3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.18.1.el7 (RHEL7.7)
7 3.10.0-1127.19.1.el7 (RHEL7.8)
8 3.10.0-1160.2.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.19.1.el8 (RHEL8.2)
15 4.4.120-92.70 (SLES12 SP2)
16 4.4.180-94.100 (SLES12 SP3)
17 4.4.0-131 (Ubuntu 16.04)
18 4.15.0-32 (Ubuntu 18.04)
19 5.4.0-48 (Ubuntu 20.04)
20 vanilla linux 5.4.0 (ZFS + ldiskfs)
21 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
22 * Client primary kernels built and tested during release cycle:
23 3.10.0-1062.18.1.el7 (RHEL7.7)
24 3.10.0-1127.19.1.el7 (RHEL7.8)
25 3.10.0-1160.2.1.el7 (RHEL7.9)
26 4.12.14-95.48 (SLES12 SP4)
27 4.12.14-122.20 (SLES12 SP5)
28 4.12.14-197.37 (SLES15 SP1)
29 5.3.18-24.24 (SLES15 SP2)
30 4.15.0-48 (Ubuntu 18.04)
31 * Other clients known to build on these kernels at some point (others may also work):
32 3.10.0-862.14.4.el7 (RHEL7.5)
33 3.10.0-957.27.2.el7 (RHEL7.6)
34 4.14.0-49.13.1.el7a (RHEL7.5)
35 4.14.0-115.2.2.el7a (RHEL7.6)
36 4.18.0-80.11.2.el8 (RHEL8.0)
37 4.18.0-147.8.1.el8 (RHEL8.1)
38 4.18.0-193.19.1.el8 (RHEL8.2)
39 4.4.120-92.70 (SLES12 SP2)
40 4.4.180-94.100 (SLES12 SP3)
41 4.4.0-131 (Ubuntu 16.04)
42 5.4.0-37 (Ubuntu 20.04)
43 5.4.0 (vanilla kernel.org)
44 * Recommended e2fsprogs version: 1.45.6.wc1 or newer
45 * Recommended ZFS version: 0.8.4
46 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
47 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
48 stack overflow. For more information, please refer to bugzilla 17630.
49 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
50 the re-exporting nodes to be mounted with "32bitapi" mount option
52 --------------------------------------------------------------------------------
56 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
57 for currently supported client and server kernel versions.
58 * Server primary kernels built and tested during release cycle:
59 3.10.0-1062.1.1.el7 (RHEL7.7)
60 * Other server kernels known to build and work at some point (others may also work):
61 3.10.0-862.14.4.el7 (RHEL7.5)
62 3.10.0-957.27.2.el7 (RHEL7.6)
63 4.18.0-80.7.1.el8 (RHEL8.0)
64 4.4.120-92.70 (SLES12 SP2)
65 4.4.180-94.100 (SLES12 SP3)
66 4.4.0-131 (Ubuntu 16.04)
67 4.15.0-32 (Ubuntu 18.04)
68 vanilla linux 4.6.7 (ZFS only)
69 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
70 * Client primary unpatched kernels built and tested during release cycle:
71 3.10.0-1062.1.1.el7 (RHEL7.7)
72 4.12.14-95.29 (SLES12 SP4)
73 4.15.0-48 (Ubuntu 18.04)
74 * Other clients known to build on unpatched kernels at some point (others may also work):
75 3.10.0-862.14.4.el7 (RHEL7.5)
76 3.10.0-957.27.2.el7 (RHEL7.6)
77 4.14.0-49.13.1.el7a (RHEL7.5)
78 4.14.0-115.2.2.el7a (RHEL7.6)
79 4.18.0-80.7.1.el8 (RHEL8.0)
80 4.4.120-92.70 (SLES12 SP2)
81 4.4.180-94.100 (SLES12 SP3)
82 4.4.0-131 (Ubuntu 16.04)
83 vanilla linux 4.15.0 (ZFS only)
84 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
85 * Recommended ZFS / SPL version: 0.7.13
86 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
87 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
88 stack overflow. For more information, please refer to bugzilla 17630.
89 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
90 the re-exporting nodes to be mounted with "32bitapi" mount option
91 * Note for Ubuntu 14.04.5:
92 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
93 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
95 --------------------------------------------------------------------------------
99 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
100 for currently supported client and server kernel versions.
