugging.
888 Frequency : only when formatting test filesystems
890 Description: Unable to run several mkfs.lustre on loop devices at the same time
891 Details : mkfs.lustre returns error 256 on the concurrent loop devices
892 formatting. The solution is to proper handle the error.
894 Severity : enhancement
896 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
898 Severity : enhancement
900 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
905 Description: resolve race between obd_disconnect and class_disconnect_exports
906 Details : if obd_disconnect will be called to already disconnected export he
907 forget release one reference and osc module can't unloaded.
909 Severity : enhancement
911 Description: Allow tuning service thread via /proc
912 Details : For each service a new
913 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
914 created that can be used to set min/max thread counts, and get the
915 current number of running threads.
920 Descriptoin: don't return error if have partially created objects for file.
921 Details : lov_update_create_set uses set->set_success as index for created
922 objects, so if some requests will be failed, they will have hole
923 at end of array and qos_shrink_lsm for allocate correct lsm.
925 Severity : enhancement
927 Description: Update OFED support to 1.4.1
932 Description: Reduce small size read RPC
933 Details : Set read-ahead limit for every file and only do read-ahead when
934 available read-ahead pages are bigger than 1M to avoid small size
937 Severity : enhancement
939 Description: move AT tunable parameters for more consistent usage
940 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
942 Severity : enhancement
944 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
945 Details : allow skip disconnected ost for send statfs request and hide error
951 Description: fix racy locking of mballoc block bitmaps causing BUG
952 Details : The locking of the mballoc buddy bitmap and the in-memory
953 block bitmap was using two different spin locks in some
954 cases. This made it possible to incorrectly access the
955 mballoc bitmap while another process was modifying it,
956 causing a sanity assertion to fail. While no on-disk corruption
957 was reported, there was some risk of this happening.
960 Frequency : rare, on llog test 6
962 Descriptoin: don't allow connect to already connected import
963 Details : allowing connect to already connected import is hide connecting problem.
968 Descriptoin: allow kill process which wait statahead result
969 Details : for some reasons 'ls' can stick in waiting result from statahead,
970 in this case need way for kill this process.
972 Severity : enhancement
974 Description: Add state history info file, enhance import info file
975 Details : Track import connection state changes in a new osc/mdc proc file;
976 add overview-type data to the osc/mdc import proc file.
978 Severity : enhancement
980 Description: MDS create should not wait for statfs RPC while holding DLM lock.
982 Severity : enhancement
984 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
989 Descriptoin: OST grant too much space to client even there are not enough space.
990 Details : Client will shrink its grant cache to OST if there are no write
991 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
992 this grant cache if there are already not enough avaible space
993 (left_space < total_clients * 32M).
996 Frequency : start MDS on uncleanly shutdowned MDS device
998 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
999 Details : stay in waiting mds<>ost recovery finished produce random bugs
1000 due race between two ll_sync thread for one lov target. send
1001 ACTIVATE event only if connect realy finished and import have
1005 Frequency : rare, connect and disconnect target at same time
1007 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1008 Details : don't call obd_disconnect under lov_lock. this long time
1009 operation and can block ptlrpcd which answer to connect request.
1014 Descriptoin: don't lose wakeup for imp_recovery_waitq
1015 Details : recover_import_no_retry or invalidate_import and import_close can
1016 both sleep on imp_recovery_waitq, but we was send only one wakeup
1020 Frequency : always with long access acl
1022 Descriptoin: mds can't pack reply with long acl.
1023 Details : mds don't control size of acl but they limited by reint/getattr
1026 Severity : enhancement
1028 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1030 Severity : enhancement
1032 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1035 Frequency : start MDS on uncleanly shutdowned MDS device
1037 Descriptoin: aborting recovery hang on MDS
1038 Details : don't throttle destroy RPCs for the MDT.
1041 Frequency : on remount
1043 Description: external journal device not working after the remount
1044 Details : clear dev_rdonly flag for external journal devices in
1050 Description: shutdown vs evict race
1051 Details : client_disconnect_export vs connect request race.
1052 if client will evicted at this time - we start invalidate
1053 thread without referece to import and import can be freed
1057 Frequency : rare, need acl's on inode.
1059 Description: client can't handle ost additional correctly
1060 Details : if ost was added after client connected to mds client can have
1061 hit lnet_try_match_md ... to big messages to wide striped files.
1062 in this case need teach client to handle config events about add
1063 lov target and update client max ea size at that event.
1065 Severity : enhancement
1067 Description: Changelogs
1068 Details : Changelogs are a lightweight mechanism to track filesystem
1069 metadata and namespace changes. The changelog is recorded
1070 permanently on the MDTs, and is periodically "consumed" / purged
1071 when records are no longer needed.
1073 Severity : enhancement
1075 Description: compact fld format with extents
1076 Details : Store range of seq rather than every seq in FLD. Seq
1077 controller update FLD rather than clients. In Case of CMD, mdt0
1078 has FLD, all other metadata server act as non persistent proxy
1079 for FLD queries and cache fld entries in fld cache.
1084 Description: don't skip ost target if they assigned to file
1085 Details : Drop slow OSCs if we can, but not for requested start idx.
1086 This means "if OSC is slow and it is not the requested
1087 start OST, then it can be skipped, otherwise skip it only
1088 if it is inactive/recovering/out-of-space.
1092 Description: more cleanup in mds_lov
1093 Details : not send LOV EA under replay, we can't know about they size at this
1094 time. Don't allow client connect to mds before any ost connected,
1095 for avoid problems with LOV EA size and returning EIO to client.
1097 Severity : enhancement
1099 Description: Interoperability at server side (Disk interoperability)
1101 Severity : enhancement
1103 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1105 Severity : enhancement
1107 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1109 Severity : enhancement
1111 Description: New client IO stack (CLIO).
1113 Severity : enhancement
1115 Description: Commit on sharing. Eliminate inter-client dependencies between
1116 uncommitted transactions by doing transaction commits.
1117 Thereby clients may recovery independently.
1120 Frequency : Create a symlink file with a very long name
1122 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1123 Details : If there is no extra space in the request for early cancels,
1124 ldlm_req_handles_avail() returns 0 instead of a negative value.
1126 Severity : enhancement
1128 Description: Add /proc entry for import status
1129 Details : The mdc, osc, and mgc import directories now have
1130 an import directory that contains useful import data for debugging
1131 connection problems.
1133 Severity : enhancement
1135 Description: Re-disable certain /proc logging
1136 Details : Enable and disable client's offset_stats, extents_stats and
1137 extents_stats_per_process stats logging on the fly.
1140 Frequency : Only on FC kernels 2.6.22+
1142 Description: oops in statahead
1143 Details : Do not drop reference count for the dentry from VFS when lookup,
1144 VFS will do that by itself.
1146 Severity : enhancement
1148 Description: Generic /proc file permissions
1149 Details : Set /Proc file permissions in a more generic way to enable non-
1150 root users operate on some /proc files.
1154 Description: Hitting mdc_commit_close() ASSERTION
1155 Details : Properly handle request reference release in
1156 ll_release_openhandle().
1160 Description: quota recovery deadlock during mds failover
1161 Details : This patch includes att18982, att18236, att18237 in bz14840.
1163 1. fix osts hang when mds does failover with quotaon
1164 2. prevent watchdog storm when osts threads wait for the
1169 Frequency : only patchless client
1170 Description: add workaround for race between add/remove dentry from hash
1172 Severity : enhancement
1174 Description: Allow OST glimpses to return PW locks
1178 Description: LBUG when llog conf file is full
1179 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1183 Description: Prevent import from entering FULL state when server in recovery
1187 Description: service mount cannot take device name with ":"
1188 Details : Only when device name contains ":/" will mount treat it as
1194 Description: replace ptlrpcd with the statahead thread to interpret the async
1195 statahead RPC callback
1199 Frequency : on recovery
1200 Description: I/O failures after umount during fail back
1201 Details : if client reconnected to restarted server we need join to recovery
1202 instead of find server handler is changed and process self
1203 eviction with cancel all locks.
1205 Severity : enhancement
1207 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1209 Severity : enhancement
1211 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1213 Severity : enhancement
1215 Description: Upcall on Lustre log has been dumped
1216 Details : Allow for a user mode script to be called once a Lustre log has
1217 been dumped. It passes the filename of the dumped log to the
1218 script, the location of the script can be specified via
1219 /proc/sys/lnet/debug_log_upcall.
1224 Description: avoid idr_remove called for id which is not allocated.
1225 Details : Move assigment s_dev for clustered nfs to end of initialization,
1226 to avoid problem with error handling.
1231 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1232 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1233 into CONN_USED_HASH and this prodice warning when put connection
1234 again in unused hash.
1237 Severity : enhancement
1239 Description: File striping can now be set to use an arbitrary pool of OSTs.
1241 Severity : enhancement
1243 Description: Export bytes_read/bytes_write count on OSC/OST.
1247 Description: Early reply size mismatch, MGC loses connection
1248 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1249 the connect flags are properly negotiated.
1254 Description: quotas are not honored with O_DIRECT
1255 Details : all writes with the flag O_DIRECT will use grants which leads to
1256 this problem. Now using OBD_BRW_SYNC to guard this.
1260 Description: add quota statistics
1261 Details : 1. sort out quota proc entries and proc code.
1262 2. add quota statistics
1264 Severity : enhancement
1266 Description: enable quota support for HEAD.
1270 Description: Properly propagate oinfo flags from lov to osc for statfs
1271 Details : restore missing copy oi_flags to lov requests.
1273 Severity : enhancement
1275 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1276 : l_getgroups(8), lst(8), routerstat(8)
1277 Details : included man pages for llobdstat(8), llstat(8),
1278 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1280 Severity : enhancement
1282 Description: configure's --enable-quota should check the
1283 : kernel .config for CONFIG_QUOTA
1284 Details : configure is terminated if --enable-quota is passed but
1285 : no quota support is in kernel
1289 Description: Remove portals compatibility
1290 Details : Remove portals compatibility, not interoperable with releases
1295 Description: Resolve device initialization race
1296 Details : Prevent proc handler from accessing devices added to the
1297 obd_devs array but yet be intialized.
1299 Severity : enhancement
1301 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1303 Severity : enhancement
1305 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1310 Description: Using wrong pointer in osc_brw_prep_request
1311 Details : Access to array[-1] can produce panic if kernel compiled with
1312 CONFIG_PAGE_ALLOC enabled
1316 Description: Client runs out of low memory
1317 Details : Consider only lowmem when counting initial number of llap pages
1321 Description: *optional* service tags registration
1322 Details : if the "service tags" package is installed on a Lustre node
1323 When the filesystem is mounted, a local-node service tag will
1324 be created. See http://inventory.sun.com/ for more information
1325 about the Service Tags asset management system.
1329 Description: Kernel BUG tries to release flock
1330 Details : Lustre does not destroy flock lock before last reference goes
1331 away. So always drop flock locks when client is evicted and
1332 perform unlock regardless of successfulness of speaking to MDS.
1336 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1340 Description: sanity test 65a fails if stripecount of -1 is set
1341 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1346 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1347 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1350 Severity : enhancement
1352 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1356 Description: do not process already freed flock
1357 Details : flock can possibly be freed by another thread before it reaches
1358 to ldlm_flock_completion_ast.
1362 Description: LBUG during stress test
1363 Details : Need properly lock accesses the flock deadlock detection list.
1367 Description: oops in page fault handler
1368 Details : kernel page fault handler can return two special 'pages' in error
1369 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1373 Description: timeout with invalidate import.
1374 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1375 should be handled by ptlrpcd. This produce long age waiting and
1376 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1378 Severity : enhancement
1380 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1384 Description: enable MGS and MDT services start separately
1385 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1386 a co-located MGS without starting the MGS, which is a complement
1387 to 'nosvc' mount option.
1391 Frequency : after recovery
1392 Description: precreate to many object's after del orphan.
1393 Details : del orphan st in oscc last_id == next_id and this triger growing
1394 count of precreated objects. Set flag LOW to skip increase count
1395 of precreated objects.
1399 Frequency : rare, on clear nid stats
1400 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1401 Details : when clean nid stats sometimes try destroy live entry,
1402 and this produce panic in free.
1406 Description: Stack overflow during MDS log replay
1407 ease stack pressure by using a thread dealing llog_process.
1411 Description: wait until IO finished before start new when do lock cancel.
1412 Details : VM protocol want old IO finished before start new, in this case
1413 need wait until PG_writeback is cleared until check dirty flag and
1414 call writepages in lock cancel callback.
1416 Severity : enhancement
1418 Description: using special macro for print time and cleanup in includes.
1422 Description: mds_mfd_close() ASSERTION(rc == 0)
1423 Details : In mds_mfd_close(), we need protect inode's writecount change
1424 within its orphan write semaphore to prevent possible races.
1428 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1432 Frequency : rare, on shutdown ost
1433 Description: don't hit live lock with umount ost.
1434 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1435 use shrink_dcache_sb instead.
1439 Description: don't panic with use echo client
1440 Details : echo client pass NULL as client nid pointer and this produce null
1441 pointer dereference.
1445 Description: fix build on ppc32
1446 Details : compile code with -m64 flag produce wrong object file for ppc32.
1450 Description: add message levels for liblustreapi
1454 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1455 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1456 will result in an -EAGAIN and in the exit call path we call
1457 alloc_rr() which will with increasing aggressiveness attempt to
1458 aquire precreated objects on the minimum number of required OSCs.
1462 Description: Use old size assignment to avoid deadlock
1463 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1464 the scheduling while holding a spinlock. We do not need locking
1465 for size in ll_update_inode() because size is only updated from
1466 the MDS for directories or files without objects, so there is no
1467 other place to do the update, and concurrent access to such inodes
1468 are protected by the inode lock.
1472 Description: resolve "_IOWR redefined" build error on SLES10
1476 Description: dump the memory debugging after all modules are unloaded to
1477 suppress false negative in conf_sanity test 39
1479 Severity : enhancement
1481 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1486 Description: fix SLES kernel versioning
1487 Details : the kernel version for our SLES 10 kernel did not include a "-"
1488 before the "smp" at the end. while this was not a problem in
1489 general, it did mean that software trying to use the kernel
1490 version to try to detect a vendor specific kernel would fail.
1491 this was most evident by the OFED build scripts.
1495 Description: Don't update lov_desc members until making sure they are valid
1496 Details : When updating lov_desc members via proc fs, need fix their
1497 validities before doing the real update.
1501 Description: don't put request into delay list while invalidate in flight.
1502 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1503 import in flight. this produce timeout for invalidate and sometimes
1504 can cause stale data.
1508 Frequency : on ppc only
1509 Description: not convert ost objects for directory because it's not exist.
1510 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1514 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1516 Severity : enhancement
1518 Description: Update to support 2.6.22.14 vanilla kernel.
1522 Frequency : rare, on recovery
1523 Description: read procfs can produce deadlock in some situation
1524 Details : Holding lprocfs lock which send rpc can produce block for destroy
1525 obd objects and this also block reconnect with -EALREADY. This
1526 isn't fix all lprocfs bugs - but make it rare.
1528 Severity : enhancement
1530 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1533 Frequency : frequent on X2 node
1535 Description: mdc_set_open_replay_data LBUG
1536 Details : Set replay data for requests that are eligible for replay.
1540 Description: lustre_mgs: operation 101 on unconnected MGS
1541 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1542 MGC, and late on MGC cannot successfully connect to MGS and a lot
1543 of the error messages complaining that MGS is not connected.
1547 Frequency : on network error
1548 Description: panic with double free request if network error
1549 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1550 but it's true only for synchronus enqueue, for async enqueue
1551 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1554 Severity : enhancement
1556 Description: client-side metadata stat-ahead during readdir(directory readahead)
1557 Details : perform client-side metadata stat-ahead when the client detects
1558 readdir and sequential stat of dir entries therein
1561 Frequency : on start mds
1563 Description: Implement get_info(last_id) in obdfilter.
1566 Frequency : occasional
1568 Description: Correctly check stale fid, not start epoch if ost not support SOM
1569 Details : open with flag O_CREATE need set old fid in op_fid3 because
1570 op_fid2 was overwritten with new generated fid, but mds can answer
1571 with one of these two fids and both is not stale. Setattr
1572 incorrectly started an epoch and assume will be called
1573 done_writing, but without SOM done_writing ever being called.
1576 Frequency : rare, depends on device drivers and load
1578 Description: MDS or OSS nodes crash due to stack overflow
1579 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1580 In some cases, in conjunction with device drivers that use a lot
1581 of stack the MDS (or possibly OSS) service threads could overflow
1582 the stack. One change which was identified to consume additional
1583 stack has been reworked to avoid the extra stack usage.
1586 Frequency : occasional
1588 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1589 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1590 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1591 oscc_has_objects and see OSC has no objests and call
1592 oscc_internal_create to resend create request.
1594 Severity : enhancement
1596 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1598 Severity : enhancement
1600 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1603 Frequency : very rare
1605 Description: Fix replay if there is an un-replied request and open
1606 Details : In some cases, older replay request will revert the
1607 mcd->mcd_last_xid on MDS which is used to record the client's
1608 latest sent request.
1610 Severity : enhancement
1612 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1614 Severity : enhancement
1616 Description: Add rhel5 support to HEAD.
1618 Serverity : enhancement
1620 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1625 Description: Don't allow skipping OSTs if index has been specified.
1626 Details : Don't allow skipping OSTs if index has been specified, make
1627 locking in internal create lots better.
1631 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1632 Details : access to bitfield in structure is always rounded to long
1633 and this produce problem with not atomic change any bit.
1637 Description: Lustre make rpms failed.
1638 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1639 builds Lustre rpms from tarball.
1642 Frequency : rare on shutdown ost
1644 Description: If llog cancel was not send before clean_exports phase, this can
1645 produce deadlock in llog code.
1646 Details : If llog thread has last reference to obd and call class_import_put
1647 this produce deadlock because llog_cleanup_commit_master wait when
1648 last llog_commit_thread exited, but this never success because was
1649 called from llog_commit_thread.
1653 Description: allow userland application know is lost one of stripes.
1654 Details : fill lvb_blocks with error code on ost and return it to
1655 application if error flag found.
1659 Description: NULL lov_tgts causing MDS oops
1660 Details : more safe checks for NULL lov_tgts for avoid oops.
1662 Severity : enhancement
1664 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1668 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1669 Details : If an OST is down the MDS will hang indefinitely in
1670 obd_statfs_rqset() waiting for the statfs data. While for
1671 MDS QOS usage of statfs, it should not stuck in waiting.
1673 Severity : enhancement
1675 Description: remote_acl support
1676 Details : Support ACL-based permission check for remote user.
1677 Support setfacl/getfacl for remote user with the utils
1678 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1679 the system "{s,g}etfacl" utils.
1681 Severity : enhancement
1683 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1685 Severity : enhancement
1687 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1691 Description: make mgs_setparam() handle fsname containing dash
1692 Details : fsname containing a dash does not work with lctl conf_param
1694 Severity : enhancement
1696 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1698 Severity : enhancement
1700 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1704 Frequency : rare, at shutdown
1705 Description: access already free / zero obd_namespace.
1706 Details : if client_disconnect_export was called without force flag set,
1707 and exist connect request in flight, this can produce access to
1708 NULL pointer (or already free pointer) when connect_interpret
1709 store ocd flags in obd_namespace.
1713 Frequency : only at startup
1714 Description: not alloc memory with spinlock held.
1715 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1716 if any spinlock held.
1718 Severity : enhancement
1720 Description: make lustre randomly fail allocating memory
1721 Details : Make lustre randomly failed allocating memory for testing purpose.
1723 Severity : enhancement
1725 Description: lost problems with lov objid file
1726 Details : Fixes some scability and access to not inited memory problems
1727 in work with lov objdid file.
1732 Description: lfs find does not continue on file error
1733 Details : Continue other files processing when a file/dir is absent.
1737 Description: Inconsistent usage of lustre_pack_reply()
1738 Details : Standardize the usage of lustre_pack_reply() such that it
1739 always generate a CERROR on failure.
1742 Frequency : occasional
1744 Description: MDS hang or stay in waiting lock
1745 Details : If client receive lock with CBPENDING flag ldlm need send lock
1746 cancel as separate rpc, to avoid situation when cancel request
1747 can't processed due all i/o threads stay in wait lock.
1751 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1755 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1759 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1763 Description: Update to RHEL4 latest kernel.
1765 Severity : enhancement
1767 Description: Build SLES10 patchless client fails
1768 Details : The configure was broken by run ./configure with
1769 --with-linux-obj=.... argument for patchless client. When the
1770 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1771 can't search header adequately. Use absolute path such as
1772 -I($LINUX)/include instead.
1776 Description: interrupt oig_wait produce painc on resend.
1777 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1778 private set, and this produce situation when rq_ptlrpcd_data not
1779 copyed to new allocated request and triggered LBUG on assert
1780 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1781 ptlrpcd set if request is changed and can be safety changed to use
1786 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1787 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1788 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1789 these macros and replace them with their difinitions instead.
1793 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1794 Details : Update which_patch & target file for SLES10 latest kernel.
1797 Description: Remove client patches from SLES 10 kernel.
1798 Details : This causes SLES 10 clients to behave as patchless clients
1799 even on a Lustre-patched (server) kernel.
1801 Severity : enhancement
1803 Description: self-adjustable client's lru lists
1804 Details : use adaptive algorithm for managing client cached locks lru
1805 lists according to current server load, other client's work
1806 pattern, memory activities, etc. Both, server and client
1807 side namespaces provide number of proc tunables for controlling
1810 Severity : enhancement
1812 Description: light-weight GSS support
1813 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1814 and reduce performance overhead.
1816 Severity : enhancement
1818 Description: Linux keyring support
1819 Details : Support using service of Linux keyring for Lustre GSS internal
1820 context refresh/cache mechanism.
1824 Description: Fix errors in lfs documentation
1825 Details : Fixes man pages
1829 Description: don't use GFP_* in generic Lustre code.
1830 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1834 Description: obdclass is limited by single OBD_ALLOC(idarray)
1835 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1839 Description: warnings with build patchless client with vanila 2.6.19 and up
1840 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1841 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1845 Frequency : Only for SLES
1846 Description: sanity_quota fail test_1
1847 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1848 file, which makes TSTID[2] inunique.
1852 Frequency : Only for SLES9
1853 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1854 Details : Update target/ChangeLog/which_patch .
1858 Frequency : Only for test_10 in sanity-quota.sh
1859 Description: a bug in quota test code
1860 Details : checking a test flag in a wrong place
1864 Frequency : Only for test_13 in sanity-quota.sh
1865 Description: a bug in quota test script
1866 Details : shouldn't check overfull allocated quota
1870 Description: jbd statistics
1871 Details : Port older jbd statistics patch for sles10
1873 Severity : enhancement
1875 Description: Add printing inode info into message about error in writepage.
1880 Description: reply_lock_interpret crash due to race with it and lock cancel.
1881 Details : Do not replay locks that are being cancelled. Do not reference
1882 locks by their address during replay, just by their handle.
1886 Frequency : When flocks are used.
1887 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1888 Details : Flock locks might destroy just granted lock if it could be merged
1889 with another existing flock, this is done in completion handler,
1890 so teach ldlm_cli_enquque_fini that this is a valid case for
1896 Description: Oops in read and write path when failing to allocate lock.
1897 Details : Check if lock allocation failed and return error back.
1901 Description: lstripe command fails for valid OST index
1902 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1903 instead of lov->desc.ld_active_tgt_count.
1907 Description: sanity.sh failed test 103
1908 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1913 Description: df doesn't work properly if diskfs blocksize != 4K
1914 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1918 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1919 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1921 Severity : enhancement
1923 Description: Adaptive timeouts
1924 Details : RPC timeouts adapt to changing server load and network
1925 conditions to reduce resend attempts and improve recovery time.
1929 Description: llapi_file_create() does not allow some changes
1930 Details : add llapi_file_open() that allows specifying the mode and
1931 open flags, and also returns an open file handle.
1935 Description: merge and cleanup kernel patches.
1936 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1940 Description: Add journal checksum support.(Kernel part)
1941 Details : The journal checksum feature adds two new flags i.e
1942 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1943 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1944 indicates that the commit block contains the checksum for
1945 the blocks described by the descriptor blocks. Now commit
1946 record can be sent to disk without waiting for descriptor
1947 blocks to be written to disk. This behavior is controlled
1948 using JBD2_FEATURE_ASYNC_COMMIT flag.
1952 Description: OSS needs mutliple precreate threads
1953 Details : Add ability to start more than one create thread per OSS.
1957 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1958 Details : execute lfs setstripe on client
1962 Description: mds_obd_create error creating tmp object
1963 Details : When the user sets quota on root, llog will be affected and can't
1964 create files and write files.
1967 Frequency : Always on ia64 patchless client, and possibly others.
1969 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1970 Details : This allows the patchless client to be loaded on architectures
1971 without this export.
1975 Description: RedHat Update kernel for RHEL5
1976 Details : Add the kernel config file for RHEL5.
1980 Description: RedHat Update kernel for RHEL5
1981 Details : Modify the kernel config file more closer RHEL5.
1985 Description: Build failure against Centos5 (RHEL5)
1986 Details : Define PAGE_SIZE when it isn't present.
1990 Description: test_7 on sanity-quota.sh can't be used on separate machines
1991 Details : add facet to handle it
1995 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1996 Details : with NFS, the anon dentry's parent was set to itself in
1997 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1998 corresponding dentry other than use rec->ur_name.
2000 Severity : enhancement
2002 Description: lfs setstripe enhancement
2003 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2007 Description: enable data checksumming by default
2008 Details : enable checksum by default, allow --disable-checksum
2009 configure option and "-o nochecksum" mount option. Checksums
2010 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2014 Description: lustre support for RHEL5
2015 Details : Add support for RHEL5.
2019 Description: Client eviction due to failover config
2020 Details : after a connection loss, the lustre client should attempt to
2021 reconnect to the last active server first before trying the
2022 other potential connections.
2026 Description: when mds and osts use different quota unit(32bit and 64bit),
2027 quota will be released repeatly.
2028 Details : void sending multiple quota reqs to mds, which will keep the
2029 status between the reqs.
2033 Description: osts not allocated evenly to files
2034 Details : change the condition to increase offset_idx
2037 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2039 Description: Data corruption for OSTs that are formatted larger than 2TB
2041 Details : When generating the bio request for lustre file writes the
2042 sector number would overflow a temporary variable before being
2043 used for the IO. The data reads correctly from Lustre (which
2044 will overflow in a similar manner) but other file data or
2045 filesystem metadata may be corrupted in some cases.
2049 Description: Tune the kernel for good SCSI performance.
2050 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2051 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2056 Description: rewrite ext2-derived code in obdclass/uuid.c
2057 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2058 from scratch preserving functionality.
2062 Description: "lfs find -obd UUID" prints directories
2063 Details : "lfs find -obd UUID" will return all directory names instead
2064 of just file names. It is incorrect because the directories
2065 do not reside on the OSTs.
2069 Description: use i_size_read and i_size_write in 2.6 port
2070 Details : replace inode->i_size access with i_size_read/write()
2073 Frequency : when removing large files
2075 Description: scheduling issue during removal of large Lustre files
2076 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2077 It causes scheduling issues when removing large files (17TB in the
2081 Frequency : only with liblustre clients on XT3
2083 Description: evictions taking too long
2084 Details : allow llrd to evict clients directly on OSTs
2087 Frequency : only on ppc
2089 Description: /proc/fs/lustre/devices broken on ppc
2090 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2091 We should make sure the type of 'index' is loff_t and then cast
2092 explicitly as needed below. Do not assign an explicitly cast
2096 Frequency : only for rhel5
2098 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2099 Details : Modify the target file & which_kernel.
2101 Severity : enhancement
2103 Description: omit set fsid for export NFS
2104 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2105 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2109 Frequency : Always, for kernels after 2.6.16
2110 Description: Fix warning idr_remove called for id=.. which is not allocated.
2111 Details : Last kernels save old s_dev before kill super and not allow
2112 to restore from callback - restore it before call kill_anon_super.
2116 Description: buffer overruns could theoretically occur
2117 Details : llapi_semantic_traverse() modifies the "path" argument by
2118 appending values to the end of the origin string, and a
2119 overrun may occur. Adding buffer overrun check in liblustreapi.
2123 Description: Fix error on 'ls .' at the top of the Lustre mount.
2124 Details : Don't revalidate dentry if it is a root dentry.
2128 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2129 Details : Modify vm-tunables-rhel4.patch.
2133 Description: Kernel config for 2.6.18-vanilla.
2134 Details : Modify targets/2.6-vanilla.target.in.
2135 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2136 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2137 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2138 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2142 Description: improve handling recoverable errors
2143 Details : if request processig with error which can be recoverable on server
2144 request should be resend, otherwise page released from cache and
2149 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2150 Details : Modify target file & which_patch.
2151 A flaw was found in the IA32 system call emulation provided
2152 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2153 value could be stored in the %RAX register, which could trigger an
2154 out-of-bounds system call table access. An untrusted local user
2155 could exploit this flaw to run code in the kernel
2156 (ie a root privilege escalation). (CVE-2007-4573).
2160 Description: O_DIRECT bypasses client statistics.
2161 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2162 read_ahead_stats were updated. copied the stats section from
2163 osc_send_oap_rpc() into async_internal().
2167 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2171 Description: change order of libsysio includes
2172 Details : '#include sysio.h' should always come before '#include xtio.h'
2176 Description: handle "serious error: objid * already exists" more gracefully
2177 Details : If LAST_ID value on disk is smaller than the objects existing in
2178 the O/0/d* directories, it indicates disk corruption and causes an
2179 LBUG(). If the object is 0-length, then we should use the existing
2180 object. This will help to avoid a full fsck in most cases.
2182 Severity : enhancement
2184 Description: adapt the lustre_config script to support the upgrade case
2185 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2186 server targets to 1.6 in parallel.
2190 Description: To avoid grant space > avaible space when the disk is almost
2191 full. Without this patch you might see the error "grant XXXX >
2192 available" or some LBUG about grant, when the disk is almost
2194 Details : In filter_check_grant, for non_grant cache write, we should
2195 check the left space by if (*left > ungranted + bytes), instead
2196 of (*left > ungranted), because only we are sure the left space
2197 is enough for another "bytes", then the ungrant space should be
2198 increase. In client, we should update cl_avail_grant only there
2199 is OBD_MD_FLGRANT in the reply.
2203 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2204 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2205 described in CVE-2007-4573. This problem could allow untrusted
2206 local users to gain root access.
2209 Frequency : when using O_DIRECT and quotas
2211 Description: Incorrect file ownership on O_DIRECT output files
2212 Details : block usage reported by 'lfs quota' does not take into account
2213 files that have been written with O_DIRECT.
2218 Description: touch file failed when fs is not full
2219 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2220 otherwise we can get ENOSP while fs is not full.
2224 Description: parallel lock callbacks
2225 Details : Instead of sending blocking and completion callbacks as separated
2226 requests, adding them to a set and sending in parallel.
2229 Frequency : only for Cray XT3
2230 Bugzilla : 12829/13455
2231 Description: Changing primary group doesn't change the group lustre assigns to
2233 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2234 overridden with the primary group provided by the group upcall,
2235 whereas the supplied fsgid can be trusted if it is in the list of
2236 supplementary groups returned by the group upcall.
2238 Severity : enhancement
2240 Description: Allow masking D_WARNING, D_ERROR messages from console
2241 Details : Console messages can now be disabled via lnet.printk.
2245 Description: User code with malformed file open parameter crashes client node
2246 Details : Before packing join_file req, all the related reference should be
2247 checked carefully in case some malformed flags cause fake
2248 join_file req on client.
2252 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2253 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2254 mode is equal to the granted mode. However, it should also check
2255 that LDLM_FL_CP_REQD is not set.
2259 Description: Heavy nfs access might result in deadlocks
2260 Details : After ELC code landed, it is now improper to enqueue any mds
2261 locks under och_sem, because enqueue might want to decide to
2262 cancel open locks for same inode we are holding och_sem for.
2266 Description: Client eviction while running blogbench
2267 Details : A lot of unlink operations with concurrent I/O can lead to a
2268 deadlock causing evictions. To address the problem, the number of
2269 oustanding OST_DESTROY requests is now throttled to
2270 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2271 callbacks are processed in priority.
2275 Description: enable ACLs on MDS by default
2276 Details : ACLs must be enabled on MDS by default.
2279 Frequency : PPC/PPC64 only
2281 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2282 Details : fix duplicated definitions between asm-ppc64/types.h and
2283 lustre_types.h on PPC.
2286 Frequency : PPC/PPC64 only
2288 Description: asm-ppc/segment.h does not exist
2289 Details : fix compile issue on PPC.
2293 Description: better handle error messages in extents code
2296 Frequency : RHEL4 only
2298 Description: mkfs is very slow on IA64/RHEL4
2299 Details : A performance regression has been discovered in the MPT Fusion
2300 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2301 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2302 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2304 Severity : enhancement
2306 Description: SNMP support enhancement
2307 Details : Adding total number of sampled request for an MDS node in snmp
2310 Severity : enhancement
2312 Description: Optimize ldlm waiting list processing for PR extent locks
2313 Details : When processing waiting list for read extent lock and meeting read
2314 lock that is same or wider to it that is not contended, skip
2315 processing rest of the list and immediatelly return current
2316 status of conflictness, since we are guaranteed there are no
2317 conflicting locks in the rest of the list.
2321 Description: Time out and refuse to reconnect
2322 Details : When the failover node is the primary node, it is possible
2323 to have two identical connections in imp_conn_list. We must
2324 compare not conn's pointers but NIDs, otherwise we can defeat
2325 connection throttling.
2329 Description: port llog fixes from b1_6 into HEAD
2330 Details : Port llog reference couting and some llog cleanups from b1_6
2331 (bug 10800) into HEAD, for protect from panic and access to already
2332 free llog structures.
2336 Description: Detect stride IO mode in read-ahead
2337 Details : When a client does stride read, read-ahead should detect that and
2338 read-ahead pages according to the detected stride pattern.
2342 Description: data checksumming impacts single node performance
2343 Details : add support for several checksum algorithm. Currently, only CRC32
2344 and Adler-32 are supported. The checksum type can be changed on
2345 the fly via /proc/fs/lustre/osc/*/checksum_type.
2349 Description: use adler32 for page checksums
2350 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2355 Description: build for x2 fails
2356 Details : fix compile issue on Cray systems.
2360 Description: Properly match for duplicate locks
2361 Details : Due to different lock order from skiplists code, we need to
2362 traverse entire list for now
2365 Frequency : only on PPC/SLES10
2367 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2368 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2369 be defined. Add a hack in mkfs_lustre.c to work around this compile
2374 Description: LASSERT on MDS when client holding flock lock dies
2375 Details : ldlm pool logic depends on number of granted locks equal to
2376 number of released locks which is not true for flock locks, so
2377 just exclude such locks from consideration.
2381 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2382 Details : Use fsfilt_sync() for both the whole filesystem sync and
2383 individual file sync to eliminate dangerous inode locking
2384 with I_LOCK that can lead to a deadlock.
2388 Description: performance in 1.6.3
2389 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2390 sure that 1MB requests are not fragmented by the block layer.
2392 Severity : enhancement
2394 Description: organize the server-side client stats on per-nid basis
2395 Details : Change the structure of stats under obdfilter and mds to
2403 The "uuid"s file would list the uuids of _active_ exports.
2404 And the clear entry is to clear all stats and stale nids.
2406 Severity : enhancement
2408 Description: eliminate client locks in face of contention
2409 Details : file contention detection and lockless i/o implementation
2410 for contended files.
2414 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2417 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2419 Description: MDS may lose file striping (and hence file data) in some cases
2420 Details : If there are additional extended attributes stored on the MDS,
2421 in particular ACLs, SELinux, or user attributes (if user_xattr
2422 is specified for the client mount options) then there is a risk
2423 of attribute loss. Additionally, the Lustre file striping
2424 needs to be larger than default (e.g. striped over all OSTs),
2425 and an additional attribute must be stored initially in the
2426 inode and then increase in size enough to be moved to the
2427 external attribute block (e.g. ACL growing in size) for file
2432 Description: skiplist implementation simplification
2433 Details : skiplists are used to group compatible locks on granted list
2434 that was implemented as tracking first and last lock of each
2435 lock group the patch changes that to using doubly linked lists
2439 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2440 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2441 because LASSERT on both the data supplied by a client, and the
2442 data on disk is dangerous and incorrect.
2444 Severity : enhancement
2446 Description: Slow truncate/writes to huge files at high offsets.
2447 Details : Directly associate cached pages to lock that protect those pages,
2448 this allows us to quickly find what pages to write and remove
2449 once lock callback is received.
2453 Description: more ldlm soft lockups
2454 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2455 starve other threads from the resource lock for a long time in
2456 case of long waiting queue, so change the debug level from
2457 D_OTHER to the less frequently used D_INFO.
2459 Severity : enhancement
2461 Description: add -gid, -group, -uid, -user options to lfs find
2465 Description: Hung threads in invalidate_inode_pages2_range
2466 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2467 once one is completed. As a result, some RPCs are stuck in the
2468 queue and are never sent.
2472 Description: filter threads hungs on waiting journal commit
2473 Details : Cleanup filter group llog code, then only filter group llog will
2474 be only created in the MDS/OST syncing process.
2478 Description: Procfs and llog threads access destoryed import sometimes.
2479 Details : Sync the import destoryed process with procfs and llog threads by
2480 the import refcount and semaphore.
2482 Severity : enhancement
2484 Description: openlock cache of b1_6 port to HEAD
2489 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2490 Details : If insertion of an extent fails, then discard the inode
2491 preallocation and free data blocks else it can lead to duplicate
2496 Description: don't always update ctime in ext3_xattr_set_handle()
2497 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2498 In some cases the ctime should not be updated, for example for
2499 2.0->1.8 compatibility it is necessary to delete an xattr and it
2500 should not update the ctime.
2504 Bugzilla : 15713/16362
2505 Description: Assertion in iopen_connect_dentry in 1.6.3
2506 Details : looking up an inode via iopen with the wrong generation number can
2507 populate the dcache with a disconneced dentry while the inode
2508 number is in the process of being reallocated. This causes an
2509 assertion failure in iopen since the inode's dentry list contains
2510 both a connected and disconnected dentry.
2514 Description: assertion failure in ldlm_handle2lock()
2515 Details : fix a race between class_handle_unhash() and class_handle2object()
2516 introduced in lustre 1.6.5 by bug 13622.
2521 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2522 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2527 Description: oops in sd_iostats_seq_show()
2528 Details : unloading/reloading the scsi low level driver triggers a kernel
2529 bug when trying to access the sd iostat file.
2534 Description: Kernel panics during QLogic driver reload
2535 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2536 patch, causing memory corruption.
2541 Description: journal_dev option does not work in b1_6
2542 Details : pass mount option during pre-mount.
2544 Severity : enhancement
2546 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2547 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2548 extent information of a file. It can be used to map logical blocks
2549 in a file to physical blocks in the block device.
2553 Description: LDLM soft lockups - improvement
2554 Details : It is be possible to send the lock handle along with each read
2555 or write request because the client is already doing a lock match
2556 itself so there isn't any reason the OST should have to re-do that
2560 Frequency : only on Cray X2
2562 Description: X2 build failures
2563 Details : fix build failures on Cray X2.
2567 Description: xid & resent requests
2568 Details : Initialize RPC XID from clock at startup (randomly if clock is
2571 Severity : enhancement
2573 Description: Add lustre_start utility to start or stop multiple Lustre servers
2578 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2579 Details : In case of memory pressure, list_del() can be called twice on
2580 req->rq_history_list, causing a kernel oops.
2584 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2585 Details : fix stack overflow in the distributed lock manager by defering
2586 export eviction after a failed AST to the elt thread instead of
2587 handling it in the dlm interpret routine.
2591 Description: Convert some comments to new format.
2592 Details : Update documenting comments to match doxygen conventions.
2596 Description: Grammar fixes.
2597 Details : A couple of trivial spelling fixes.
2601 Description: OSD_COUNTERS-mandatory
2602 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2603 to simplify the code.
2607 Description: simplify cmm_device freeing logic.
2608 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2612 Description: Add lockdep support to dt_object_operations locking interface.
2613 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2614 parameter indicating lock ordering. Update mdd code to use new
2619 Description: Introduce failloc constants for lockless IO tests.
2620 Details : Add two new failloc constants to test lockless IO. Only one of
2621 them in implemented---another is checked in yet to be landed
2626 Description: Add lockdep support for inode mutex.
2627 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2628 in the situations where Lustre has to lock more than one inode
2633 Description: Add optional invariants checking support.
2634 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2635 configure switch. This macro is to be used for consistency and
2636 sanity checks that are too expensive to be left in `production'
2641 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2642 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2647 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2648 Details : Use spin_lock_nested() in (the only) situation where more than
2649 one ldlm_lock is locked simultaneously. Also, fix possible
2650 dead-lock in ldlm_lock_change_resource() by enforcing particular
2655 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2656 Details : Instead of specifying each ldlm_lock call-back through separate
2657 parameter, wrap them into struct ldlm_callback_suite.
2661 Description: Kill join_lru obd method and its callers.
2662 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2663 in memory---a requirement implemented through ->o_join_lru() obd
2664 method in HEAD. Remove this method and its users.
2668 Description: Add asynchronous ldlm ENQUEUE completion handler.
2669 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2670 case that stock ldlm_completion_ast() cannot handle as it waits
2671 until lock is granted. Introduce new ldlm_completion_ast_async()
2672 for this. Also comment ldlm_completion_ast().
2676 Description: ldlm_error <-> errno conversion.
2677 Details : Add functions to map (rather arbitrary) between LDLM error codes
2678 and standard errno values. CLIO needs this to prevent LDLM specific
2679 constants from escaping ldlm and osc.
2683 Description: Kill unused ldlm_handle2lock_ns() function.
2684 Details : Kill unused ldlm_handle2lock_ns() function.
2688 Description: Add lu_ref support to ldlm_lock
2689 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2690 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2691 and ldlm_resource. LDLM interface has to be changed, because code that
2692 releases a reference on a lock, has to "know" what reference this is.
2693 In the most frequent case
2695 lock = ldlm_handle2lock(handle);
2697 LDLM_LOCK_PUT(lock);
2699 no changes are required. When any other reference (received _not_ from
2700 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2701 instead of LDLM_LOCK_PUT().
2703 Arguably, changes are pervasive, and interface requires some discipline
2704 for proper use. On the other hand, it was very instrumental in finding
2705 a few leaked lock references.
2709 Description: Add ldlm_lock_addref_try().
2710 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2711 attempts to addref a lock that might be being canceled
2716 Description: Add ldlm_weigh_callback().
2717 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2718 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2719 hard-coded `number of pages' logic.