101 * Server known to build on the following kernels (others may also work):
102 2.6.32-696.18.7.el6 (RHEL6.9)
103 3.10.0-862.14.4.el7 (RHEL7.5)
104 3.10.0-957.el7 (RHEL7.6)
105 4.4.120-92.70 (SLES12 SP2)
106 4.4.162-94.69 (SLES12 SP3)
107 4.4.0-131 (Ubuntu 16.04)
108 4.15.0-32 (Ubuntu 18.04)
109 vanilla linux 4.6.7 (ZFS only)
110 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
111 * Client known to build on unpatched kernels (others may also work):
112 2.6.32-696.18.7.el6 (RHEL6.9)
113 2.6.32-754.6.3.el6 (RHEL6.10)
114 3.10.0-862.14.4.el7 (RHEL7.5)
115 3.10.0-957.el7 (RHEL7.6)
116 4.14.0-49.13.1.el7a (RHEL7.5)
117 4.14.0-115.2.2.el7a (RHEL7.6)
118 4.4.120-92.70 (SLES12 SP2)
119 4.4.162-94.69 (SLES12 SP3)
120 4.4.0-131 (Ubuntu 16.04)
121 4.15.0-32 (Ubuntu 18.04)
122 vanilla linux 4.15.0 (ZFS only)
123 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
124 * Recommended ZFS / SPL version: 0.7.9
125 * Tested with ZFS / SPL version: 0.7.9
126 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
127 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
128 stack overflow. For more information, please refer to bugzilla 17630.
129 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
130 the re-exporting nodes to be mounted with "32bitapi" mount option
131 * Note for Ubuntu 14.04.5:
132 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
133 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
135 --------------------------------------------------------------------------------
137 2018-04-03 Intel Corporation
139 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
140 for currently supported client and server kernel versions.
141 * Server known to build on patched kernels:
142 2.6.32-431.29.2.el6 (RHEL6.5)
143 2.6.32-504.30.3.el6 (RHEL6.6)
144 2.6.32-573.26.1.el6 (RHEL6.7)
145 2.6.32-642.15.1.el6 (RHEL6.8)
146 2.6.32-696.18.7.el6 (RHEL6.9)
147 3.10.0-693.21.1.el7 (RHEL7.4)
148 3.0.101-0.47.71 (SLES11 SP3)
149 3.0.101-107 (SLES11 SP4)
150 3.12.74-60.64.40 (SLES12 SP1)
151 4.4.114-92.64 (SLES12 SP2)
152 4.4.114-94.11 (SLES12 SP3)
153 3.13.0-101 (Ubuntu 14.04, ZFS only)
154 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
155 4.4.0-116 (Ubuntu 16.04)
156 vanilla linux 4.6.7 (ZFS only)
157 * Client known to build on unpatched kernels:
158 2.6.32-431.29.2.el6 (RHEL6.5)
159 2.6.32-504.30.3.el6 (RHEL6.6)
160 2.6.32-573.26.1.el6 (RHEL6.7)
161 2.6.32-642.15.1.el6 (RHEL6.8)
162 2.6.32-696.18.7.el6 (RHEL6.9)
163 3.10.0-693.21.1.el7 (RHEL7.4)
164 3.0.101-0.47.71 (SLES11 SP3)
165 3.0.101-107 (SLES11 SP4)
166 3.12.74-60.64.40 (SLES12 SP1)
167 4.4.114-92.64 (SLES12 SP2)
168 4.4.114-94.11 (SLES12 SP3)
169 3.13.0-101 (Ubuntu 14.04)
170 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
171 4.4.0-116 (Ubuntu 16.04)
173 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
174 * Recommended ZFS / SPL version: 0.7.6
175 * Tested with ZFS / SPL version: 0.7.6
176 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
177 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
178 stack overflow. For more information, please refer to bugzilla 17630.
179 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
180 the re-exporting nodes to be mounted with "32bitapi" mount option
181 * Note for Ubuntu 14.04.5:
182 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
183 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
185 --------------------------------------------------------------------------------
187 2017-07-05 Intel Corporation
189 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
190 for currently supported client and server kernel versions.