2723 Description: Add lockdep annotations to llog code.
2724 Details : Use appropriately tagged _nested() locking calls in the places
2725 where llog takes more than one ->lgh_lock lock.
2729 Description: Add loi_kms_set().
2730 Details : Wrap kms updates into a helper function.
2734 Description: Constify instances of struct lsm_operations.
2735 Details : Constify instances of struct lsm_operations.
2739 Description: lu_conf support.
2740 Details : On a server, a file system object is uniquely identified
2741 by a fid, which is sufficient to locate and load all object
2742 state (inode). On a client, on the other hand, more data are
2743 necessary instantiate an object. Change lu_object_find() and
2744 friends to take additional `lu_conf' argument describing object.
2745 Typically this includes layout information.
2749 Description: lu_context fixes.
2750 Details : Introduce new lu_context functions that are needed on the client
2751 side, where some system threads (ptlrpcd) are shared by multiple
2752 modules, and so cannot be stopped during module shutdown.
2756 Description: Add start and stop methods to lu_device_type_operations.
2757 Details : Introduce two new methods in lu_device_type_operations, that are
2758 invoked when first instance of a given type is created and last one
2759 is destroyed respectively. This is need by CLIO.
2763 Description: Add lu_ref support to struct lu_device.
2764 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2765 track leaked references.
2769 Description: Introduce lu_kmem_descr.
2770 Details : lu_kmem_descr and its companion interface allow to create
2771 and destroy a number of kmem caches at once.
2775 Description: Fix lu_object finalization race.
2776 Details : Fix a race between lu_object_find() finding an object and its
2777 concurrent finalization. This race is (most likely) not possible
2778 on the server, but might happen on the client.
2782 Description: Introduce lu_ref interface.
2783 Details : lu_ref is a debugging module allowing to track references to
2784 a given object. It is quite cpu expensive, and has to be
2785 explicitly enabled with --enable-lu_ref. See usage description
2790 Description: Factor lu_site procfs stats into a separate function.
2791 Details : Separate lu_site stats printing code into a separate function
2792 that can be reused on a client.
2796 Description: Constify instances of struct {lu,dt,md}_device_operations.
2797 Details : Constify instances of struct {lu,dt,md}_device_operations.
2801 Description: Introduce struct md_site and move meta-data specific parts of
2802 struct lu_site here.
2803 Details : Move md-specific fields out of struct lu_site into special struct
2804 md_site, so that lu_site can be used on a client.
2808 Description: Kill mdd_lov_destroy().
2809 Details : Remove unused mdd code.
2813 Description: Add st_block checking to multistat.c.
2814 Details : Add st_block checking to multistat.c.
2818 Description: Add lu_ref support to struct obd_device.
2819 Details : Add lu_ref tracking to obd_device.
2823 Description: Kill obd_set_fail_loc().
2824 Details : Remove unused code.
2828 Description: Add special type for ptlrpc_request interpret functions.
2829 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2830 there. Actual usage will be in CLIO.
2834 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2835 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2836 doesn't work with lockdep.
2840 Description: Add rwv.c test program.
2841 Details : New testing program exercising readv(2) and writev(2) (Qian).
2845 Description: Add sendfile.c test program.
2846 Details : New testing program exercising sendfile(2) (Jay).
2850 Description: Ratelimit a message that can be very frequent.
2851 Details : Ratelimit a memory allocation failure message that can
2856 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2857 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2861 Description: Liblustre build fixes.
2862 Details : Liblustre build fixes.
2866 Description: libcfs: add cfs_{need,cond}_resched() interface.
2867 Details : libcfs: add cfs_{need,cond}_resched() definition and
2868 implementations for Linux, NT, and liblustre.
2870 Severity : enhancement
2872 Description: More exported tunables for mballoc
2873 Details : Add support for tunable preallocation window and new tunables for
2874 large/small requests
2878 Description: Detect corruption of block bitmap and checking for preallocations
2879 Details : Checks validity of on-disk block bitmap. Also it does better
2880 checking of number of applied preallocations. When corruption is
2881 found, it turns filesystem readonly to prevent further corruptions.
2885 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2886 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2887 window to 0. If it does hit cache miss, and read-pattern is still
2888 stride-io mode, does not reset the stride window, but also does not
2889 increase the stride window length in this case.
2893 Frequency : only for big-endian servers
2894 Description: Check if system is big-endian while mounting fs with extents feature
2895 Details : Mounting a filesystem with extents feature will fail on big-endian
2896 systems since ext3-based ldiskfs is not supported on big-endian
2897 systems. This can be over-riden with "bigendian_extents" mount option.
2899 Severity : enhancement
2901 Description: The root squash functionality
2902 Details : A security feature, which is to prevent users from being able
2903 to mount lustre on their desktop, run as root, and delete
2904 all of the files in the filesystem. The goal is accomplished by
2905 remapping user id (UID) and group id (GID) of the root user to
2906 a UID and GID specified by the system administartor via Lustre
2907 configuration management server (MGS). The functionality also
2908 allows to specify sets of clients for which the remapping does
2913 Description: Excessive recovery window
2914 Details : With AT enabled, the recovery window can be excessively long (6000+
2915 seconds). To address this problem, we no longer use
2916 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2917 timeout no longer depends on the service time, it is set to
2918 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2919 time via pb_service_time.
2923 Description: Watchdog triggered on MDS failover
2924 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2925 the OSTs know that the MDS "UUID" can be reused for the same export
2926 from a different NID, so we do not need to wait for the export to be
2930 Frequency : rare, only if using MMP with Linux RAID
2932 Description: MMP doesn't work with Linux RAID
2933 Details : While using HA for Lustre servers with Linux RAID, it is possible
2934 that MMP will not detect multiple mounts. To make this work we
2935 need to unplug the device queue in RAID when the MMP block is being
2936 written. Also while reading the MMP block, we should read it from
2937 disk and not the cached one.
2939 Severity : enhancement
2941 Description: open file using fid
2942 Details : A file can be opened using just its fid, like
2943 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2946 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2949 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2950 Details : add proper locking for creation of jbd_4k slab cache
2954 Description: MMP check in ext3_remount() fails without displaying any error
2955 Details : When multiple mount protection fails during remount, proper error
2958 Severity : enhancement
2960 Description: Allow stripe size to be up to 4G-64k
2961 Details : Fix math logic to allow large stripe sizes.
2965 Description: add check for >8TB ldiskfs filesystems
2966 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2967 Don't allow >8TB ldiskfs filesystems to be mounted without
2968 force_over_8tb mount option
2973 Descriptoin: OSS complains "too many blocks" and crashes
2974 Details : In filter_commitrw_write():
2975 1. lquota_chkquota();
2976 2. obdo_from_inode();
2977 3. lquota_pending_commit();
2978 "2." has changed uid and gid in this case, then caused
2979 lquota_chkquota() and lquota_pending_commit() to use
2980 different uid and gid. That is the root of the bug.
2984 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2985 Details : Detect and print out the 'stripe count = -1' case properly.
2989 Description: Out or order replies might be lost on replay
2990 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2991 request with tid smaller than one currently being added, add it
2992 to the start, not end of the list.
2994 --------------------------------------------------------------------------------
2996 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
2998 * Support for kernels:
2999 2.6.5-7.283 (SLES 9),
3000 2.6.9-55.EL (RHEL 4),
3001 2.6.16.46-0.14 (SLES 10),
3002 2.6.18.8 vanilla (kernel.org)
3003 * Client support for unpatched kernels:
3004 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3005 2.6.16 - 2.6.22 vanilla (kernel.org)
3006 * Due to recently discovered recovery problems, we do not recommend
3007 using patchless RHEL 4 clients with this or any earlier release.
3008 * Recommended e2fsprogs version: 1.39.cfs8
3009 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3010 * Starting with this release, the ldiskfs backing filesystem required
3011 by Lustre is now in its own package, lustre-ldiskfs. This package
3012 should be installed. It is versioned separately from Lustre and
3013 may be released separately in future.
3017 Description: block reactivating mgc import until all deactivates complete
3018 Details : Fix race when failing back MDT/MGS to itself (testing)
3021 Frequency : at statup only
3023 Description: mds_lov_synchronize race leads to various problems
3024 Details : simultaneous MDT->OST connections at startup can cause the
3025 sync to abort, leaving the OSC in a bad state.
3027 Severity : enhancement
3029 Description: add optional extra BUILD_VERSION info
3030 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3031 to override the lustre version.
3034 Frequency : 2.6.18 servers only
3036 Description: ll_kern_mount() doesn't release the module reference
3037 Details : The ldiskfs module reference count never drops down to 0
3038 because ll_kern_mount() doesn't release the module reference.
3043 Description: server LBUG when using old ost_num_threads parameter
3044 Details : Accept the old ost_num_threads parameter but warn that it
3045 is deprecated, and fix an off-by-one error that caused an LBUG.
3050 Description: Transient SCSI error results in persistent IO issue
3051 Details : iobuf->dr_error is not reinitialized to 0 between two
3055 Frequency : sometimes when underlying device returns I/O errors
3057 Description: OSTs not going read-only during write failures
3058 Details : OSTs are not remounted read-only when the journal commit threads
3059 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3060 instead of the ext3 wrappers.
3064 Description: poor connect scaling with increasing client count
3065 Details : Don't run filter_grant_sanity_check for more than 100 exports
3066 to improve scaling for large numbers of clients.
3069 Frequency : SLES10 only
3071 Description: sanity-quota.sh quotacheck failed: rc = -22
3072 Details : Quotas cannot be enabled on SLES10.
3075 Frequency : liblustre clients only
3077 Description: getdirentries does not give error when run on compute nodes
3078 Details : getdirentries does not fail when the size specified as an argument
3079 is too small to contain at least one entry
3081 Severity : enhancement
3083 Description: Add LNET router traceability for debug purposes
3084 Details : If a checksum failure occurs with a router as part of the
3085 IO path, the NID of the last router that forwarded the bulk data
3086 is printed so it can be identified.
3091 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3092 Details : Due to a race condition, liblustre clients were occasionally
3093 evicted incorrectly.
3095 Severity : enhancement
3097 Description: lfs setstripe use optional parameters instead of postional
3100 Severity : enhancement
3102 Description: Nanosecond timestamp support for ldiskfs
3103 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3104 resolution timestamps. There is not yet support for this at
3105 the Lustre filesystem level.
3108 Frequency : during server recovery
3110 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3111 Details : request with rq_no_resend flag not awake l_wait_event if they get
3115 Frequency : nfs export on patchless client
3117 Description: connectathon hang when test nfs export over patchless client
3118 Details : Disconnected dentry cannot be found with lookup, so we do not need
3119 to unhash it or make it invalid
3122 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3123 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3124 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3129 Description: open req refcounting wrong on reconnect
3130 Details : If reconnect happened between getting open reply from server and
3131 call to mdc_set_replay_data in ll_file_open, we will schedule
3132 replay for unreferenced request that we are about to free.
3133 Subsequent close will crash in variety of ways.
3134 Check that request is still eligible for replay in
3135 mdc_set_replay_data().
3140 Description: disable writes to filesystem when reading health_check file
3141 Details : the default for reading the health_check proc file has changed
3142 to NOT do a journal transaction and write to disk, because this
3143 can cause reads of the /proc file to hang and block HA state
3144 checking on a healthy but otherwise heavily loaded system. It
3145 is possible to return to the previous behaviour during configure
3146 with --enable-health-write.
3148 Severity : enhancement
3150 Description: 64-bit inode version
3151 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3152 to the inode. This will be required for version-based recovery.
3157 Description: MDS fails to start if a duplicate client export is detected
3158 Details : in some rare cases it was possible for a client to connect to
3159 an MDS multiple times. Upon recovery the MDS would detect this
3160 and fail during startup. Handle this more gracefully.
3162 Severity : enhancement
3164 Description: Add -o localflock option to simulate old noflock
3166 Details : This will achieve local-only flock/fcntl locks
3172 Description: log_commit_thread vs filter_destroy race leads to crash
3173 Details : Take import reference before releasing llog record semaphore
3178 Description: Wrong request locking in request set processing
3179 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3180 imp_delayed_list, in this place should be used imp_lock.
3183 Frequency : when reconnection
3185 Description: Grant Leak when osc reconnect to OST
3186 Details : When osc reconnect ost, OST(filter) should check whether it
3187 should grant more space to client by comparing fed_grant and
3188 cl_avail_grant, and return the granted space to client instead
3189 of "new granted" space, because client will call osc_init_grant
3190 to update the client grant space info.
3193 Frequency : when client reconnect to OST
3195 Description: Grant Leak when osc do resend and replay bulk write
3196 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3197 bulk write request, because the grant info while be sync between
3198 OSC and OST when reconnect, and we should ignore the grant info
3199 these of resend/replay write req.
3204 Description: Grant space more than available left space sometimes.
3205 Details : When then OST is about to be full, if two bulk writing from
3206 different clients came to OST. Accord the avaliable space of the
3207 OST, the first req should be permitted, and the second one
3208 should be denied by ENOSPC. But if the seconde arrived before
3209 the first one is committed. The OST might wrongly permit second
3210 writing, which will cause grant space > available space.
3213 Frequency : when client is evicted
3215 Description: Grant might be wrongly erased when osc is evicted by OST
3216 Details : when the import is evicted by server, it will fork another
3217 thread ptlrpc_invalidate_import_thread to invalidate the
3218 import, where the grant will be set to 0. While the original
3219 thread will update the grant it got when connecting. So if
3220 the former happened latter, the grant will be wrongly errased
3221 because of this race.
3226 Description: Checking Stale with correct fid
3227 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3228 to check whether it is stale, because sometimes, we want the
3229 enqueue happened anyway, and op_data.fid2 will not be initialized.
3231 Severity : enhancement
3233 Description: update patchless client
3234 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3237 Frequency : only with 2.4 kernel
3239 Description: random memory corruption
3240 Details : size of struct ll_inode_info is to big for union inode.u and this
3241 can be cause of random memory corruption.
3246 Description: Memory leak in recovery
3247 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3248 It should also check obd_fail before fsfilt_start, otherwise if
3249 fsfilt_start return -EROFS,(failover mds during mds recovery).
3250 then the req will return with repmsg->transno = 0 and rc = EROFS.
3251 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3252 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3253 be freed no matter whether fsfilt_commit success or not.
3256 Frequency : only with huge count clients
3258 Description: Prevents from taking the superblock lock in llap_from_page for
3260 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3261 for llap_from_page call in ll_removepage prevents from taking the
3262 superblock lock for a soon died page.
3267 Description: Not check open intent error before release open handle
3268 Details : in some rare cases, the open intent error is not checked before
3269 release open handle, which may cause
3270 ASSERTION(open_req->rq_transno != 0), because it tries to release
3271 the failed open handle.
3276 Description: Set cat log bitmap only after create log success.
3277 Details : in some rare cases, the cat log bitmap is set too early. and it
3278 should be set only after create log success.
3282 Description: Accessing a block bevice can re-enable I/O when Lustre is
3283 tearing down a device.
3284 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3289 Description: service threads may hog cpus when there are a lot of requests
3290 Details : Insert cond_resched to give other threads a chance to use some CPU
3295 Description: the cat log was not initialized in recovery
3296 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3297 unlink log on mds will not be initialized until mds post
3298 recovery. And also in mds post recovery, the unlink log will
3299 initialization will be done asynchronausly, so there will be race
3300 between add unlink log and unlink log initialization.
3304 Description: brw_stats were being printed incorrectly
3305 Details : brw_stats were being printed as log2 but all of them were not
3306 recorded as log2. Also remove some code duplication arising from
3307 filter_tally_{read,write}.
3311 Frequency : rare, only in recovery.
3312 Description: ASSERTION(req->rq_type != LI_POISON) failed
3313 Details : imp_lock should be held while iterating over imp_sending_list for
3314 prevent destroy request after get timeout in ptlrpc_queue_wait.
3318 Description: replay-single.sh test 52 fails
3319 Details : A lock's skiplist need to be cleanup when it being unlinked
3320 from its resource list.
3324 Description: Short directio read returns full requested size rather than
3326 Details : Direct I/O operations should return actual amount of bytes
3327 transferred rather than requested size.
3329 Severity : enhancement
3331 Description: metadata RPC reduction (e.g. for rm performance)
3332 Details : decrease the amount of synchronous RPC between clients and servers
3333 by canceling conflicing lock before the operation on the client
3334 and packing thier handles into the main operation RPC to server.
3336 Severity : enhancement
3338 Description: Async OSC create to avoid the blocking unnecessarily.
3339 Details : If an OST has no remaining object, system will block on the
3340 creation when it needs to create a new object on this OST. Now,
3341 ways use pre-created objects when available, instead of blocking on
3342 an empty osc while others are not empty. If we must block, we
3343 block for the shortest possible period of time.
3347 Description: improve handling recoverable errors
3348 Details : if request processig with error which can be recoverable on server
3349 request should be resend, otherwise page released from cache and
3352 Severity : enhancement
3354 Description: refine locking for avoid write wrong info into lov_objid file
3355 Details : fix possible races with add new target and write/update data in
3358 --------------------------------------------------------------------------------
3360 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3365 Frequency : on some architectures
3367 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3368 Details : Uninitialized flags sometimes cause configuration commands to
3372 Frequency : patchless clients only
3374 Description: missing __iget() symbol export
3375 Details : The __iget() symbol export is missing. To avoid the need for
3376 this on patchless clients the deathrow inode reaper is turned
3377 off, and we depend on the VM to clean up old inodes. This
3378 dependency was during via the fix for bug 12181.
3380 --------------------------------------------------------------------------------
3382 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3384 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3385 INTEROPERATE with older versions automatically. In many cases a
3386 special upgrade step is needed. Please read the
3387 user documentation before upgrading any part of a live system.
3388 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3389 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3390 * WARNING: Lustre configuration and startup changes are required with
3391 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3395 Severity : enhancement
3397 Description: MountConf
3398 Details : Lustre configuration is now managed via mkfs and mount
3399 commands instead of lmc and lconf. New obd types (MGS, MGC)
3400 are added for dynamic configuration management. See
3401 https://mail.clusterfs.com/wikis/lustre/MountConf for
3404 Severity : enhancement
3406 Description: dynamic OST addition
3407 Details : OSTs can now be added to a live filesystem
3409 Severity : enhancement
3411 Description: startup order invariance
3412 Details : MDTs and OSTs can be started in any order. Clients only
3413 require the MDT to complete startup.
3415 Severity : enhancement
3417 Description: parallel, asynchronous orphan cleanup
3418 Details : orphan cleanup is now performed in separate threads for each
3419 OST, allowing parallel non-blocking operation.
3421 Severity : enhancement
3423 Description: optimized stripe assignment
3424 Details : stripe assignments are now made based on ost space available,
3425 ost previous usage, and OSS previous usage, in order to try
3426 to optimize storage space and networking resources.
3428 Severity : enhancement
3430 Description: Permanently set tunables
3431 Details : All writable /proc/fs/lustre tunables can now be permanently
3432 set on a per-server basis, at mkfs time or on a live system.
3434 Severity : enhancement
3436 Description: Lustre message v2
3437 Details : Add lustre message format v2.
3439 Severity : enhancement
3441 Description: client OST exclusion list
3442 Details : Clients can be started with a list of OSTs that should be
3443 declared "inactive" for known non-responsive OSTs.
3447 Description: SPEC SFS validation failure on NFS v2 over lustre.
3448 Details : Changes the blocksize for regular files to be 2x RPC size,
3449 and not depend on stripe size.
3451 Severity : enhancement
3453 Description: Multiple MD RPCs in flight.
3454 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3455 To avoid overly-overloading MDS, introduce a limit on number of
3456 MDS RPCs in flight for a single client and add /proc controls
3457 to adjust this limit.
3459 Severity : enhancement
3461 Description: client read/write statistics
3462 Details : Add client read/write call usage stats for performance
3463 analysis of user processes.
3464 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3465 file access. extents_stats shows chunk size distribution.
3466 extents_stats_per_process show chunk size distribution per
3469 Severity : enhancement
3471 Description: mds statistics
3472 Details : Add detailed mds operations statistics in
3473 /proc/fs/lustre/mds/*/stats.
3477 Description: Failure of copying files with lustre special EAs.
3478 Details : Client side always return success for setxattr call for lustre
3479 special xattr (currently only "trusted.lov").
3484 Description: Refcount LNET uuids
3485 Details : The global LNET uuid list grew linearly with every startup;
3486 refcount repeated list entries instead of always adding to
3489 Severity : enhancement
3491 Description: Dynamic service threads
3492 Details : Within a small range, start extra service threads
3493 automatically when the request queue builds up.
3496 Frequency : mixed-endian client/server environments
3498 Description: mixed-endian crashes
3499 Details : The new msg_v2 system had some failures in mixed-endian
3502 Severity : enhancement
3504 Description: Easy OST removal
3505 Details : OSTs can be permanently deactivated with e.g. 'lctl
3506 conf_param lustre-OST0001.osc.active=0'
3508 Severity : enhancement
3510 Description: MGS proc entries
3511 Details : Added basic proc entries for the MGS showing what filesystems
3514 Severity : enhancement
3516 Description: provide MGS failover
3517 Details : Added config lock reacquisition after MGS server failover.
3519 Severity : enhancement
3521 Description: add Linux 2.4 support
3522 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3526 Description: a large application tries to do I/O to the same resource and dies
3527 in the middle of it.
3528 Details : Check the req->rq_arrival time after the call to
3529 ost_brw_lock_get(), but before we do anything about
3530 processing it & sending the BULK transfer request. This
3531 should help move old stale pending locks off the queue as
3532 quickly as obd_timeout.
3535 Frequency : when an incorrect nid is specified during startup
3537 Description: ptlrpc connect to non-existant node causes kernel crash
3538 Details : LNET can't be re-entered from an event callback, which
3539 happened when we expire a message after the export has been
3540 cleaned up. Instead, hand the zombie cleanup off to another
3543 Severity : enhancement
3545 Description: plain/inodebits lock performance improvement
3546 Details : Grouping plain/inodebits in granted list by their request modes
3547 and bits policy, thus improving the performance of search through
3551 Frequency : only if OST filesystem is corrupted
3553 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3554 Details : for a short time RPCs with bulk IO are in the replay list,
3555 but replay of bulk IOs is unimplemented. If the OST filesystem
3556 is corrupted due to disk cache incoherency and then replay is
3557 started it is possible to trip an assertion. Avoid putting
3558 committed RPCs into the replay list at all to avoid this issue.
3561 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3563 Description: System hang on startup
3564 Details : This bug allowed the liblustre (e.g. catamount) client to
3565 return to the app before handling all startup RPCs. This
3566 could leave the node unresponsive to lustre network traffic
3567 and manifested as a server ptllnd timeout.
3569 Severity : enhancement
3571 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3572 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3573 boolean to control whether to print peer debug info when a
3574 client's RPC times out.
3577 Frequency : only for kernels with patches from Lustre below 1.4.3
3579 Description: Remove old rdonly API
3580 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3583 Frequency : only for devices with external journals
3585 Description: Set external device read-only also
3586 Details : During a commanded failover stop, we set the disk device
3587 read-only while the server shuts down. We now also set any
3588 external journal device read-only at the same time.
3591 Frequency : when upgrading from 1.4 while trying to change parameters
3593 Description: The wrong (new) MDC name was used when setting parameters for
3594 upgraded MDT's. Also allows changing of OSC (and MDC)
3595 parameters if --writeconf is specified at tunefs upgrade time.
3598 Frequency : when setting specific ost indicies
3600 Description: QOS code breaks on skipped indicies
3601 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3602 created with --index need not be sequential.
3604 Severity : enhancement
3606 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3607 Details : The uninit_groups feature works in conjunction with the kernel
3608 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3609 up the pass1 processing of e2fsck. This is a read-only feature
3610 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3611 mount filesystems that have had this feature enabled.
3613 Severity : enhancement
3615 Description: Improve multi-block allocation algorithm to avoid fragmentation
3616 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3617 allocation locality and avoid filesystem fragmentation.
3619 ------------------------------------------------------------------------------
3621 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3623 * Support for kernels:
3624 2.4.21-47.0.1.EL (RHEL 3)
3625 2.6.5-7.283 (SLES 9)
3626 2.6.9-42.0.10.EL (RHEL 4)
3627 2.6.12.6 vanilla (kernel.org)
3628 2.6.16.27-0.9 (SLES 10)
3629 * Recommended e2fsprogs version: 1.39.cfs5
3631 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3635 Frequency : occasional, depends on client load and configuration
3636 Bugzilla : 12181, 12203
3637 Description: data loss for recently-modified files
3639 Details : In some cases it is possible that recently written or created
3640 files may not be written to disk in a timely manner (this should
3641 normally be within 30s unless client IO load is very high).
3642 The problem appears as zero-length files or files that are a
3643 multiple of 1MB in size after a client crash or client eviction
3644 that are missing data at the end of the file.
3646 This problem is more likely to be hit on clients where files are
3647 repeatedly created and unlinked in the same directory, clients
3648 have a large amount of RAM, have many CPUs, the filesystem has
3649 many OSTs, the clients are rebooted frequently, and/or the files
3650 are not accessed by other nodes after being written.
3652 The presence of the problem can be detected by looking at
3653 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3654 smaller than the second (nr_unused) then dirty files will not
3655 be flushed automatically to disk. "sync; sleep 10" should be
3656 run several times on the node before unmounting it to update
3657 Lustre (this is also safe to run on nodes without this problem).
3659 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3660 that can cause this same problem, so customers using that kernel
3661 also need to update the kernel in addition to Lustre. In order
3662 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3664 It is normal that files written just before a client crash (less
3665 than 30s) may not yet have been flushed to disk, even for local
3669 Frequency : frequent on thin XT3 nodes
3671 Description: UUID collision on thin XT3 Linux nodes
3672 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3673 randomly, since we relied on an insufficiently-seeded PRNG.
3678 Description: OSS hangs after "All ost request buffers busy"
3679 Details : A deadlock between quota and journal operations caused OSS
3680 hangs after printing "All ost request buffers busy."
3683 Frequency : always on liblustre builds
3685 Description: Cleanup compiler warnings on liblustre
3688 Frequency : always on liblustre builds on XT3
3690 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3694 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3695 > 32000 subdirectories
3696 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3697 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3698 subdirectory count crosses 32000. This will aid e2fsck to
3699 correctly handle more than 32000 subdirectories.
3702 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3704 Description: System hang on startup
3705 Details : This bug allowed the liblustre (e.g. catamount) client to
3706 return to the app before handling all startup RPCs. This
3707 could leave the node unresponsive to lustre network traffic
3708 and manifested as a server ptllnd timeout.
3710 Severity : enhancement
3712 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3713 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3714 boolean to control whether to print peer debug info when a
3715 client's RPC times out.
3720 Description: make O_SYNC working on 2.6 kernels
3721 Details : 2.6 kernels use different method for mark pages for write,
3722 so need add a code to lustre for O_SYNC work.
3727 Description: Failure to close file and release space on NFS
3728 Details : Put inode details into lock acquired in ll_intent_file_open.
3729 Use mdc_intent_lock in ll_intent_open to properly
3730 detect all kind of errors unhandled by mdc_enqueue.
3735 Description: proc file read during shutdown sometimes raced obd removal,
3737 Details : Add lock to prevent obd access after proc file removal.
3740 Frequency : Only for files larger than 4GB on 32-bit clients.
3742 Description: improperly doing page alignment of locks
3743 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3744 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3749 Description: RPCs being resent when they shouldn't be
3750 Details : Some RPCs that should not be resent are being resent. This
3751 can cause inconsistencies in the RPC state machine. Do not
3752 resend such requests.
3755 Frequency : rare, only with NFS export
3757 Description: Crash on NFS re-export node
3758 Details : under very unusual load conditions an assertion is hit in
3759 ll_intent_file_open()
3762 Frequency : only if OST filesystem is corrupted
3764 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3765 Details : for a short time RPCs with bulk IO are in the replay list,
3766 but replay of bulk IOs is unimplemented. If the OST filesystem
3767 is corrupted due to disk cache incoherency and then replay is
3768 started it is possible to trip an assertion. Avoid putting
3769 committed RPCs into the replay list at all to avoid this issue.
3774 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3775 Details : Large single O_DIRECT read and write calls can fail to allocate
3776 a sufficiently large buffer to process the request. In case of
3777 allocation failure the allocation is retried with a smaller
3778 buffer and broken into smaller requests.
3780 Severity : enhancement
3782 Description: Add -o localflock option to simulate old noflock behaviour.
3783 Details : This will achieve local-only flock/fcntl locks coherentness.
3788 Description: versioning check is incomplete
3789 Details : Checking the version difference of client vs. server, report
3790 error if the gap is too big.
3795 Description: add support PG_writeback bit
3796 Details : add support for PG_writeback bit for Lustre, for more carefull
3797 work with page cache in 2.6 kernel. This also fix some deadlocks
3798 and remove hack for work O_SYNC with 2.6 kernel.
3800 Severity : enhancement
3802 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3803 Details : The uninit_groups feature works in conjunction with the kernel
3804 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3805 up the pass1 processing of e2fsck. This is a read-only feature
3806 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3807 mount filesystems that have had this feature enabled.
3809 Severity : enhancement
3811 Description: Improve multi-block allocation algorithm to avoid fragmentation
3812 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3813 allocation locality and avoid filesystem fragmentation.
3815 ------------------------------------------------------------------------------
3817 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3819 * Support for kernels:
3820 2.6.9-42.0.3.EL (RHEL 4)
3821 2.6.5-7.276 (SLES 9)
3822 2.4.21-47.0.1.EL (RHEL 3)
3823 2.6.12.6 vanilla (kernel.org)
3824 2.6.16.21-0.8 (SLES10)
3825 * Recommended e2fsprogs version: 1.39.cfs2-0
3827 * The backwards-compatible /proc/sys/portals symlink has been removed
3828 in this release. Before upgrading, please ensure that you change
3829 any configuration scripts or /etc/sysctl.conf files that access
3830 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3831 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3832 in advance of the upgrade on any system running Lustre 1.4.6 or
3833 newer, since /proc/sys/lnet was added in that version.
3834 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3838 Frequency : only when quota is used
3840 Description: avoid scanning export list for quota master
3841 Details : Change the algorithms to avoid scanning export list in order
3842 to improve the efficiency.
3845 Frequency : MDS failover only, very rarely
3847 Description: "went back in time" messages on mds failover
3848 Details : The greatest transno may be lost when the current operation
3849 finishes with an error (transno==0) and the client's last_rcvd
3850 record is over-written. Save the greatest transno in the
3851 mds_last_transno for this case.
3854 Frequency : always for specific kernels and striping counts
3856 Description: client may get "Matching packet too big" without ACL support
3857 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3858 when trying to access files in certain configurations. The
3859 clients should in fact be denied when mounting because they do
3860 not understand ACLs.
3863 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3865 Description: many clients connecting with IO in progress causes connect timeouts
3866 Details : Avoid synchronous journal commits to avoid delays caused by many
3867 clients connecting/disconnecting when bulk IO is in progress.
3868 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3869 OST_IO_PORTAL to avoid delays behind potentially many pending
3873 Frequency : occasionally with multiple writers to a single file
3875 Description: shared writes to file may result in wrong size reported by stat()
3876 Details : Allow growing of kms when extent lock is cancelled
3879 Frequency : always with random mmap IO to multi-striped file
3881 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3882 Details : Only if the hole is at the end of OST object so that kms is too
3883 small. Fix is to increase kms accordingly in ll_nopage.
3886 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3888 Description: writes to a missing object would leak memory on the OST
3889 Details : If there is an inconsistency between the MDS and OST filesystems,
3890 such that the MDS references an object that doesn't exist, writes
3891 to that object will leak memory due to incorrect cleanup in the
3892 error handling path, eventually running out of memory on the OST.
3897 Description: Creating too long symlink causes lustre errors
3898 Details : Check symlink and name lengths before sending requests to MDS.
3901 Frequency : only if flock is enabled (not on by default)
3903 Description: posix locks not released on fd closure on 2.6.9+
3904 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3905 kernels, this caused such locks not to be released on fd close and
3906 then assertions on fs unmount about still used locks.
3909 Frequency : MDS failover only, very rarely
3911 Description: clients may get ASSERTION(granted_lock != NULL)
3912 Details : When request was taking a long time, and a client was resending
3913 a getattr by name lock request. The were multiple lock requests
3914 with the same client lock handle and
3915 mds_getattr_name->fixup_handle_for_resent_request found one of the
3916 lock handles but later failed with ASSERTION(granted_lock != NULL).
3921 Description: handle->h_buffer_credits > 0, assertion failure
3922 Details : h_buffer_credits is zero after truncate, causing assertion
3923 failure. This patch extends the transaction or creates a new
3927 Frequency : NFS re-export or patchless client
3928 Bugzilla : 11179, 10796
3929 Description: Crash on NFS re-export node (__d_move)
3930 Details : We do not want to hash the dentry if we don't have a lock.
3931 But if this dentry is later used in d_move, we'd hit uninitialised
3932 list head d_hash, so we just do this to init d_hash field but
3933 leave dentry unhashed.
3936 Frequency : NFS re-export or patchless client
3938 Description: NFS exports has problem with symbolic link
3939 Details : lustre client didn't properly install dentry when re-exported
3940 to NFS or running patchless client.
3943 Frequency : NFS re-export or patchless client
3945 Description: Various nfs/patchless fixes.
3946 Details : fixes reuse disconected alias for lookup process - this fixes
3947 warning "find_exported_dentry: npd != pd",
3948 fix permission error with open files at nfs.
3949 fix apply umask when do revalidate.
3952 Frequency : occasional
3954 Description: Crash on NFS re-export node
3955 Details : calling clear_page() on the wrong pointer triggered oops in
3956 generic_mapping_read().
3959 Frequency : rarely, using O_DIRECT IO
3961 Description: unaligned directio crashes client with LASSERT
3962 Details : check for unaligned buffers before trying any requests.
3965 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3967 Description: stale data returned from RAID cache
3968 Details : If only a small amount of IO is done to the RAID device before
3969 reading it again it is possible to get stale data from the RAID
3970 cache instead of reading it from disk.
3973 Frequency : always for sles10 kernel
3975 Description: sles10 support
3976 Details : ll_follow_link: compile fixes and using of nd_set_link
3977 under newer kernels.
3980 Frequency : depends on arch, kernel and compiler version, always on sles10
3983 Description: recursive or deep enough symlinks cause stack overflow
3984 Details : getting rid of large stack-allocated variable in
3988 Frequency : depends on hardware
3990 Description: lustre write performance loss in the SLES10 kernel
3991 Details : the performance loss is caused by using of write barriers in the
3992 ext3 code. The SLES10 kernel turns barrier support on by
3993 default. The fix is to undo that change for ldiskfs.
3995 ------------------------------------------------------------------------------
3997 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3999 * Support for kernels:
4000 2.6.9-42.0.3EL (RHEL 4)
4001 2.6.5-7.276 (SLES 9)
4002 2.4.21-47.0.1.EL (RHEL 3)
4003 2.6.12.6 vanilla (kernel.org)
4007 Frequency : quota enabled and large files being deleted
4009 Description: releasing more than 4GB of quota at once hangs OST
4010 Details : If a user deletes more than 4GB of files on a single OST it
4011 will cause the OST to spin in an infinite loop. Release
4012 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4017 Description: statfs data retrieved from /proc may be stale or zero
4018 Details : When reading per-device statfs data from /proc, in the
4019 {kbytes,files}_{total,free,avail} files, it may appear
4020 as zero or be out of date.
4023 Frequency : systems with MD RAID1 external journal devices
4025 Description: lconf's call to blkid is confused by RAID1 journal devices
4026 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4027 of returning all block devices that match the journal UUID.
4030 Frequency : always, for aggregate stripe size over 4GB
4032 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4033 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4034 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4037 Frequency : always if "lfs find" used on a local file/directory
4039 Description: "lfs find" segfaults if used on a local file/directory
4040 Details : The case where a directory component was not specified wasn't
4041 handled correctly. Handle this properly.
4044 Frequency : always on ppc64
4046 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4047 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4048 perform endianness conversion.
4051 Frequency : rarely (truncate to non-zero file size after write under load)
4052 Bugzilla : 10730, 10687
4053 Description: Files padded with zeros to next 4K multiple
4054 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4055 it is possible that files that are truncated to a non-zero size
4056 immediately after being written are filled with zero bytes beyond
4057 the truncated size. No file data is lost.
4059 Severity : enhancement
4061 Description: Allow recovery/failover for liblustre clients.
4062 Details : liblustre clients were unaware of failover configurations until
4065 Severity : enhancement
4067 Description: user file locks should fail when not mounting with flock option
4068 Details : Set up an error-returning stub in ll_file_operations.lock field
4069 to prevent incorrect behaviour when client is mounted without
4070 flock option. Also, set up properly f_op->flock field for
4074 Frequency : always on ia64
4076 Description: "lfs df" loops on printing out MDS statfs information
4077 Details : The obd_ioctl_data was not initialized and in some systems
4078 this caused a failure during the ioctl that did not return
4079 an error. Initialize the struct and return an error on failure.
4082 Frequency : SLES 9 only
4084 Description: Error of copying files with lustre special EAs as root
4085 Details : Client side always return success for setxattr call for lustre
4086 special xattr (currently only "trusted.lov").
4089 Frequency : rarely on clusters with both ia64+i386 clients
4091 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4092 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4093 clients are doing concurrent writes to the same file it is
4094 possible that the ia64 clients may LASSERT because the OST
4095 extent locks are not PAGE_SIZE aligned. Ensure that grown
4096 locks are always aligned on the request boundary.
4099 Frequency : specific use, occasional
4101 Description: Overwriting in use executable truncates on-disk binary image
4102 Details : If one node attempts to overwrite an executable in use by
4103 another node, we now correctly return ETXTBSY instead of
4104 truncating the file.
4106 Severity : enhancement
4108 Description: Async OSC create to avoid the blocking unnecessarily.
4109 Details : If an OST has no remaining object, system will block on the creating
4110 when need to create a new object on this OST. Now, ways use
4111 pre-created objects when available, instead of blocking on an
4112 empty osc while others are not empty. If we must block, we block
4113 for the shortest possible period of time.
4118 Description: chmod on Lustre root is propagated to other clients
4119 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4120 invalid by the follow_mount time.
4123 Frequency : liblustre clients only
4125 Description: Race in 'instant cancel' lock handling could lead to such locks
4126 never to be granted in case of SMP MDS
4127 Details : Do not destroy not yet granted but cbpending locks in
4131 Frequency : replay/resend of open
4133 Description: non null lock assetion failure in mds_intent_policy
4134 Details : Trying to replay/resend lockless open requests resulted in
4135 mds_open() returning 0 with no lock. Now it sets a flag if
4136 a lock is going to be returned.
4138 Severity : enhancement
4140 Description: Checksum enhancements
4141 Details : New checksum enhancements allow for resending RPCs that failed
4144 Severity : enhancement
4146 Description: Tunables on number of dirty pages in cacche
4147 Details : Allow to set limit on number of dirty pages cached.
4152 Description: client crash on unmount - lock still has references
4153 Details : In some error handling cases it was possible to leak a lock
4154 reference on a client while accessing a file. This was not
4155 harmful to the client during operation, but would cause the
4156 client to crash when the filesystem is unmounted.
4159 Frequency : specific case, rare
4161 Description: ETXTBSY on mds though file not in use by client
4162 Details : ETXTBSY is no longer incorrectly returned when attempting to
4163 chmod or chown a directory that the user previously tried to
4164 execute or a currently-executing binary.
4167 Frequency : extremely rare except on liblustre-based clients
4169 Description: Lustre space not freed when files are deleted
4170 Details : Clean up open-unlinked files after client eviction. Previously
4171 the unlink was skipped and the files remained as orphans.
4176 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4177 Details : In some cases it was possible to send a blocking callback to
4178 a client doing a glimpse, even though that client didn't get
4179 a lock granted. When the glimpse lock is cancelled on the OST
4180 the freed lock is left on the waiting list and corrupted the list.
4183 Frequency : all core dumps
4185 Description: Broke core dumps to lustre
4186 Details : Negative dentry may be unhashed if parent does not have UPDATE
4187 lock, but some callers, e.g. do_coredump, expect dentry to be
4188 hashed after successful create, hash it in ll_create_it.
4190 ------------------------------------------------------------------------------
4192 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4194 * Support for kernels:
4195 2.6.9-42.0.2.EL (RHEL 4)
4196 2.6.5-7.276 (SLES 9)
4197 2.4.21-40.EL (RHEL 3)
4198 2.6.12.6 vanilla (kernel.org)
4202 Frequency : always on RHEL 3
4204 Description: Number of open files grows over time
4205 Details : The number of open files grows over time, whether or not
4206 Lustre is started. This was due to a filp leak introduced
4207 by one of our kernel patches.
4209 ------------------------------------------------------------------------------
4211 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4213 * Support for kernels:
4214 2.6.9-42.EL (RHEL 4)
4215 2.6.5-7.276 (SLES 9)
4216 2.4.21-40.EL (RHEL 3)
4217 2.6.12.6 vanilla (kernel.org)
4222 Bugzilla : 5719, 9635, 9792, 9684
4223 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4224 Details : If a server is under heavy load and cannot reply to new
4225 connection requests before the client resends the (re)connect,
4226 the connection handling code can behave badly if two service
4227 threads are concurrently handing separate (re)connections from
4228 the same client. Add better locking to the connection handling
4229 code, and ensure that only a single connection will be processed
4230 for a given client UUID, even if the lock is dropped.
4232 Severity : enhancement
4234 Description: add TCP zero-copy support to kernel
4235 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4236 sends if the hardware supports scatter-gather and checksumming.
4237 This allows socklnd to do client-write and server-read more
4238 efficiently and reduce CPU utilization from skbuf copying.
4241 Frequency : only if NFS exporting from client
4243 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4244 Details : If setting attributes on a file created under NFS that had
4245 never been opened it would be possible to oops the client
4246 if the file had no objects.
4249 Frequency : always for liblustre
4251 Description: liblustre client does MDS+OSTs setattr RPC for each write
4252 Details : When doing a write from a liblustre client, the client
4253 incorrectly issued an RPC to the MDS and each OST the file was
4254 striped over in order to update the timestamps. When writing
4255 with small chunks and many clients this could overwhelm the MDS
4256 with RPCs. In all cases it would slow down the write because
4257 these RPCs are unnecessary.
4259 Severity : enhancement
4261 Description: allow number of MDS service threads to be changed at module load
4262 Details : It is now possible to change the number of MDS service threads
4263 running. Adding "options mds mds_num_threads={N}" to the MDS's
4264 /etc/modprobe.conf will set the number of threads for the next
4265 time Lustre is restarted (assuming the "mds" module is also
4266 reloaded at that time). The default number of threads will
4267 stay the same, 32 for most systems.