191 * Server known to build on patched kernels:
192 2.6.32-431.29.2.el6 (RHEL6.5)
193 2.6.32-504.30.3.el6 (RHEL6.6)
194 2.6.32-573.26.1.el6 (RHEL6.7)
195 2.6.32-642.15.1.el6 (RHEL6.8)
196 2.6.32-696.3.1.el6 (RHEL6.9)
197 3.10.0-514.21.1.el7 (RHEL7.3)
198 3.0.101-0.47.71 (SLES11 SP3)
199 3.0.101-100 (SLES11 SP4)
200 3.12.74-60.64.40 (SLES12 SP1)
201 4.4.59-92.17 (SLES12 SP2)
202 vanilla linux 4.6.7 (ZFS only)
203 * Client known to build on unpatched kernels:
204 2.6.32-431.29.2.el6 (RHEL6.5)
205 2.6.32-504.30.3.el6 (RHEL6.6)
206 2.6.32-573.26.1.el6 (RHEL6.7)
207 2.6.32-642.15.1.el6 (RHEL6.8)
208 2.6.32-696.3.1.el6 (RHEL6.9)
209 3.10.0-514.21.1.el7 (RHEL7.3)
210 3.0.101-0.47.71 (SLES11 SP3)
211 3.0.101-100 (SLES11 SP4)
212 3.12.74-60.64.40 (SLES12 SP1)
213 4.4.59-92.17 (SLES12 SP2)
215 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
216 * Recommended ZFS / SPL version: 0.6.5.9
217 * Tested with ZFS / SPL version: 0.6.5.9
218 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
219 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
220 stack overflow. For more information, please refer to bugzilla 17630.
221 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
222 the re-exporting nodes to be mounted with "32bitapi" mount option
224 --------------------------------------------------------------------------------
226 12-07-2016 Intel Corporation
228 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
229 for currently supported client and server kernel versions.
230 * Server known to build on patched kernels:
231 2.6.32-431.29.2.el6 (RHEL6.5)
232 2.6.32-504.30.3.el6 (RHEL6.6)
233 2.6.32-573.26.1.el6 (RHEL6.7)
234 2.6.32-642.11.1.el6 (RHEL6.8)
235 3.10.0-514.el7 (RHEL7.3)
236 3.0.101-0.47.71 (SLES11 SP3)
237 3.0.101-80 (SLES11 SP4)
238 3.12.62-60.64.8 (SLES12 SP1)
239 vanilla linux 4.5.7 (ZFS only)
240 * Client known to build on unpatched kernels:
241 2.6.32-431.29.2.el6 (RHEL6.5)
242 2.6.32-504.30.3.el6 (RHEL6.6)
243 2.6.32-573.26.1.el6 (RHEL6.7)
244 2.6.32-642.11.1.el6 (RHEL6.8)
245 3.10.0-514.el7 (RHEL7.3)
246 3.0.101-0.47.71 (SLES11 SP3)
247 3.0.101-80 (SLES11 SP4)
248 3.12.62-60.64.8 (SLES12 SP1)
250 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
251 * Recommended ZFS / SPL version: 0.6.5.8
252 * Tested with ZFS / SPL version: 0.6.5.8
253 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
254 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
255 stack overflow. For more information, please refer to bugzilla 17630.
256 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
257 the re-exporting nodes to be mounted with "32bitapi" mount option
259 --------------------------------------------------------------------------------
261 02-29-2016 Intel Corporation
263 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
264 for currently supported client and server kernel versions.
265 * Server known to build on patched kernels:
266 2.6.32-431.29.2.el6 (RHEL6.5)
267 2.6.32-504.30.3.el6 (RHEL6.6)
268 2.6.32-573.12.1.el6 (RHEL6.7)
269 3.10.0-327.3.1.el7 (RHEL7.2)
270 3.0.101-0.47.71 (SLES11 SP3)
271 3.0.101-68 (SLES11 SP4)
272 vanilla linux 4.2.1 (ZFS only)
273 * Client known to build on unpatched kernels:
274 2.6.32-431.29.2.el6 (RHEL6.5)
275 2.6.32-504.30.3.el6 (RHEL6.6)
276 2.6.32-573.12.1.el6 (RHEL6.7)
277 3.10.0-327.3.1.el7 (RHEL7.2)
278 3.0.101-0.47.71 (SLES11 SP3)
279 3.0.101-68 (SLES11 SP4)
282 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
283 * Recommended ZFS / SPL version: 0.6.4.2
284 * Tested with ZFS / SPL version: 0.6.4.2
285 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
286 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
287 stack overflow. For more information, please refer to bugzilla 17630.