4272 Description: OST crash if filesystem is unformatted or corrupt
4273 Details : If an OST is started on a device that has never been formatted
4274 or if the filesystem is corrupt and cannot even mount then the
4275 error handling cleanup routines would dereference a NULL pointer.
4280 Description: NULL pointer deref in llap_from_page.
4281 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4282 incorrect) mapping if the page was truncated/reclaimed while it was
4283 searched for. Check for this condition and skip such pages when
4284 doing readahead. Introduce extra check to llap_from_page() to
4285 verify page->mapping->host is non-NULL (so page is not anonymous).
4288 Frequency : Sometimes when using sys_sendfile
4290 Description: "page not covered by a lock" warnings from ll_readpage
4291 Details : sendfile called ll_readpage without right page locks present.
4292 Now we introduced ll_file_sendfile that does necessary locking
4293 around call to generic_file_sendfile() much like we do in
4297 Frequency : with certain MDS communication failures at client mount time
4299 Description: NULL pointer deref after failed client mount
4300 Details : a client connection request may delayed by the network layer
4301 and not be sent until after the PTLRPC layer has timed out the
4302 request. If the client fails the mount immediately it will try
4303 to clean up before the network times out the request. Add a
4304 reference from the request import to the obd device and delay
4305 the cleanup until the network drops the request.
4308 Frequency : occasionally during client (re)connect
4310 Description: assertion failure during client (re)connect
4311 Details : processing a client connection request may be delayed by the
4312 client or server longer than the client connect timeout. This
4313 causes the client to resend the connection request. If the
4314 original connection request is replied in this interval, the
4315 client may trip an assertion failure in ptlrpc_connect_interpret()
4316 which thought it would be the only running connect process.
4319 Frequency : only with obd_echo servers and clients that are rebooted
4321 Description: kernel BUG accessing uninitialized data structure
4322 Details : When running an obd_echo server it did not start the ping_evictor
4323 thread, and when a client was evicted an uninitialized data
4324 structure was accessed. Start the ping_evictor in the RPC
4325 service startup instead of the OBD startup.
4327 Severity : enhancement
4328 Bugzilla : 10193 (patchless)
4329 Description: Remove dependency on various unexported kernel interfaces.
4330 Details : No longer need reparent_to_init, exit_mm, exit_files,
4331 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4334 Frequency : rare (only users of deprecated and unsupported LDAP config)
4336 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4337 Details : LDAP apparently contains 'lustreName' attributes instead of
4338 'name'. A simple remapping of the name is sufficient.
4341 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4343 Description: waiting_locks_callback trips kernel BUG if client is evicted
4344 Details : Running with the dump_on_timeout debug flag turned on makes
4345 it possible that the waiting_locks_callback() can try to dump
4346 the Lustre kernel debug logs from an interrupt handler. Defer
4347 this log dumping to the expired_lock_main() thread.
4349 Severity : enhancement
4351 Description: Support NFS exporting on 2.6 kernels.
4352 Details : Implement non-rawops metadata methods for NFS server to use without
4353 changing NFS server code.
4356 Frequency : very rare (synthetic metadata workload only)
4358 Description: two racing renames might cause an MDS thread to deadlock
4359 Details : Running the "racer" program may cause one MDS thread to rename
4360 a file from being the source of a rename to being the target of
4361 a rename at exactly the same time that another thread is doing
4362 so, and the second thread has already enqueued these locks after
4363 doing a lookup of the target and is trying to relock them in
4364 order. Ensure that we don't try to re-lock the same resource.
4367 Frequency : only very large systems with liblustre clients
4369 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4370 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4371 all OSTs in parallel. This allows the eviction of stale liblustre
4372 clients to proceed much faster than if they were done in series,
4373 and also offers similar improvements for other set_info RPCs.
4378 Description: excessive CPU usage during initial read phase on client
4379 Details : During the initial read phase on a client, it would agressively
4380 retry readahead on the file, consuming too much CPU and impacting
4381 performance (since 1.4.5.8). Improve the readahead algorithm
4382 to avoid this, and also improve some other common cases (read
4383 of small files in particular, where "small" is files smaller than
4384 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4389 Description: MDS crash when receiving packet with unknown intent.
4390 Details : Do not LBUG in unknown intent case, just return -EFAULT
4392 Severity : enhancement
4393 Bugzilla : 9293, 9385
4394 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4395 Details : Do not serialize getattr (non-intent version) and statfs.
4398 Frequency : occasional, when OST network is overloaded/intermittent
4400 Description: client evicted by OST after bulk IO timeout
4401 Details : If a client sends a bulk IO request (read or write) the OST
4402 may evict the client if it is unresposive to its data GET/PUT
4403 request. This is incorrect if the network is overloaded (takes
4404 too long to transfer the RPC data) or dropped the OST GET/PUT
4405 request. There is no need to evict the client at all, since
4406 the pinger and/or lock callbacks will handle this, and the
4407 client can restart the bulk request.
4410 Frequency : Always when mmapping file with no objects
4412 Description: client crashes when mmapping file with no objects
4413 Details : Check that we actually have objects in a file before doing any
4414 operations on objects in ll_vm_open, ll_vm_close and
4420 Description: Request leak when working with deleted CWD
4421 Details : Introduce advanced request refcount tracking for requests
4422 referenced from lustre intent.
4424 Severity : Enhancement
4426 Description: Cache open file handles on client.
4427 Details : MDS now will return special lock along with openhandle, if
4428 requested and client is allowed to hold openhandle, even if unused,
4429 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4430 closing files for every read/write openration.
4432 Severity : Enhancement
4434 Description: Cache open negative dentries on client when possible.
4435 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4436 negative dentries on lock revocation.
4441 Description: Remounting a client read-only wasn't possible with a zconf mount
4442 Details : It wasn't possible to remount a client read-only with llmount.
4444 Severity : enhancement
4445 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4446 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4447 MPICH is included to add Lustre-specific ADIO interfaces.
4448 This is based closely on the UFS ADIO layer and only differs
4449 in file creation, in order to allow the OST striping to be set.
4450 This is user-contributed code and not supported by CFS.
4455 Description: extended inode attributes (immutable, append-only) work improperly
4456 when 2.4 and 2.6 kernels are used on client/server or vice versa
4457 Details : Introduce kernel-independent values for these flags.
4459 Severity : enhancement
4462 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4463 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4464 granularity. Now, Fractional MB granularity is be supported,
4465 this is very useful for low memory system.
4467 Severity : enhancement
4469 Description: Getattr by fid
4470 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4471 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4472 have locks and avoids getattr rpc requests for it on every stat.
4475 Frequency : Always, for filesystems larger than 2TB
4477 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4478 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4479 filesystems larger than 2TB to be mounted with mballoc enabled.
4482 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4484 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4485 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4486 filesystems larger than 2TB is mounted then the kernel will
4487 silently corrupt the start of the filesystem. CONFIG_LBD is
4488 enabled for all CFS-supported kernels, but the possibility of
4489 this happening with a modified kernel config exists.
4491 Severity : enhancement
4493 Description: add client O_DIRECT support for 2.6 kernels
4494 Details : It is now possible to do O_DIRECT reads and writes to files
4495 in the Lustre client mountpoint on 2.6 kernel clients.
4497 Severity : enhancement
4499 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4500 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4501 parallel, not waiting for response from every OST before sending
4502 a rpc to the next OST.
4507 Description: setattr vs write race when updating file timestamps
4508 Details : Client processes that update a file timestamp into the past
4509 right after writing to the file (e.g. tar) it is possible that
4510 the updated file modification time can be reset to the current
4511 time due to a race between processing the setattr and write RPC.
4513 Severity : enhancement
4515 Description: Bring 'lfs find' closer in line with regular Linux find.
4516 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4517 -print0 options and obtains all the needed info through the lustre
4520 Severity : enhancement
4522 Description: support up to 1024 configured devices on one node
4523 Details : change obd_dev array from statically allocated to dynamically
4524 allocated structs as they are first used to reduce memory usage
4529 Description: Flush dirty partially truncated pages during truncate
4530 Details : Immediatelly flush partially truncated pages in filter_setattr,
4531 this way we completely avoid having any pages in page cache on OST
4532 and can retire ugly workarounds during writes to flush such pages.
4537 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4538 Details : protect inode from truncation within vfs_unlink() context
4539 just take a reference before calling vfs_unlink() and release it
4540 when parent's i_sem is free.
4545 Description: last_id value checked outside lock on OST caused LASSERT failure
4546 Details : If there were multiple MDS->OST object precreate requests in
4547 flight, it was possible that the OST's last object id was checked
4548 outside a lock and incorrectly tripped an assertion. Move checks
4549 inside locks, and discard old precreate requests.
4552 Frequency : always, if extents are used on OSTs
4554 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4555 Details : OSTs using the extents format would not zero the high 16 bits of
4556 the index physical block number. This is not a problem for any
4557 OST filesystems smaller than 16TB, and no kernels support ext3
4558 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4559 new/modified files) and can be fixed for existing filesystems
4560 with e2fsprogs-1.39-cfs1.
4565 Description: import connection selection may be incorrect if timer wraps
4566 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4567 import connections to be ignored if the 32-bit jiffies counter
4568 wraps. Use a 64-bit jiffies counter.
4571 Frequency : very large clusters immediately after boot
4573 Description: LNET request buffers exhausted under heavy short-term load
4574 Details : If a large number of client requests are generated on a service
4575 that has previously never seen so many requests it is possible
4576 that the request buffer growth cannot keep up with the spike in
4577 demand. Instead of dropping incoming requests, they are held in
4578 the LND until the RPC service can accept more requests.
4581 Frequency : Sometimes during replay
4583 Description: Assertion failure in ll_local_open after replay.
4584 Details : If replay happened on an open request reply before we were able
4585 to set replay handler, reply will become not swabbed tripping the
4586 assertion in ll_local_open. Now we set the handler right after
4587 recognising of open request
4590 Frequency : very rare
4592 Description: kernel reports "badness in vsnprintf"
4593 Details : Reading from the "recovery_status" /proc file in small chunks
4594 may cause a negative length in lprocfs_obd_rd_recovery_status()
4595 call to vsnprintf() (which is otherwise harmless). Exit early
4596 if there is no more space in the output buffer.
4598 Severity : enhancement
4600 Description: clear OBD RPC statistics by writing to them
4601 Details : It is now possible to clear the OBD RPC statistics by writing
4602 to the "stats" file.
4607 Description: Client mtime is not the same on different clients after utimes
4608 Details : In some cases, the client was using the utimes() syscall on
4609 a file cached on another node. The clients now validate the
4610 ctime from the MDS + OSTs to determine which one is right.
4615 Description: Inability to activate failout mode
4616 Details : lconf script incorrectly assumed that in pythong string's numeric
4617 value is used in comparisons.
4620 Frequency : always with multiple stripes per file
4622 Description: Inefficient object allocation for mutli-stripe files
4623 Details : When selecting which OSTs to stripe files over, for files with
4624 a stripe count that divides evenly into the number of OSTs,
4625 the MDS is always picking the same starting OST for each file.
4626 Return the OST selection heuristic to the original design.
4631 Description: mount failures may take full timeout to return an error
4632 Details : Under some heavy load conditions it is possible that a
4633 failed mount can wait for the full obd_timeout interval,
4634 possibly several minutes, before reporting an error.
4635 Instead return an error as soon as the status is known.
4637 Frequency : quota enabled and large files being deleted
4639 Description: releasing more than 4GB of quota at once hangs OST
4640 Details : If a user deletes more than 4GB of files on a single OST it
4641 will cause the OST to spin in an infinite loop. Release
4642 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4647 Description: statfs data retrieved from /proc may be stale or zero
4648 Details : When reading per-device statfs data from /proc, in the
4649 {kbytes,files}_{total,free,avail} files, it may appear
4650 as zero or be out of date.
4653 Frequency : systems with MD RAID1 external journal devices
4655 Description: lconf's call to blkid is confused by RAID1 journal devices
4656 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4657 of returning all block devices that match the journal UUID.
4660 Frequency : always, for aggregate stripe size over 4GB
4662 Description: assertion fails when trying to use 4GB stripe size
4663 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4664 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4667 Frequency : always on ppc64
4669 Description: the first write on an ext3 filesystem with mballoc got stuck
4670 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4671 perform endianness conversion.
4673 ------------------------------------------------------------------------------
4675 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4677 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4678 INTEROPERATE with older versions automatically. Please read the
4679 user documentation before upgrading any part of a live system.
4680 * WARNING: Lustre networking configuration changes are required with
4681 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4684 * Support for kernels:
4685 2.6.9-22.0.2.EL (RHEL 4)
4686 2.6.5-7.244 (SLES 9)
4687 2.6.12.6 vanilla (kernel.org)
4690 Severity : enhancement
4691 Bugzilla : 7981/8208
4692 Description: Introduced Lustre Networking (LNET)
4693 Details : LNET is new networking infrastructure for Lustre, it includes
4694 a reorganized network configuration mode (see the user
4695 documentation for full details) as well as support for routing
4696 between different network fabrics. Lustre Networking Devices
4697 (LNDs) for the supported network fabrics have also been
4698 created for this new infrastructure.
4700 Severity : enhancement
4701 Description: Introduced Access control lists
4702 Details : clients can set ACLs on files and directories in order to have
4703 more fine-grained permissions than the standard Unix UGO+RWX.
4704 The MDS must be started with the "-o acl" mount option.
4706 Severity : enhancement
4707 Description: Introduced filesystem quotas
4708 Details : Administrators may now establish per-user quotas on the
4711 Severity : enhancement
4713 Description: Configuration change for the XT3
4714 The PTLLND is now used to run Lustre over Portals on the XT3
4715 The configure option(s) --with-cray-portals are no longer used.
4716 Rather --with-portals=<path-to-portals-includes> is used to
4717 enable building on the XT3. In addition to enable XT3 specific
4718 features the option --enable-cray-xt3 must be used.
4723 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4724 Details : A race between allocating a new llcd and re-getting the llcd_lock
4725 allowed another thread to grab newly-allocated llcd.
4727 Severity : enhancement
4729 Description: 2.6 OST async journal commit and locking fix to improve performance
4730 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4731 2.6 kernels are now async as they already were in 2.4 kernels so
4732 that they can commit concurrently with the network bulk transfer.
4733 For block-allocated files the filter allocation semaphore is held
4734 to avoid filesystem fragmentation during allocation. BKL lock
4735 removed for 2.6 xattr operations where it is no longer needed.
4740 Description: lconf incorrectly determined whether two IP networks could talk
4741 Details : In some more complicated routing and multiple-network
4742 configurations, lconf will avoid trying to make a network
4743 connection to a disjoint part of the IP space. It was doing the
4744 math incorrectly for one set of cases.
4749 Description: Fix for potential infinite loop processing records in an llog.
4750 Details : If an llog record is corrupted/zeroed, it is possible to loop
4751 forever in llog_process(). Validate the llog record length
4752 and skip the remainder of the block on error.
4755 Frequency : occasional (liblustre only)
4757 Description: liblustre could not open files whose last component is a symlink
4758 Details : sysio_path_walk() would incorrectly pass the open intent to
4759 intermediate path components.
4762 Frequency : rare (liblustre only with non-standard tuning)
4763 Bugzilla : 7201 (7350)
4764 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4765 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4766 after it has been released, only do this for referenced locks
4768 Severity : enhancement
4770 Description: specifying an (invalid) directory default stripe_size of -1
4771 would reset the directory default striping
4772 Details : stripe_size -1 was used internally to signal directory stripe
4773 removal, now use "all default" to signal dir stripe removal
4774 as a directory striping of "all default" is not useful
4777 Frequency : common for large clusters running liblustre clients
4779 Description: doing an ls when liblustre clients are running is slow
4780 Details : sending a glimpse AST to a liblustre client waits for every AST
4781 to time out, as liblustre clients will not respond. Since they
4782 cannot cache data we refresh the OST lock LVB from disk instead.
4784 Severity : enhancement
4786 Description: doing an ls at the same time as file IO can be slow
4787 Details : enqueue and other "small" requests can be blocked behind many
4788 large IO requests. Create a new OST IO portal for non-IO
4789 requests so they can be processed faster.
4792 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4794 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4795 created file with mode 000
4796 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4797 allow an ftruncate() as a non-root user to a file with mode 000.
4798 Lustre now respects this flag to disable mode checks when
4799 truncating a file owned by the user
4802 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4805 Description: Revoking locks from clients that went dead or catatonic might take
4807 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4808 cancellation of such locks instant on servers without waiting for
4809 any reply from clients. Clients drops these locks when cancel
4810 notification from server is received without replying.
4813 Frequency : liblustre-only, when liblustre client dies or becomes busy
4815 Description: Doing ls on Linux clients can take a long time with active
4817 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4818 granting such locks to it in the first place if possible. Locks
4819 are taken by proxy on the OST during the read or write and
4820 dropped immediately afterward. Add connect flags handling, do
4821 not grant locks to liblustre clients for glimpse ASTs.
4823 Severity : enhancement
4825 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4826 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4827 sequential reads and behaves badly with applications doing random
4828 reads. Improve it by reading ahead at least read region, and
4829 avoiding excessive large RPC for small reads.
4831 Severity : enhancement
4833 Description: Creating more than 1000 files for a single job may cause a load
4834 imbalance on the OSTs if there are also a large number of OSTs.
4835 Details : qos_prep_create() uses an OST index reseed value that is an
4836 even multiple of the number of available OSTs so that if the
4837 reseed happens in the middle of the object allocation it will
4838 still utilize the OSTs as uniformly as possible.
4843 Description: OST or MDS may oops in ping_evictor_main()
4844 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4845 but doesn't restart at beginning of obd_exports_timed list
4848 Severity : enhancement
4850 Description: improve by-nid export eviction on the MDS and OST
4851 Details : allow multiple exports with the same NID to be evicted at one
4852 time without re-searching the exports list.
4855 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4857 Description: MDS may oops in groups_free()
4858 Details : in rare race conditions a newly allocated group_info struct is
4859 freed again, and this can be NULL. The 2.4 compatibility code
4860 for groups_free() checked for a NULL pointer, but 2.6 did not.
4863 Frequency : common for liblustre clients doing little filesystem IO
4864 Bugzilla : 9352, 7313
4865 Description: server may evict liblustre clients accessing contended locks
4866 Details : if a client is granted a lock or receives a completion AST
4867 with a blocking AST already set it would not reply to the AST
4868 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4871 Frequency : lfs setstripe, only systems with more than 160 OSTs
4873 Description: unable to set striping with a starting offset beyond OST 160
4874 Details : llapi_create_file() incorrectly limited the starting stripe
4875 index to the maximum single-file stripe count.
4878 Frequency : LDAP users only
4880 Description: lconf did not handle in-kernel recovery with LDAP properly
4881 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4883 Severity : enhancement
4885 Description: bind OST threads to NUMA nodes to improve performance
4886 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4887 node and do their allocations there to localize memory access
4889 Severity : enhancement
4891 Description: llmount can determine client NID directly from Myrinet (GM)
4892 Details : the client NID code from gmnalnid was moved directly into
4893 llmount, removing the need to use this or specifying the
4894 client NID explicitly when mounting GM clients with zeroconf
4897 Frequency : if client is started with down MDS
4899 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4900 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4901 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4902 ptlrpc_import_delayed_req() because we didn't check if the
4903 request was interrupted, and we also didn't break out of the
4904 event loop if there was no timeout
4909 Description: data loss during non-page-aligned writes to a single file from
4910 both multiple nodes and multiple threads on one node at same time
4911 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4912 inconsistency led to false short-reads, that were cached and later
4913 used by ->prepare_write() to fill in partially written page,
4914 leading to data loss.
4917 Frequency : always, if lconf --abort_recovery used
4919 Description: lconf --abort_recovery fails with 'Operation not supported'
4920 Details : lconf was attempting to abort recovery on the MDT device and not
4923 Severity : enhancement
4925 Description: remove cleanup logs
4926 Details : replace lconf-generated cleanup logs with lustre internal
4927 cleanup routines. Eliminates the need for client-cleanup and
4930 Severity : enhancement
4932 Description: add support for EAs (user and system) on lustre filesystems
4933 Details : it is now possible to store extended attributes in the Lustre
4934 client filesystem, and with the user_xattr mount option it
4935 is possible to allow users to store EAs on their files also
4937 Severity : enhancement
4939 Description: Add possibility (config option) to show minimal available OST free
4941 Details : When compiled with --enable-mindf configure option, statfs(2)
4942 (and so, df) will return least minimal free space available from
4943 all OSTs as amount of free space on FS, instead of summary of
4944 free spaces of all OSTs.
4946 Severity : enhancement
4948 Description: do not expand extent locks acquired on OST-side
4949 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4950 by server: they are not cached anyway.
4953 Frequency : when mmap is used/binaries executed from Lustre
4955 Description: Unmmap pages before throwing them away from read cache.
4956 Details : llap_shrink cache now attempts to unmap pages before discarding
4957 them (if unmapping failed - do not discard). SLES9 kernel has
4958 extra checks that trigger if this unmapping is not done first.
4963 Description: lconf didn't resolve symlinks before checking to see whether a
4964 given mountpoint was already in use
4967 Frequency : when migrating failover services
4968 Bugzilla : 6395, 9514
4969 Description: When migrating a subset of services from a node (e.g. failback
4970 from a failover service node) the remaining services would
4971 time out and evict clients.
4972 Details : lconf --force (implied by --failover) sets the global obd_timeout
4973 to 5 seconds in order to quickly disconnect, but this caused
4974 other RPCs to time out too quickly. Do not change the global
4975 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4977 Severity : enhancement
4978 Frequency : if MDS is started with down OST
4979 Bugzilla : 9439,5706
4980 Description: Allow startup/shutdown of an MDS without depending on the
4981 availability of the OSTs.
4982 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4983 Add appropriate locking and lov-osc refcounts for safe
4984 cleaning. Add osc abort_inflight calls in case the
4985 synchronize never started.
4988 Frequency : occasional (Cray XT3 only)
4990 Description: root not authorized to access files in CRAY_PORTALS environment
4991 Details : The client process capabilities were not honoured on the MDS in
4992 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4993 been accessed by an authorized user then root was able to access
4994 the file on the local client also. The root user capabilities
4995 are now allowed on the MDS, as this environment has secure UID.
4998 Frequency : occasional
5000 Description: ldiskfs "too long searching" message happens too often
5001 Details : A debugging message (otherwise harmless) prints too often on
5002 the OST console. This has been reduced to only happen when
5003 there are fragmentation problems on the filesystem.
5008 Description: Division by zero in statfs when all OSCs are inactive
5009 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5010 lov_statfs divides by value returned by lov_get_stripecnt().
5014 Bugzilla : 9489, 3273
5015 Description: First write from each client to each OST was only 4kB in size,
5016 to initialize client writeback cache, which caused sub-optimal
5017 RPCs and poor layout on disk for the first written file.
5018 Details : Clients now request an initial cache grant at (re)connect time
5019 and so that they can start streaming writes to the cache right
5020 away and always do full-sized RPCs if there is enough data.
5021 If the OST is rebooted the client also re-establishes its grant
5022 so that client cached writes will be honoured under the grant.
5027 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5028 Details : Now I/O RPCs go to different portal number and (presumably) fast
5029 lock requests (and glimses) and other RPCs get their own service
5030 threads pool that should be able to service those RPCs
5033 Severity : enhancement
5035 Description: Ability to exchange lustre version between client and servers and
5036 issue warnings at client side if client is too old. Also for
5037 liblustre clients there is ability to refuse connection of too old
5039 Details : New 'version' field is added to connect data structure that is
5040 filled with version info. That info is later checked by server and
5044 Frequency : rare, liblustre only.
5045 Bugzilla : 9296, 9581
5046 Description: Two simultaneous writes from liblustre at offset within same page
5047 might proceed at the same time overwriting eachother with stale
5049 Details : I/O lock withing llu_file_prwv was released too early, before data
5050 actually was hitting the wire. Extended lock-holding time until
5051 server acknowledges receiving data.
5054 Frequency : extremely rare. Never observed in practice.
5056 Description: avoid generating lustre_handle cookie of 0.
5057 Details : class_handle_hash() generates handle cookies by incrementing
5058 global counter, and can hit 0 occasionaly (this is unlikely, but
5059 not impossible, because initial value of cookie counter is
5060 selected randonly). Value of 0 is used as a sentinel meaning
5061 "unassigned handle" --- avoid it. Also coalesce two critical
5062 sections in this function into one.
5064 Severity : enhancement
5066 Description: allow liblustre clients to delegate truncate locking to OST
5067 Details : To avoid overhead of locking, liblustre client instructs OST to
5068 take extent lock in ost_punch() on client's behalf. New connection
5069 flag is added to handle backward compatibility.
5071 Severity : enhancement
5072 Bugzilla : 4928, 7341, 9758
5073 Description: allow number of OST service threads to be specified
5074 Details : a module parameter allows the number of OST service threads
5075 to be specified via "options ost ost_num_threads={N}" in the
5076 OSS's /etc/modules.conf or /etc/modprobe.conf.
5080 Bugzilla : 6146, 9635, 9895
5081 Description: servers crash with bad pointer in target_handle_connect()
5082 Details : In rare cases when a client is reconnecting it was possible that
5083 the connection request was the last reference for that export.
5084 We would temporarily drop the export reference and get a new
5085 one, but this may have been the last reference and the export
5086 was just destroyed. Get new reference before dropping old one.
5088 Severity : enhancement
5089 Frequency : if client is started with failover MDS
5091 Description: Allow multiple MDS hostnames in the mount command
5092 Details : Try to read the configuration from all specified MDS
5093 hostnames during a client mount in case the "primary"
5096 Severity : enhancement
5098 Description: Stop sending data to evicted clients as soon as possible.
5099 Details : Check if the client we are about to send or are sending data to
5100 was evicted already. (Check is done every second of waiting,
5101 for which l_wait_event interface was extended to allow checking
5102 of exit condition at specified intervals).
5105 Frequency : rare, normally only when NFS exporting is done from client
5107 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5109 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5112 Frequency : always, if srand() is called before liblustre initialization
5114 Description: Liblustre uses system PRNG disturbing its usage by user application
5115 Details : Introduce internal to lustre fast and high-quality PRNG for
5116 lustre usage and make liblustre and some other places in generic
5117 lustre code to use it.
5119 Severity : enhancement
5120 Bugzilla : 9477, 9557, 9870
5121 Description: Verify that the MDS configuration logs are updated when xml is
5122 Details : Check if the .xml configuration logs are newer than the config
5123 logs stored on the MDS and report an error if this is the case.
5124 Request --write-conf, or allow starting with --old_conf.
5126 Severity : enhancement
5128 Description: Handle symlinks in the path when checking if Lustre is mounted.
5129 Details : Resolve intermediate symlinks when checking if a client has
5130 mounted a filesystem to avoid duplicate client mounts.
5135 Description: lconf can hit an error exception but still return success.
5136 Details : The lconf command catches the Command error exception at the top
5137 level script context and will exit with the associated exit
5138 status, but doesn't ensure that this exit status is non-zero.
5143 Description: failure of ptlrpc thread startup can cause oops
5144 Details : Starting a ptlrpc service thread can fail if there are a large
5145 number of threads or the server memory is very fragmented.
5146 Handle this without oopsing.
5149 Frequency : always, only if liblustre and non-default acceptor port was used
5151 Description: liblustre cannot connect to servers with non-default acceptor port
5152 Details : tcpnal_set_default_params() was not called and was therefore
5153 ignoring the environment varaible TCPNAL_PORT, as well as other
5154 TCPNAL_ environment variables
5159 Description: two objects could be created on the same OST for a single file
5160 Details : If an OST is down, in some cases it was possible to create two
5161 objects on a single OST for a single file. No problems other
5162 than potential performance impact and spurious error messages.
5166 Bugzilla : 5681, 9562
5167 Description: Client may oops in ll_unhash_aliases
5168 Details : Client dcache may become inconsistent in race condition.
5169 In some cases "getcwd" can fail if the current directory is
5175 Description: Inode refcounting problems in NFS export code
5176 Details : link_raw functions used to call d_instantiate without obtaining
5177 extra inode reference first.
5181 Bugzilla : 9942, 9903
5182 Description: Referencing freed requests leading to crash, memleaks with NFS.
5183 Details : We used to require that call to ll_revalidate_it was always
5184 followed by ll_lookup_it. Also with revalidate_special() it is
5185 possible to call ll_revalidate_it() twice for the same dentry
5186 even if first occurence returned success. This fix changes semantic
5187 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5188 reference on a request referred from the intent.
5189 ll_intent_release() then releases such a request.
5192 Frequency : rare, normally benchmark loads only
5194 Description: unlinked inodes were kept in memory on the client
5195 Details : If a client is repeatedly creating and unlinking files it
5196 can accumulate a lot of stale inodes in the inode slab cache.
5197 If there is no other client load running this can cause the
5198 client node to run out of memory. Instead flush old inodes
5199 from client cache that have the same inode number as a new inode.
5202 Frequency : SLES9 2.6.5 kernel and long filenames only
5203 Bugzilla : 9969, 10379
5204 Description: utime reports stale NFS file handle
5205 Details : SLES9 uses out-of-dentry names in some cases, which confused
5206 the lustre dentry revalidation. Change it to always use the
5210 Frequency : rare, unless heavy write-truncate concurrency is continuous
5211 Bugzilla : 4180, 6984, 7171, 9963, 9331
5212 Description: OST becomes very slow and/or deadlocked during object unlink
5213 Details : filter_destroy() was holding onto the parent directory lock
5214 while truncating+unlinking objects. For very large objects this
5215 may block other threads for a long time and slow overall OST
5216 responsiveness. It may also be possible to get a lock ordering
5217 deadlock in this case, or run out of journal credits because of
5218 the combined truncate+unlink. Solution is to do object truncate
5219 first in one transaction without parent lock, and then do the
5220 final unlink in a new transaction with the parent lock. This
5221 reduces the lock hold time dramatically.
5224 Frequency : rare, 2.4 kernels only
5226 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5227 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5228 before unmounting their filesystem in order to avoid deadlock.
5229 The kernel_locked() function in 2.4 kernels only checks whether
5230 the kernel lock is held, not whether it is this process that is
5231 holding it as 2.6 kernels do.
5236 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5237 Details : The client ptlrpc code may be trying to reconnect to a down
5238 server before a previous connection attempt has timed out.
5239 Increase the reconnect interval to be longer than the connection
5240 timeout interval to avoid sending duplicate connections to
5244 Frequency : echo_client brw_test command
5246 Description: fix echo_client to work with OST preallocated code
5247 Details : OST preallocation code (5137) didn't take echo_client IO path
5248 into account: echo_client calls filter methods outside of any
5249 OST thread and, hence, there is no per-thread preallocated
5250 pages and buffers to use. Solution: hijack pga pages for IO. As
5251 a byproduct, this avoids unnecessary data copying.
5255 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5256 Description: Client can oops in mdc_commit_close() after open replay
5257 Details : It was possible for the MDS to return an open request with no
5258 transaction number in mds_finish_transno() if the client was
5259 evicted, but without actually returning an error. Clients
5260 would later try to replay that open and may trip an assertion
5261 Simplify the client close codepath, and always return an error
5262 from the MDS in case the open is not successful.
5265 Frequency : rare, 2.6 OSTs only
5267 Description: OST may deadlock under high load on fragmented files
5268 Details : If there was a heavy load and highly-fragmented OST filesystems
5269 it was possible to have all the OST threads deadlock waiting on
5270 allocation of biovecs, because the biovecs were not released
5271 until the entire RPC IO was completed. Instead, release biovecs
5272 as soon as they are complete to ensure forward IO progress.
5274 Severity : enhancement
5276 Description: Support for specifying external journal device at mount
5277 Details : If an OST or MDS device is formatted with an external journal
5278 device, this device major/minor is stored in the ext3 superblock
5279 and may not be valid for failover. Allow detecting and
5280 specifying the external journal at mount time.
5285 Description: Mounting an MDS with pending unlinked files may cause oops
5286 Details : target_finish_recovery() calls mds_postrecov() which returned
5287 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5288 considers this an error and immediately begins cleaning up the
5289 lov, just after starting the mds_lov process
5291 Severity : enhancement
5293 Description: Implement 'lfs df' to report actual free space on per-OST basis
5294 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5295 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5296 usage of objects; '-h' to report in human readable format.
5298 ------------------------------------------------------------------------------
5300 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5307 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5308 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5309 at startup during orphan unlinking. Instead, skip buddy update
5310 before setup, as it will be regenerated after recovery is complete.
5315 Description: If an OST is inactive, its locks might reference stale inodes.
5316 Details : lov_change_cbdata() must iterate over all namespaces, even if
5317 they are inactive to clear inode references from the lock.
5319 Severity : enhancement
5320 Frequency : occasional, if non-standard max_dirty_mb used
5322 Description: Client will block write RPCs if not enough grant
5323 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5324 then the client will block writes while waiting for another RPC
5325 to complete instead of consuming its dirty limit. With change
5326 we get improved performance when max_dirty_mb is small.
5328 Severity : enhancement
5329 Bugzilla : 3389, 6253
5330 Description: Add support for supplementary groups on the MDS.
5331 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5332 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5333 lookups for user supplementary groups into a cache.
5337 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5338 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5339 for liblustre O_EXCL is enforced by MDS.
5343 Description: readdir never returns NULL in liblustre.
5344 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5345 dentry in struct dirent.
5349 Frequency : liblustre only, depends on application IO pattern
5350 Description: liblustre clients evicted if not contacting servers
5351 Details : Don't put liblustre clients into the ping_evictor list, so
5352 they will not be evicted by the pinger ever.
5354 Severity : enhancement
5356 Description: Add ability to evict clients by NID from MDS.
5357 Details : By echoing "nid:$NID" string into
5358 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5359 $NID would be instantly evicted from this MDS and from all active
5360 OSTs connected to it.
5364 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5365 Details : lookup_it_finish() used to query file size from OSTs that was not
5370 Description: service threads change working directory to that of init
5371 Details : Starting lustre service threads may pin the working directory
5372 of the parent thread, making that filesystem busy. Threads
5373 now change to the working directory of init to avoid this.
5377 Frequency : during shutdown only
5378 Description: shutdown with a failed MDS or OST can cause unmount to hang
5379 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5384 Frequency : 2.6 only
5385 Description: chmod/chown may include an extra supplementary group
5386 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5387 supplementary group and if none is specified this is used.
5390 Bugzilla : 5479 (6816)
5392 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5393 Details : If lookup is in progress on a file that is unlinked we might try
5394 to revalidate the inode and fail in revalidate after lookup is
5395 complete and ll_file_open() enqueues the open again but
5396 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5399 Frequency : always, if lconf --abort_recovery used
5401 Description: lconf --abort_recovery fails with 'Operation not supported'
5402 Details : lconf was attempting to abort recovery on the MDT device and not
5405 ------------------------------------------------------------------------------
5407 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5412 Frequency : rare (only unsupported configurations with a node running as an
5414 Bugzilla : 6514, 5137
5415 Description: Mounting a Lustre file system on a node running as an OST could
5417 Details : OSTs now preallocates memory needed to write out data at
5418 startup, instead of when needed, to avoid having to
5419 allocate memory in possibly low memory situations.
5420 Specifically, if the file system is mounted on on OST,
5421 memory pressure could force it to try to write out data,
5422 which it needed to allocate memory to do. Due to the low
5423 memory, it would be unable to do so and the node would
5424 become unresponsive.
5426 Severity : enhancement
5428 Description: Addition of lconf --service command line option
5429 Details : lconf now accepts a '--service <arg>' option, which is
5430 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5432 Severity : enhancement
5434 Description: Failover mode is now the default for OSTs.
5435 Details : By default, OSTs will now run in failover mode. To return to
5436 the old behaviour, add '--failout' to the lmc line for OSTs.
5438 Severity : enhancement
5440 Description: Health checks are now provided for MDS and OSTs
5441 Details : Additional detailed health check information on MSD and OSTs
5442 is now provided through the procfs health_check value.
5445 Frequency : occasional, depends on IO load
5447 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5448 after numerous create/delete cycles
5449 Details : The ext3 inode allocation policy would not allocate new inodes
5450 very well on the OSTs because there are no new directories
5451 being created. Instead we look for groups with free space if
5452 the parent directories are nearly full.
5457 Description: Network or server problems during mount may cause partially
5458 mounted clients instead of returning an error.
5459 Details : The config llog parsing code may overwrite the error return
5460 code during mount error handling, returning success instead
5466 Description: MDS can fail to allocate large reply buffers
5467 Details : After long uptimes the MDS can fail to allocate large reply
5468 buffers (e.g. zconf client mount config records) due to memory
5469 fragmentation or consumption by the buffer cache. Preallocate
5470 some large reply buffers so that these replies can be sent even
5471 under memory pressure.
5475 Frequency : rare (liblustre)
5476 Description: fsx running with liblustre complained that using truncate() to
5477 extend the file doesn't work. This patch corrects that issue.
5478 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5479 ATTR_SIZE and lsm use in llu_setattr_raw.
5483 Frequency : rare, only 2.6 kernels
5484 Description: Unusual file access patterns on the MDS may result in inode
5485 data being lost in very rare circumstances.
5486 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5487 code in the 2.6 kernel caused the inode and/or EA data not to
5488 be read from disk, causing single-file corruption.
5492 Frequency : rare, only 2.6 filesystems using extents
5493 Description: Heavy concurrent write and delete load may cause data corruption.
5494 Details : It was possible under high-load situations to have an extent
5495 metadata block in the block device cache from a just-unlinked
5496 file overwrite a newly-allocated data block. We now unmap any
5497 metadata buffers that alias just-allocated data blocks.
5501 Frequency : filesystems with default stripe_count larger than 77
5502 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5503 Details : lconf specifies an inode size of 4096 bytes when the default
5504 stripe_count is larger than 77. This conflicts with the default
5505 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5506 this case to avoid pinning too much memory for large EAs.
5508 ------------------------------------------------------------------------------
5510 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5515 Frequency : rare (extremely heavy IO load with hundreds of clients)
5517 Description: Client is evicted, gets IO error writing to file
5518 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5519 caused clients to be evicted for AST timeouts. The fixes in
5520 bug 5192 mean we no longer need to have such short AST timeouts
5521 so ldlm_timeout has been increased.
5524 Frequency : occasional during --force or --failover shutdown under load
5525 Bugzilla : 5949, 4834
5526 Description: Server oops/LBUG if stopped with --force or --failover under load
5527 Details : a collection of import/export refcount and cleanup ordering
5528 issues fixed for safer force cleanup
5531 Frequency : only filesystems larger than 120 OSTs
5532 Bugzilla : 5990, 6223
5533 Description: lfs getstripe would oops on a very large filesystem
5534 Details : lov_getconfig used kfree on vmalloc'd memory
5537 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5538 Bugzilla : 6242, 6243
5539 Description: reading from files that had been truncated to a non-zero size
5540 but never opened returned no data
5541 Details : ll_file_read() reads zeros from no-object files to EOF
5546 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5547 Details : There's little harm in aborting MDS/OSS recovery and letting it
5548 try again, so I removed the LASSERT and return an error instead.
5550 Severity : enhancement
5552 Description: New debugging infrastructure for tracking down data corruption
5553 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5554 (b) cover more of the client-side code path, and (c) try to narrow
5555 down where problems occurred
5559 Bugzilla : 3819, 4364, 4397, 6313
5560 Description: Racing close and eviction MDS could cause assertion in mds_close
5561 Details : It was possible to get multiple mfd references during close and
5562 client eviction, leading to one thread referencing a freed mfd.
5564 Severity: : enhancement
5565 Bugzilla : 3262, 6359
5566 Description: Attempts to reconnect to servers are now more aggressive.
5567 Details : This builds on the enhanced upcall-less recovery that was added
5568 in 1.4.2. When trying to reconnect to servers, clients will
5569 now try each server in the failover group every 10 seconds. By
5570 default, clients would previously try one server every 25 seconds.
5575 Description: After recovery, certain operations trigger a failed
5576 assertion on a client.
5577 Details : Failing over an mds, using lconf -d --failover, while a
5578 client was doing a readdir() call would cause the client to
5579 LBUG after recovery completed and the readdir() was resent.
5581 Severity : enhancement
5583 Description: Default groups are now added by lconf
5584 Details : You can now run lconf --group <servicename> without having to
5585 manually add groups with lmc.
5588 Frequency : occasional
5590 Description: Nodes with an elan id of 0 trigger a failed assertion
5593 Frequency : always when accessing e.g. tty/console device nodes
5595 Description: tty and some other devices nodes cannot be used on lustre
5596 Details : file's private_data field is used by device data and lustre
5597 values in there got lost. New field was added to struct file to
5598 store fs-specific private data.
5601 Frequency : when exporting Lustre via NFS
5603 Description: NFSD failed occasionally when looking up a path component
5604 Details : NFSD is looking up ".." which was broken in ext3 directories
5605 that had grown large enough to become hashed.
5608 Frequency : Clusters with multiple interfaces not on the same subnet
5610 Description: Nodes will repeatedly try to reconnect to an interface which it
5611 cannot reach and report an error to the log.
5612 Details : Extra peer list entries will be created by lconf with some peers
5613 unreachable. lconf now validates the peer before adding it.
5616 Frequency : Only if a default stripe is set on the filesystem root.
5618 Description: Setting a default stripe on the filesystem root prevented the
5619 filesystem from being remounted.
5620 Details : The client was sending extra request flags in the root getattr
5621 request and did not allocate a reply buffer for the dir EA.
5624 Frequency : occasional, higher if lots of files are accessed by one client
5625 Bugzilla : 6159, 6097
5626 Description: Client trips assertion regarding lsm mismatch/magic
5627 Details : While revalidating inodes the VFS looks up inodes with ifind()
5628 and in rare cases can find an inode that is being freed.
5629 The ll_test_inode() code will free the lsm during ifind()
5630 when it finds an existing inode and then the VFS later attaches
5631 this free lsm to a new inode.
5635 Bugzilla : 6422, 7030
5636 Description: MDS deadlock between mkdir and client eviction
5637 Details : Creating a new file via mkdir or mknod (starting a transaction
5638 and getting the ns lock) can deadlock with client eviction
5639 (gets ns lock and trying to finish a synchronous transaction).
5642 Frequency : occasional
5643 Description: While starting a server, the fsfilt_ext3 module could not be
5645 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5646 kernels. Previously, lconf would still use the ext3 name
5647 when trying to load modules. Now, it will correctly use
5648 ext3 on 2.4 and ldiskfs on 2.6.