288 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
289 the re-exporting nodes to be mounted with "32bitapi" mount option
291 --------------------------------------------------------------------------------
293 03-10-2015 Intel Corporation
295 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
296 for currently supported client and server kernel versions.
297 * Server known to build on patched kernels:
298 2.6.32-431.29.2.el6 (RHEL6.5)
299 2.6.32-504.8.1.el6 (RHEL6.6)
300 3.0.101-0.46 (SLES11 SP3)
301 * Client known to build on unpatched kernels:
302 2.6.32-431.29.2.el6 (RHEL6.5)
303 2.6.32-504.8.1.el6 (RHEL6.6)
304 3.10.0-123.20.1.el7 (RHEL7)
305 3.0.101-0.46 (SLES11 SP3)
306 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
307 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
308 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
309 stack overflow. For more information, please refer to bugzilla 17630.
310 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
311 the re-exporting nodes to be mounted with "32bitapi" mount option
313 Severity : enhancement
315 Description: control OST-index in IDIF via ROCOMPAT flag.
316 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
317 last_rcvd file. For new formatted OST device, it will be auto set;
318 for the case of upgrading from old OST device, you can enable it
319 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
320 enabled, for new created OST-object, its IDIF-in-LMA will contain
321 the OST-index; for the existing OST-object, the OSD will convert
322 old format IDIF as new format IDIF with OST-index stored in the
323 LMA EA when accessing such OST-object or via OI scrub. Once such
324 flag is enabled, it cannot be reverted back, so the system cannot
325 be downgraded to the orignal incompatible version.
327 --------------------------------------------------------------------------------
329 07-30-2014 Intel Corporation
331 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
332 for currently supported client and server kernel versions.
333 * Server known to build on patched kernels:
334 2.6.32-431.20.3.el6 (RHEL6.5)
335 3.0.101-0.31 (SLES11 SP3)
336 * Client known to build on unpatched kernels:
337 2.6.32-431.20.3.el6 (RHEL6.5)
338 3.0.101-0.31 (SLES11 SP3)
341 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
342 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
343 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
344 stack overflow. For more information, please refer to bugzilla 17630.
345 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
346 the re-exporting nodes to be mounted with "32bitapi" mount option
348 --------------------------------------------------------------------------------
350 10-31-2013 Intel Corporation
352 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
353 * Server builds for kernels:
354 2.6.32-358.18.1.el6 (RHEL6.4)
355 * Client builds for unpatched kernels:
356 2.6.32-358.18.1.el6 (RHEL6.4)
357 3.0.92-0.8 (SLES11 SP3)
359 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
360 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
361 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
362 stack overflow. For more information, please refer to bugzilla 17630.
363 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
364 the re-exporting nodes to be mounted with "32bitapi" mount option
366 --------------------------------------------------------------------------------
368 05-30-2013 Intel Corporation
370 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
371 * Server builds for kernels:
372 2.6.32-358.6.2.el6 (RHEL6.4)
373 * Client builds for unpatched kernels:
374 2.6.32-358.6.2.el6 (RHEL6.4)
375 3.0.74-0.6.6 (SLES11 SP2)
377 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
378 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
379 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
380 stack overflow. For more information, please refer to bugzilla 17630.
381 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
382 the re-exporting nodes to be mounted with "32bitapi" mount option
384 --------------------------------------------------------------------------------
386 9-30-2012 Whamcloud, Inc.
388 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
389 * Server support for kernels:
390 2.6.32-279.5.1.el6 (RHEL6)
391 * Client support for unpatched kernels:
392 2.6.32-279.5.1.el6 (RHEL6)
394 2.6.18-238.19.1.el5 (RHEL5)
395 2.6.16.60-0.69.1 (SLES10)
396 * Recommended e2fsprogs version: 1.41.90.wc4
397 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
398 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
399 stack overflow. For more information, please refer to bugzilla 17630.
400 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
401 the re-exporting nodes to be mounted with "32bitapi" mount option
403 Severity : enhancement
405 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
407 Severity : enhancement
409 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
411 Severity : enhancement
413 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
415 Severity : enhancement
417 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
419 Severity : enhancement
421 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
423 --------------------------------------------------------------------------------
425 04-30-2012 Whamcloud, Inc.
427 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
428 * Server support for kernels:
429 2.6.32-220.4.2.el6 (RHEL6)
430 * Client support for unpatched kernels:
431 2.6.32-220.4.2.el6 (RHEL6)
433 2.6.18-238.19.1.el5 (RHEL5)
434 2.6.16.60-0.69.1 (SLES10)
435 * Recommended e2fsprogs version: 1.41.90.wc4
436 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
437 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
438 stack overflow. For more information, please refer to bugzilla 17630.
439 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
440 the re-exporting nodes to be mounted with "32bitapi" mount option
442 Severity : enhancement
444 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
446 Severity : enhancement
448 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
450 Severity : enhancement
452 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
454 Severity : enhancement
456 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
458 --------------------------------------------------------------------------------
460 09-30-2011 Whamcloud, Inc.
462 * Server support for kernels:
463 2.6.18-238.19.1.el5 (RHEL5)
464 2.6.32-131.6.1.el6 (RHEL6)
465 * Client support for unpatched kernels:
466 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
467 2.6.18-238.19.1.el5 (RHEL5)
468 2.6.32-131.6.1.el6 (RHEL6)
469 2.6.29.4-167.fc11 (FC11)
470 2.6.16.60-0.69.1 (SLES10)
472 * Recommended e2fsprogs version: 1.41.90.wc3
473 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
474 Lustre filesystem with 4K stack may cause a stack overflow. For more
475 information, please refer to bugzilla 17630.
476 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
477 mounted with "32bitapi" mount option
479 Severity : enhancement
480 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
482 Severity : enhancement
483 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
485 Severity : enhancement
486 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
489 Frequency : only with SLES10
491 Description: Use OFED "KMP" provided by Novell
492 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
493 Lustre is now built against this package. That means you need to
494 install the ofed-kmp package from Novell for the patchless client.
495 Note that the ofed-kmp that Novell ships may not exactly match the
496 kernel version but should still be compatible.
498 Severity : enhancement
500 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
501 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
503 --------------------------------------------------------------------------------
505 2010-07-15 Oracle, Inc.
507 * Server support for kernels:
508 2.6.18-164.11.1.el5 (RHEL 5)
509 2.6.18-164.11.1.0.1.el5 (OEL 5)
510 * Client support for unpatched kernels:
511 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
512 2.6.18-164.11.1.el5 (RHEL 5),
513 2.6.18-164.11.1.0.1.el5 (OEL 5)
514 2.6.16.60-0.42.8 (SLES 10),
516 2.6.29.4-167.fc11 (FC11)
517 * Recommended e2fsprogs version: 1.41.10-sun2
519 Severity : enhancement
521 Description: Distribute lfs_migrate script with lustre
522 Details : The lfs_migrate mentioned in the manual is now included in the
523 lustre tree in lustre/scripts and will distributed in lustre RPMs.
525 Severity : enhancement
527 Description: Failover nid list added to import proc file.
529 Severity : enhancement
531 Description: "som_preview" mount option required to use SOM
532 Details : The clients will need to use som_preview mount option
536 Description: Lustre does not do 1MB IOs to HW RAID
537 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
539 Severity : enhancement
541 Description: lfs quota output is not very convenient for awk/sed-parsing
542 Details : Some positions in lfs quota output table could be empty or
543 non-empty which made it hard to parse it with scripts, now
544 a dash is put instead of space where there is not supposed
549 Description: Update to OFED 1.5.1
550 Details : OFED 1.5.1 will be used in environments that don't use
551 vendor supplied OFED.
555 Description: Remove lnet. prefix from lctl params display
556 Details : Also renamed lnet.memused to lnet_memused to
561 Description: Use conf_param -d to remove permanant settings.
565 Description: Add list_param -R
566 Details : lctl list_param now supports recursive mode.
569 Bugzilla : 21829, 22312
570 Description: Fix llobdstat.
571 Details : 1. Now updating the write_bytes counter
572 2. Port script fixes from 1.8.
576 Description: decrease the usage of memory on clients.
577 Details : 1. On clients, recycle dentries and inodes unused.
578 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
583 Description: llobdstat fix and enhancement
584 Details : add a counter to set a limit to how many samples will be returned
585 fix a wildcard in the path to limit to obdfilter stats only
589 Description: lfs check only allowed for root
590 Details : Code cleanups and added a regresion test of non-root use of
595 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
596 Details : Fix hr->hr_index overload.
600 Description: Fix FIEMAP functionality.
604 Description: FC11 patchless client support added
608 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
609 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
613 Description