5650 Severity : enhancement
5651 Description: The default stripe count has been changed to 1
5652 Details : The interpretation of the default stripe count (0, to lfs
5653 or lmc) has been changed to mean striping across a single
5654 OST, rather than all available. For general usage we have
5655 found a stripe count of 1 or 2 works best.
5657 Severity : enhancement
5658 Description: Add support for compiling against Cray portals.
5659 Details : Conditional compiling for some areas that are different
5663 Frequency : occasional
5664 Bugzilla : 6409, 6834
5665 Description: Creating files with an explicit stripe count may lead to
5666 a failed assertion on the MDS
5667 Details : If some OSTs are full or unavailable, creating files may
5668 trigger a failed assertion on the MDS. Now, Lustre will
5669 try to use other servers or return an error to the
5673 Frequency : occasional
5675 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5676 caused lock timeout during readahead (since 1.4.2).
5677 Details : Processes doing readahead might match a lock that hasn't been
5678 granted yet if there are overlapping and conflicting lock
5679 requests. The readahead process waits on ungranted lock
5680 (original lock is CBPENDING), while OST waits for that process
5681 to cancel CBPENDING read lock and eventually evicts client.
5683 Severity : enhancement
5685 Description: Initial enabling of flock support for clients
5686 Details : Implements fcntl advisory locking and file status functions.
5687 This feature is provided as an optional mount flag (default
5688 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5689 locking are implemented yet, and those that are are not guaranteed
5690 to be completely correct in production environments.
5691 mount -t lustre -o [flock|noflock] ...
5694 Frequency : occasional
5696 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5697 assertion in the ext3 JBD (journaling) layer.
5698 Details : The b_committed_data struct is protected by the big kernel lock
5699 in 2.4 kernels, serializing journal_commit_transaction() and
5700 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5701 access to this struct. In 2.6 kernels there is finer grained
5702 locking to improve SMP performance of the JBD layer.
5706 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5711 Description: merge and cleanup kernel patches.
5714 -----------------------------------------------------------------------------
5716 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5718 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5719 versions of Lustre. Please update all servers and clients to
5720 version 1.4.2 or later at the same time. You must also run
5721 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5722 to update the configuration logs.
5724 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5725 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5726 - swab llog records of type '0' so we get proper header size/idx (5861)
5727 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5728 - fix rename of one directory over another leaking an inode (5953)
5729 - avoid SetPageDirty on 2.6 (5981)
5730 - don't re-add just-being-destroyed locks to the waiting list (5653)
5731 - when creating new directories, inherit the parent's custom
5732 striping settings if present parent (3048)
5733 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5734 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5735 - don't hold client locks on temporary worklist from l_lru (5666)
5736 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5737 - automatically evict dead clients (5921)
5738 - Update file size properly in create+truncate+fstat case (6196)
5739 - Do not unhash mountpoint dentries, do not allow removal of
5741 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5742 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5743 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5744 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5745 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5746 - Correctly handle reads of files with no objects (6243)
5747 - lctl recover will also mark a device active if deactivate used (5933)
5749 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5750 - allow --write-conf on an MDS with different nettype than client (5619)
5751 - don't write config llogs to MDS for mounts not from that MDS (5617)
5752 - lconf should create multiple TCP connections from a client (5201)
5753 - init scripts are now turned off by default; run chkconfig --on
5754 lustre and chkconfig --on lustrefs to use them
5755 - upcalls are no longer needed for clients to recover to failover
5757 - add --abort-recovery option to lconf to abort recovery on device
5759 - add support for an arbitrary number of OSTs (3026)
5760 - Quota support protocol changes.
5761 - forward compatibility changes to wire structs (6007)
5762 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5763 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5764 - improved "lustre status" script
5765 - initialize blocksize for non-regular files (6062)
5766 - added --disable-server and --disable-client configure options (5782)
5767 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5768 - Vanilla 2.4.29 support
5769 - increase maximum number of obd devices to 520 (6242)
5770 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5771 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5772 SLES 9 2.6.5 kernels
5773 - limit stripes per file to 160 (the maximum EA size) (6093)
5775 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5778 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5779 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5780 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5781 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5782 - small ext3 extents cleanups and fixes (5733)
5783 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5784 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5785 - avoid needless client->OST connect, fix handle mismatch (5317)
5786 - fix DLM error path that led to out-of-sync client, long delays (5779)
5787 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5788 - fix several locking issues related to i_size (5492,5624,5654,5672)
5789 - don't move pending lock onto export if it is already evicted (5683)
5790 - fix kernel oops when creating .foo in unlinked directory (5548)
5791 - fix deadlock in obdfilter statistics vs. object create (5811)
5792 - use time_{before,after} to avoid timer jiffies wrap (5882)
5793 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5794 - Do not leak request if server was not able to process it (5154)
5795 - If mds_open unable to find parent dir, make that negative lookup(5154)
5796 - don't create new directories with extent-mapping (5909, 5936)
5798 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5799 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5800 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5801 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5802 - support for RHEL4 kernel on i686 (5773)
5803 - provide error messages when incompatible logs are encountered (5898)
5805 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5806 * version 1.4.0.10 (1.4.1 release candidate 1)
5808 - don't keep a lock reference when lock is not granted (4238)
5809 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5810 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5811 - drop import inflight refcount on signal_completed_replay error (5255)
5812 - unlock page after async write error during send (3677)
5813 - handle missing objects in filter_preprw_read properly (5265)
5814 - no transno return for symlink open, don't save no-trasno open (3440)
5815 - don't try to complete elan receive that already failed (4012)
5816 - free RPC server reply state on error (5406)
5817 - clean up thread from ptlrpc_start_thread() on error (5160)
5818 - readahead could read extra page into cache that wasn't ejected (5388)
5819 - prevent races in class_attach/setup/cleanup/detach (5260)
5820 - don't dereference de->d_inode after l_dput of de (5458)
5821 - use "int" for stripe value returned from lock_to_stripe (5544)
5822 - mballoc allocation and error-checking fixes in 2.6 (5504)
5823 - block device patches to fix I/O request sizes in 2.6 (5482)
5824 - look up hostnames for IB nals (5602)
5825 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5826 - don't start multiple acceptors for the same port (5277)
5827 - fix incorrect LASSERT in mds_getattr_name (5635)
5828 - export a proc file for general "ping" checking (5628)
5829 - fix "lfs check" to not block when the MDS is down (5628)
5831 - service request history (4965)
5832 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5833 - create an "evict_client" /proc entry on OSTs, like the MDS has
5834 - fix mount usage message, return errors per mount(8) (5168)
5835 - change grep [] to grep "[]" in tests so they work in more UMLs
5836 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5837 - remove ancient LOV_MAGIC_V0 EA support (5047)
5838 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5839 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5840 - allow readahead allocations to fail when low on memory (5383)
5841 - mmap locking landed again, after considerable improvement (2828)
5842 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5844 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5847 - send OST transaction number in read/write reply to free req (4966)
5848 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5849 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5850 - the watchdog thread now runs as interruptible (5246)
5851 - flock/lockf fixes (but it's still disabled, pending 5135)
5852 - don't use EXT3 constants in llite code (5094)
5853 - memory shortage at startup could cause assertion (5176)
5855 - reorganization of lov code
5856 - single portals codebase
5858 - add extents/mballoc support (5025)
5859 - direct I/O reads in the obdfilter (4048)
5860 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5862 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5865 - send OST transaction number in read/write reply to free req (4966)
5866 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5867 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5868 - the watchdog thread now runs as interruptible (5246)
5869 - handle missing objects in filter_preprw_read properly (5265)
5870 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5871 - drop import inflight refcount on signal_completed_replay error (5255)
5872 - unlock page after async write error during send (3677)
5873 - return original error code on reconstructed replies (3761)
5874 - no transno return for symlink open, don't save no-trasno open (3440)
5876 - add pid to ldlm debugging output (4922)
5877 - bump the watchdog timeouts -- we can't handle 30sec yet
5878 - extra debugging for orphan dentry/inode bug (5259)
5880 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5883 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5884 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5885 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5886 - fix NULL dereference in /proc/sys/portals/routes (4827)
5887 - allow failed mdc_close() operations to be interrupted (4561)
5888 - stop precreate on OST before MDS would time out on it (4778)
5889 - don't send partial-page writes before EOF from client (4410)
5890 - discard client grant for sub-page writes on large-page clients (4520)
5891 - don't free dentries not owned by NFS code, check generation (4806)
5892 - fix lsm leak if mds_create_objects() fails (4801)
5893 - limit debug_daemon file size, always print CERROR messages (4789)
5894 - use transno after validating reply (3892)
5895 - process timed out requests if import state changes (3754)
5896 - update mtime on OST during writes, return in glimpse (4829)
5897 - add mkfsoptions to LDAP (4679)
5898 - use ->max_readahead method instead of zapping global ra (5039)
5899 - don't interrupt __l_wait_event() during strace
5901 - add software watchdogs to catch hung threads quickly (4941)
5902 - make lustrefs init script start after nfs is mounted
5903 - fix CWARN/ERROR duplication (4930)
5904 - return async write errors to application if possible (2248)
5905 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5906 - print NAL number in %x format (4645)
5907 - update barely-supported suse-2.4.21-171 series (4842)
5908 - support for sles 9 %post scripts
5909 - support for building 2.6 kernel-source packages
5910 - support for sles km_* packages
5912 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5915 - ignore -ENOENT errors in osc_destroy (3639)
5916 - notify osc create thread that OSC is being cleaned up (4600)
5917 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5918 - reconstruct ost_handle() like mds_handle() (4657)
5919 - create a new thread to do import eviction to avoid deadlock (3969)
5920 - let lconf resolve symlinked-to devices (4629)
5921 - don't unlink "objects" from directory with default EA (4554)
5922 - hold socknal file ref over connect in case target is down (4394)
5923 - allow more than 32000 subdirectories in a single directory (3244)
5924 - fix blocks count for O_DIRECT writes (3751)
5925 - OST returns ENOSPC from object create when no space left (4539)
5926 - don't send truncate RPC if file size isn't changing (4410)
5927 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5928 - bind to privileged port in socknal and tcpnal (3689)
5930 - rate limit CERROR/CWARN console message to avoid overload (4519)
5931 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5932 - basic mmap support (3918)
5933 - kernel patch series update from b1_4 (4711)
5935 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5938 - avoid crash during MDS cleanup with OST shut down (2775)
5939 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5940 - don't use bad inodes on the MDS (3744)
5941 - dynamic object preallocation to improve recovery speed (4236)
5942 - don't hold spinlock over lock dumping or change debug flags (4401)
5943 - don't zero obd_dev when it is force cleaned (3651)
5944 - print grants to console if they go negative (4431)
5945 - "lctl deactivate" will stop automatic recovery attempts (3406)
5946 - look for existing locks in ldlm_handle_enqueue() (3764)
5947 - don't resolve lock handle twice in recovery avoiding race (4401)
5948 - revalidate should check working dir is a directory (4134)
5950 - don't always mark "slow" obdfilter messages as errors (4418)
5952 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5955 - don't close LustreDB during write_conf until it is done (3860)
5956 - fix typo in lconf for_each_profile (3821)
5957 - allow dumping logs from multiple threads at one time (3820)
5958 - don't allow multiple threads in OSC recovery (3812)
5959 - fix debug_size parameters (3864)
5960 - fix mds_postrecov to initialize import for llog ctxt (3121)
5961 - replace config semaphore with spinlock (3306)
5962 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5963 - don't allow enqueue to complete on a destroyed export (3822)
5964 - down write_lock before checking llog header bitmap (3825)
5965 - recover from lock replay timeout (3764)
5966 - up llog sem before sending rpc (3652)
5967 - reduce ns lock hold times when setting kms (3267)
5968 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5969 - fix NULL deref and obd_dev leak on setup error (3312)
5970 - replace some LBUG about llog ops with error handling (3841)
5971 - don't match INVALID dentries from d_lookup and spin (3784)
5972 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5973 - fix invalid assertion in ptlrpc_set_wait (3880)
5975 - add libwrap support for the TCP acceptor (3996)
5976 - add /proc/sys/portals/routes for non-root route listing (3994)
5977 - allow setting MDS UUID in .xml (2580)
5978 - print the stack of a process that LBUGs (4228)
5980 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5983 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5984 - make sure to unset replay flag from failed open requests (3440)
5985 - if default stripe count is 0, use OST count for inode size (3636)
5986 - update parent mtime/ctime on client for create/unlink (2611)
5987 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5988 - free recovery state on server during a forced cleanup (3571)
5989 - unregister_reply for resent reqs (3063)
5990 - loop back devices mounting and status check on 2.6 (3563)
5991 - fix resource-creation race that can provoke i_size == 0 (3513)
5992 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5993 - more debugging for page-accounting assertion (3746)
5994 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5995 - avoid deadlock after precreation failure (3758)
5996 - fix race and lock order deadlock in orphan handling (3450, 3750)
5997 - add validity checks when grabbing inodes from l_ast_data (3599)
5999 - add /proc/.../recovery_status to obdfilter (3428)
6000 - lightweight CDEBUG infrastructure, debug daemon (3668)
6001 - change default OSC RPC parameters to be better on small clusters
6002 - turn off OST read cache for files smaller than 32MB
6003 - install man pages and include them in rpms (3100)
6004 - add new init script for (un)mounting lustre filesystems (2593)
6005 - run chkconfig in %post for init scripts (3701)
6006 - drop scimac NAL (unmaintained)
6008 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6011 - clean kiobufs before and after use (3485)
6012 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6013 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6014 - change init script to fail more clearly if not run as root (1528)
6015 - allow clients to reconnect during replay (1742)
6016 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6017 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6018 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6019 - fix cond_resched (3554)
6020 - only evict unfinished clients after recovery (3515)
6021 - allow bulk resend, prevent data loss (3570)
6022 - dynamic ptlrpc request buffer allocation (2102)
6023 - don't allow unlinking open directory if it isn't empty (2904)
6024 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6025 - remove extraneous obd dereference causing LASSERT failure (3334)
6026 - don't use get_cycles() when creating temp. files on the mds (3156)
6027 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6028 - handle EEXIST for set-stripe, set proper directory name (3336)
6030 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6031 - fix ksocknal_fmb_callback() error messages (2918)
6033 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6036 - don't copy lvb into (possibly NULL) reply on error (2983)
6037 - don't deref dentry after dput, don't free lvb on error (2922)
6038 - use the kms to determine writeback rpc length (2947)
6039 - increment oti_logcookies when osc is inactive (2948)
6040 - update client's i_blocks count via lvb messages (2543)
6041 - handle intent open/close of special files properly (1557)
6042 - mount MDS with errors=remount-ro, like obdfilter (2009)
6043 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6044 - don't use cancelling-locks' kms values (2947)
6045 - use highest lock extent for kms, not last one (2925)
6046 - don't dereference ERR_PTR() dentry in error handling path (3107)
6047 - fix thread race in portals_debug_dumplog() (3122)
6048 - create lprocfs device entries at setup instead of at attach (1519)
6049 - common AST error handler, don't evict client on completion race (3145)
6050 - zero nameidata in detach_mnt in 2.6 (3118)
6051 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6052 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6053 - initialize RPC timeout timer earlier for 2.6 (3219)
6054 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6055 - print nal/nid for unknown nid (3258)
6056 - additional checks for oscc recovery before doing precreate (3284)
6057 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6058 - don't crash in mdc_close for bad permissions on open (3285)
6059 - zero i_rdev for non-device files (3147)
6060 - clear page->private before handing to FS, better assertion (3119)
6061 - tune the read pipeline (3236)
6062 - fix incorrect decref of invalidated dentry (2350)
6063 - provide read-ahead stats and refine rpc in flight stats (3328)
6064 - don't hold journal transaction open across create RPC (3313)
6065 - update atime on MDS at close time (3265)
6066 - close LDAP connection when recovering to avoid server load (3315)
6067 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6068 - don't leak open file on MDS after open resend (3325)
6069 - serialize filter_precreate and filter_destroy_precreated (3329)
6070 - loop device shouldn't call sync_dev() for nul device (3092)
6071 - clear page cache after eviction (2766)
6072 - resynchronize MDS->OST in background (2824)
6073 - refuse to mount the same filesystem twice on same mountpoint (3394)
6074 - allow llmount to create routes for mounting behind routers (3320)
6075 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6076 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6077 - fix rare problem with rename on htree directories (3417)
6079 - allow default OST striping configuration per directory (1414)
6080 - fix compilation for qswnal for 2.6 kernels (3125)
6081 - increase maximum number of MDS request buffers for large systems
6082 - change liblustreapi to be useful for external progs like lfsck (3098)
6083 - increase local configuration timeout for slow disks (3353)
6084 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6086 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6089 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6090 - don't overwrite extent policy data in reply if lock was blocked (2901)
6091 - drop filter export grants atomically with removal from device (2663)
6092 - del obd_self_export from work_list in class_disconnect_exports (2908)
6093 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6094 - swab reply message in mdc_close, other PPC fixes (2464)
6095 - fix destroying of named logs (2325)
6096 - overwrite old logs when running lconf --write_conf (2264)
6097 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6098 - fix race in target_handle_connect (2898)
6099 - mds_reint_create() should take same inode create lock (2926)
6100 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6101 - don't close files for self_export to avoid uninitialized obd (2936)
6102 - allow MDS with the same name as client node (2939)
6103 - hold dentry reference for closed log files for unlink (2325)
6104 - reserve space for all logs during transactions (2059)
6105 - don't evict page beyond end of stripe extent (2925)
6106 - don't oops on a deleted current working directory (2399)
6107 - handle hard links to targets without a parent properly (2517)
6108 - don't dereference NULL lock when racing during eviction (2867)
6109 - don't grow lock extents when lots of conflicting locks (2919)
6111 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6114 - account for cache space usage on clients to avoid data loss (974)
6115 - lfsck support in lustre kernel code (2349)
6116 - reduce journal credits needed for BRW writes (2370)
6117 - orphan handling to avoid losing space on client/server crashes
6118 - ptlrpcd can be blocked, stopping ALL progress (2477)
6119 - use lock value blocks to assist in proper KMS, faster stat (1021)
6120 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6121 - recovery for initial connections (2355)
6122 - fixes for mds_cleanup_orphans (1934)
6123 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6124 - block all file creations until orphan recovery completes (1901)
6125 - client remove rq_connection from request struct (2423)
6126 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6127 - recovery timer race (2670)
6128 - mdc_close recovey bug (2532)
6129 - ptlrpc cleanup bug (2710)
6130 - mds timeout on local locks (2588)
6131 - namespace lock held during RPCs (2431)
6132 - handle interrupted sync write properly (2503)
6133 - don't try to handle a message that hasn't been replied to (2699)
6134 - client assert failure during cleanup after abort recovery (2701)
6135 - leak mdc device after failed mount (2712)
6136 - ptlrpc_check_set allows timedout requests to complete (2714)
6137 - wait for inflight reqs when ptlrpcd finishes (2710)
6138 - make sure unregistered services are removed from the srv_list
6139 - reset bulk XID's when resending them (caught by 1138 test)
6140 - unregister_bulk after timeout
6141 - fix lconf error (2694)
6142 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6143 - readahead locks pages, leaves pending causing memory pressure (2673)
6144 - increase OST request buffers to 4096 on large machines (2729)
6145 - fix up permission of existing directories in simple_mkdir (2661)
6146 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6147 - don't assign transno to errored transactions (2742)
6148 - don't delete objects on OST if given a bogus objid from MDS (2751)
6149 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6150 - if rq_no_resend, then timeout request after recovery (2432)
6151 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6152 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6153 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6154 - MDS recovery shouldn't delete objects that it has given out (2730)
6155 - if enqueue arrives after completion, don't clobber LVB (2819)
6156 - don't unlock pages twice when trigger_group_io returns error (2814)
6157 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6158 - don't write pages to disk if there was an error (1450)
6159 - don't ping imports that have recovery disabled (2676)
6160 - take buffered bytes into account when balancing socknal conn (2817)
6161 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6162 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6163 - remove little-endian swabbing of llog records (1987)
6164 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6165 - retry reposting request buffers if they fail (1191)
6166 - grow extent at grant time to avoid granting a revoked lock (2809)
6167 - lock revoke doesn't evict page if covered by a second lock (2765)
6168 - disable VM readahead to avoid reading outside lock extents (2805)
6170 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6171 - updated kernel patches for hp-2.4.20 kernel (2681)
6173 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6176 - fix truncated write corruption (2366)
6177 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6179 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6180 - ensure len(uuid) < 37 in lmc (1171)
6181 - fix ia64 OOPS in llog_test (2255)
6182 - zero end of page at obdfilter for partial page writes (2648)
6183 - don't leave stale dentries around after renames (bug 2428)
6184 - fix timeouts when evicting a client with a single lock held (2642)
6185 - set deadline for the initial HELLO message to drain (2634)
6186 - print out dotted-quad IP addresses in the socknal (2302)
6188 - additional debugging for MDS client eviction problem (2443)
6189 - fix mkfsoptions support for osts (2603, 2604)
6191 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6194 - add series for the vanilla 2.6.0 kernel
6195 - add series for the vanilla 2.4.24 kernel
6196 - add series for a cray x86/64 UL kernel drop
6197 - fix xattr patches for the vanilla 2.4.19 series
6199 - generate true UUIDs in lmc (1171)
6200 - have portals stack dumping break in UML (2466)
6201 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6202 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6203 - wake callback waiting threads on client eviction (2460)
6204 - Add --ptldebug and --subsystem to lmc (1719)
6205 - update assertion to allow safe interrupt allocation
6206 - set rq_no_resend for cancel requests (2432)
6207 - recalculate ptlrpcd timeout after resend (2494)
6208 - call vfs_rmdir when removing pending directories (2368)
6209 - fix renaming a file to itself (2429)
6210 - lmc creates a default one-stripe lov (2454)
6211 - expand procfs space to handle large clusters (2326)
6212 - increase UML stack to avoid overflow
6213 - update lconf's list of debug and subsystem masks
6214 - fix lfs find --obd (2510)
6215 - /proc tunable for disabling filter read caching (2591)
6216 - stop rpm packages from altering slapd.conf (2301)
6217 - disable nagle in the socknal under 0conf (2578)
6218 - choose mds inode size based on stripe count (2572)
6219 - fix kernel-source rpm problems (2516)
6221 - add --disable-doc to avoid pdf generation (2421)
6222 - update documentation, tests, type-os, comments
6223 - avoid format warnings on ia64
6224 - remove the TOE NAL
6225 - tiny code cleanups by removing unused fields
6227 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6230 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6231 - remove the most blatant lies from BUILDING file (2371)
6232 - change default debug level to reasonable production setting
6233 - reduce client side cache size to reduce cache flush time
6234 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6235 - make TCP zerocopy and pinger support enabled by default (2476)
6236 - sync writes completed after process exits caused crashes (2319)
6237 - maintain correct mount count on the MDS (2356)
6238 - backout 1557, because 2316 wasn't really fixed
6239 - better file I/O statistics gathering in /proc
6240 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6241 - another kernel patch to fix zero-copy TCP function export
6242 - don't take duplicate lock when processing re-sent getattr (2420)
6243 - lctl uses obd_self_export instead of creating new conn (2353)
6244 - MDS/OST recovery case which requires object creation asserted (2425)
6245 - move lfs from /usr/sbin to /usr/bin in packages
6246 - fix race between mds_client_add and mds_client_free (2417)
6247 - use kmalloc instead of slabs in portals (2430)
6248 - don't create duplicate records when a failover MDS is present (2442)
6249 - remove unnecessary mount age check (2332)
6250 - don't remove directory inodes from locks prematurely (2451)
6251 - don't break if MDS service name is the same as hostname (2103)
6252 - fix races in client write RPC generation when cache full (2482)
6254 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6257 - remove now-unused request->rq_obd (278)
6258 - if an allocation fails, print out how much memory we've used (1933)
6259 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6260 - add the "configurable stack size" patch to most series files (1256)
6261 - ability to write large log records, for 100+ OST configs (2306)
6262 - fix NULL deref when filter_prep fails (2314)
6263 - fix operator precedence error in filter_sync
6264 - dynamic allocation of socknal TX descriptors (2315)
6265 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6266 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6267 - ensure that CWARN messages in Portals always get to the syslog
6268 - __init/__exit are not for prototype decls (ldlm_init/exit)
6269 - x86-64 compile warning fixes
6270 - fix gateway LMC keyword conflict (2318)
6271 - fix MDS lock inversions in getattr/reint paths (1844)
6272 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6273 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6274 - DLM race condition prevented some lock evictions (2328)
6275 - ENOMEM detection and retry on socknal sends (2230)
6276 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6277 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6278 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6279 - better error messages when a client is rejected during recovery (1505)
6280 - avoid cancelling locks which were never granted, after failure (2330)
6281 - fix i_sem/journal inversion in mds_client_add (2333)
6282 - fix truncate/getattr lock cycle deadlock (2334)
6283 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6284 - fix two rare exit paths which could leak an l_lock() ref (2321)
6285 - fencepost error in MDS/OST orphan recovery (2226)
6286 - make log record alignment 8 bytes (1988)
6287 - lstripe now fails when requested offset > ost_count (2237)
6288 - ensure that all kernel series have a complete list.h (1607)
6289 - fix crashes in special-file operations (2316)
6290 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6292 - allow configurable automake binary, for testing new versions
6293 - small update to the lfs documentation
6295 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6297 * fix negative export reference count in fsfilt_sync (2312)
6299 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6300 * release candidate 0.9.1
6302 - orphans are moved into the PENDING directory for possible recovery
6303 - replayed opens now open by fid for orphan/rename safety (1042)
6304 - last close of an orphan inode generates a transno (683)
6305 - chdir() and mount() now pin the directory entry (1020)
6306 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6307 - discard very old requests without processing them (1502)
6308 - remove some common, well-understood CERRORs (1505)
6309 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6310 - clear "grant" flags in OST replies until OST grant code lands (1644)
6311 - fix read performance by not clobbering i_blksize on client (1598)
6312 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6313 - make LRU size a /proc tunable, clears locks when reduced (707)
6314 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6315 - server locks take references on exports now (1558)
6316 - build fixes for 2.4.20-rh trees (1663)
6317 - return an error from lov_create if all OSCs are inactive (1751)
6318 - fix import levels when a reconnect happens without a timeout (1597)
6319 - exit early from mds_open if we get a lookup error (1749)
6320 - partial page read at EOF wouldn't wait for disk before sending (1642)
6321 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6322 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6323 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6324 - write/truncate lock inversion (1639)
6325 - Don't auto-load obdclass, portals modules during cleanup (1495)
6326 - fix timestamps from jumping to "now" (1763)
6327 - extra journal assertions (1648)
6328 - add an extra multiunlink test (1771)
6329 - fix read_record/write_record API (1776)
6330 - fix leak of offset_extent, possible incorrect i_size later (1772)
6331 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6332 - Debugging for the kqswnal_get_idle_tx problems (1820)
6333 - Allow recovery to be attempted multiple times (1536)
6334 - Write out MDS last_rcvd file after it is first created (1600)
6335 - Fix tx_descriptor leak in failed transmit situations (1827)
6336 - ext3 journaling fixes for assertion failure after IO error (1871)
6337 - class_export_put() on freed export after completion AST error (1896)
6338 - Fix revalidate looping in VFS (1322)
6339 - Don't access a freed export during MDS_REINT timeout (1521)
6340 - Add open-unlink recovery support on the MDS (1673,1764)
6341 - Return an error if no MDS data was read from last_rcvd (1946)
6342 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6343 - Don't setup a disk device that doesn't match exported UUID (317)
6344 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6345 - avoid committing NULL handle in force close
6346 - local.sh is now a one-stripe LOV configuration
6347 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6348 - fix ext3 htree duplicate directory entry corruption (1516)
6349 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6350 - update to new LOV EA format (2097)
6351 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6353 2003-06-15 Phil Schwan <phil@clusterfs.com>
6356 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6357 - per-import/export recovery handling (958, 931, 959)
6358 - multiple last-rcvd slots, for serving multiple FSes (949)
6359 - connections are again shared between multiple imp/exports (963, 964)
6360 - "umount -f" would hang if any requests needed to be sent (393, 978)
6361 - avoid pinning large req buffer by copying for queued messages (989)
6362 - add "uuid" to "lctl device" command to help upcalls (991)
6363 - "open" RPCs with transnos would confuse recovery counters (1037)
6364 - do proper endian conversion of all wire messages (288, 340, 891)
6365 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6366 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6367 - fix ptlrpc_connection leak in target_handle_connect (1174)
6368 - fix import refcounting bug in OST and MDS cleanup (1134)
6369 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6370 - fix very unlikely obd_types race condition (501)
6371 - remove osc_open hack for echo_client (1187)
6372 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6373 - a failure in read_inode2 leads to deadlock (1139)
6374 - cancel ack-locks as soon as transaction is committed (1072)
6375 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6376 - make sure to commitrw after any preprw to avoid deadlock (1162)
6377 - failing to execute a file in a lustre FS would lock inode (1203)
6378 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6379 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6380 - obd_uuid2tgt would walk off the end of the list (1255)
6381 - on IA64 the osc would give portals incorrect bulk size (1258)
6382 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6383 - fix lock inversion caused by new llite matching code (1282)
6384 - limit the number of dirty pages on a client to 10MB (1286)
6385 - timed out locks were not being corrected cancelled (1289)
6386 - fix O_DIRECT above 4GB on IA-32 (1292)
6387 * major user-visible changes
6388 - fail out/fail over policy now controlled by the upcall (993)
6390 - add OBD_PING to check server availability and failure (954)
6391 - lustre messages are now sent in sending host order (288, 340, 891)
6392 - add eadatalen to MDS getattr reply (340)
6393 - OST read replies may contain second buffer, with per-page status (593)
6395 2003-03-11 Phil Schwan <phil@clusterfs.com>
6398 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6399 - failed open()s could cause deadlock; fixed (867, 869)
6400 - stop cancelling OST locks when files are closed (481)
6401 - overlapping XID spaces caused network corruption (851, 853)
6402 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6403 - setattr_raw added, to do single-RPC, server-side setattrs
6404 - lmc/lconf syntax change for OST UUIDs
6405 - fix crashy race condition between ptlrpc_free_req and osc_close
6406 - don't use request in mdc_enqueue if we hit a timeout (889)
6407 - don't set the inode i_size for regular files from the MDS (896)
6408 - handle out of order completion AST (842)
6409 - don't LBUG if a lock request times out after receiving AST (913)
6410 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6411 - if a bad lock AST arrives, send an error instead of dropping entirely
6412 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6413 - fix leak in bulk IO when only partially completed (899, 900, 926)
6414 - fix O_DIRECT for ia64 (55)
6415 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6416 - C-z after timeout could hang a process forever; fixed (977)
6418 - client-side I/O cache (678, 924, 929, 941, 970)
6420 - READPAGE and SETATTRs which don't take server-side locks get
6423 2003-02-11 Phil Schwan <phil@clusterfs.com>
6426 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6427 - Fix fsfilt_extN_readpage() to read a full page of directory
6428 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6429 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6430 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6431 - Most intent operations were removed, in favour of a new RPC mode
6432 that does a single RPC to the server and bypasses most of the VFS
6433 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6434 - Aggressively cancel local locks on DLM servers
6435 - mds_reint_unlink sends EA to the client if it's the last nlink.
6436 client uses that EA to unlink OST objects.
6437 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6438 - recursive symlinks were fixed (439)
6439 - fixed NULL deref in DEBUG_REQ
6440 - filter_update_lastobjid no longer calls sync, which annoyed extN
6441 - fixed multi-client small-writes to a single file problem (445)
6442 - fixed mtime updates during file writes (607)
6443 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6444 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6445 - fixed Linux OST size reporting problem (444, 656)
6446 - OST now updates object mtime with writes or setattr (607, 619)
6447 - client verifies file size before zeroing page past EOF (445)
6448 - OST now writes last allocated objid to disk with allocation (108)
6449 - LOV on echo now works (409)
6451 - mds_reint_unlink sends a new buffer, with the EA included. this
6452 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6453 set if a regular file was being unlinked, and it was the last link
6454 - use PtlGet from the target for bulk writes (315)
6455 - OST now updates object mtime with writes or setattr (607, 619)
6456 - LDLM now has a grant-time callback to revalidate locked items, if
6458 - Many MDS operations were reorganized to combat race conditions
6460 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6461 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6463 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6466 - Fully reactivate OST imports after reconnection (512, others)
6467 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6468 - More graceful error handling for truncating on dead OST (515)
6469 - Don't error out unless we're actually accessing dead stripes (474)
6470 - Fix garbage sizes when stripes are missing (410)
6471 - LRU counters were broken, causing constant lock purge (433, 432)
6472 - garbage on read from stripes with failed OSTs (441)
6473 - mark OSCs as active before reconnecting during recovery (438)
6474 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6475 - lfind did not preserve OST order in output (443)
6476 - symlinks cause hung clients, incorrect data (439)
6477 - stop dereferencing request after dropping refcount (457)
6478 - don't LASSERT(spin_is_locked) on non-SMP (455)
6479 - fixes for many rename() bugs
6480 - fstat didn't correctly synchronize attributes (399)
6481 - server must handle lock cancellation during blocking AST prep (487)
6482 - bulk descriptors were free()d too soon (511)
6483 - fix paths in lconf, which would load incorrect modules (451, 507)
6484 - fix confusing lconf 'host not found' error message (386)
6485 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6486 - fix race condition in mdc_blocking_ast() for inode access (526)
6487 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6488 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6489 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6490 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6491 - fix lstripe in several ways (kernel side) (527)
6492 - fix request leak in ldlm_cli_enqueue (262)
6493 - incorrect OSC was marked inactive after OST failure
6494 - call mds_fs_cleanup before unmounting filesystem (524)
6495 - fix races between taking ns_lock and ldlm_lock_change_resource
6496 - fix races updating LOV export open file list
6497 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6498 - fix recovery NULL deref in ldlm_cli_cancel_unused
6499 - fix some DLM races by using new hash table for lock handles (419)
6500 - permit the client to specify desired inodes, at replay
6501 - duplicate requests when we queue them for replay reintegration
6502 - fix last_rcvd offset calculation
6503 - sync after each recovered transaction, so we always make progress
6504 - never, not always, ERESTART requests without transnos
6505 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6507 - skip replay if the MDS says that the client is already connected
6508 - don't check for a recovery-enabled export to match lctl's UUID
6509 - don't INC_USE_COUNT for phantom exports
6510 - don't crash when cleaning up phantom exports (567)
6511 - don't double-finish or set replay data for errored mdc_open requests
6512 - abort requests when they time out, so we don't get old replies
6513 - send/receive replies for AST messages again
6514 - if the client says that it doesn't have the lock, cancel it on the
6516 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6517 mark it as destroyed, it will all work out when decref is called
6518 - fix module use counts (22, 581)
6520 - ASTs now expect a reply (server cancels lock on error reply)
6522 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6525 - fix many simultaneous client startup (392)
6526 - fix dentry->d_it clobbering
6527 - credentials weren't being shipped for readdir/getattr operations
6528 - remove invalid assertions triggered during some concurrent MD
6530 - proper Lustre versions added (336, 389)
6531 - fix memory leak for create error case (398)
6532 - fix LOV locking bug that would get cli/srv out of sync
6533 - fix echo client over LOV (409)
6534 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6535 - fix double-O_EXCL intent crash (424)
6536 - avoid sending multiple lock CANCELs (352)
6538 - MDS can do multi-client recovery (modulo bugs in new code)
6540 - many updates, edits, cleanups
6542 2002-11-18 Phil Schwan <phil@clusterfs.com>
6545 - fix null d_it dereference (346)
6546 - fix full OST/dbench hang (333)
6547 - fix permission problem with file removal (286)
6548 - fix removal of OSCs from LOV when they fail
6549 - fix NULL deref during bulk timeout (214)
6550 - fix problems related to multiple filesystems on one MDS (241)
6551 - fixed serious subtle metadata locking bugs
6552 - free locks on clients when inodes are removed due to memory
6554 - fix inode pointer in lock data (285)
6555 - partial support for multiple MDS on a single host (241)
6556 - data locks weren't cancelled at clear_inode time (290, 311)
6557 - intent locks could lead to unbounded lock growth (205)
6558 - added a maximum lock count, an LRU list, and a flusher
6559 - fix multiple rename (365)
6560 - properly abstracted the echo client
6561 - OSC locked 1 byte too many; fixed
6562 - rewrote brw callback code:
6563 - fixed recovery bugs related to LOVs (306)
6564 - fixed too-many-pages-in-one-write crash (191)
6565 - fixed (again) crash in sync_io_timeout (214)
6566 - probably fixed callback-related race (385)
6568 - Add capability to MDS protocol
6569 - LDLM cancellations and callbacks on different portals
6571 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6574 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6575 - defer open object destruction to close time (601981, 138)
6576 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6577 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6578 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6579 - don't call obd_setattr in ll_file_release for destroyed objects
6580 * protocol change to lustre_msg: move |version| and add |flags|
6581 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6582 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6583 * added replay of create, unlink, link and rename operations during
6584 MDS failover; recovery should be much more robust now
6585 * remove failed OSCs from LOVs (only lov_create uses this so far)
6586 * the lustre-HOWTO was brought (more) up to date (582544)
6588 2002-10-23 Phil Schwan <phil@clusterfs.com>
6591 - in-use dentries weren't being reused properly (617851)
6592 - prevent multiple LDLM setup (599178)
6593 - fix LOV size calculations for truncate (617853)
6594 - fix client handling of MDS intent errors (POSIX)
6595 - fix permission bug in lovstripe.c test (624321)
6596 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6597 - truncate past end of file could corrupt data
6598 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6599 - a race in recovery could return ETIMEDOUT to apps (623947)
6600 - building outside the source directory was fixed
6601 * the lustre-HOWTO was brought (more) up to date (582544)
6602 * major progress was made on recovery functionality
6604 2002-10-10 Phil Schwan <phil@clusterfs.com>
6607 - recovery deadlock fix
6608 - rm -rf causes LBUG fix (617817)
6609 - file open by multiple tasks fix (618962)
6610 - directory permissions bugs (602707 and 620007)
6611 - journal_stop fixed with locking (611313)
6612 - O_APPEND failures resolved (618273, perhaps 614459)
6613 - lconf PATH fix (619770)
6614 - IA64 build fix (621450)
6615 - RPC buffer sizes scale with amount of memory
6617 2002-10-01 Phil Schwan <phil@clusterfs.com>
6620 - locks would be cancelled without throwing away data pages,
6621 resulting in inconsistent data (605627)
6622 - inode attributes were not always being refreshed (605627, 612449)
6623 - lconf now continues to cleanup after lctl reports an error
6624 - MDS now enforces user permissions (602707)
6625 - lprocfs cleanup fixed, but not yet enabled (614157)
6626 - fixed infinite server hang, should a client not respond to an AST
6627 - avoid going into recovery if user calls readlink() with a buffer
6628 that's too small (613941)
6629 - AST RPCs no longer require replies (614867) -- this may be changed
6630 - don't crash server if client sends an IOV that's too big (611336)
6631 - fixed lock conversion deadlock (611892)
6632 - fixed the following of symlinks (614622)
6633 * recovery: the server can remove locks from a client that dies, other
6634 clients can make progress
6635 * more extN patch fixes
6636 * compile-time configurable ptlrpc buffer allocations
6638 - collaborative read cache document
6639 - Lustre Lite Performance CDR document-in-progress
6641 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6644 - fix typo in patch-2.4.18
6646 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6649 - clear ptlrpc request each time in handle_incoming_request()
6650 - unlink of files now destroys the object on the OST
6652 2002-09-19 Peter Braam <braam@clusterfs.com>
6654 * add hard link support
6655 * change obdfile creation method
6656 * kernel patch changed
6658 2002-09-19 Peter Braam <braam@clusterfs.com>
6661 - stack overflow bug in extN fixed
6663 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6665 * documentation updates
6666 - add man pages for config tools
6667 - update tests/README to describe testing with new config tools
6668 - finish metadata API descriptions
6669 * bug fixes and cleanups
6670 - statfs workaround for 16TB limit
6671 - LOV stripe allocation improved, can stripe on subset of OSTs
6672 - LOV file size/IO offset was wrong for files > 4GB in size
6673 - object EA data was being dropped, caused files to be unreadable
6674 - memory overflow with non-LOV OST caused memory corruption
6675 - fixed regression tests to work with new config tools, obdfilter
6676 - fixed bug when directory size became larger than 1 block
6677 - fixed bug (for single client case) when PWD was deleted
6678 - invalidate local directory pages when doing intent-based ops
6679 - avoid LDLM oops when lock callback contained bad data
6681 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6683 * documentation updates
6684 * bug fixes and cleanups
6685 - configuration tools
6688 - 64-bit compile warnings
6689 - 64-bit internal statfs data
6691 * test_brw on persistent OST devices
6693 * lprocfs (disabled)
6695 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6697 * documentation updates
6698 * bug fixes and cleanups
6699 * configuration tools
6701 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6704 * many small fixes to 0_5_4
6705 * io/network handling
6706 * thinkos in MDS operations
6708 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6711 * crucial basic fixes to 0.5.3
6712 * IOR, Iozone work over Elan
6715 2002-08-07 Phil Schwan <phil@clusterfs.com>
6716 * version 0_5_3, our first alpha
6717 * we use the new Portals iovs
6718 * documentation updates
6719 * bug fixes and cleanups
6720 * small changes in the DLM wire protocol
6722 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6723 * version 0_5_1 with some initial stability,
6724 * locking on MD and file I/O.
6725 * documentation updates
6726 * several bug fixes since 0.5.0
6727 * small changes in wire protocol
6729 2002-07-18 Phil Schwan <phil@clusterfs.com>
6731 * delivered as Lustre Light Alpha
6732 * fixed a crash after handling invalid MDS requests
6733 * fixed directory pages for architectures with non-4k pages sizes
6735 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6736 * release version v0_4_4
6737 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6739 * Fixes a number of interruption problems with OST operations.
6740 * Update documentation for portals header changes
6741 * Move all wire protocol structs/defines to lustre_idl.h
6742 * Fixes symlink length bug.
6743 * Add tcpdump to repository.
6745 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6746 * release version v0_4_3
6747 * Fixes statfs for inodes on extN.
6748 * Fixes bug in runtests which would delete /etc/hosts.
6749 * Use 64-bit object IDs wherever possible (not into VFS though)
6750 Remove ost_get_info, which is unused by lustre, and out of date.
6752 2002-07-03 Peter Braam <braam@clusterfs.com>
6753 * release version v0_4_2 Fixes a lookup error (type not passed)
6754 * move forward to head of Portals
6755 * move forward to latest Lustre kernel
6757 2002-06-25 Peter Braam <braam@clusterfs.com>
6758 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob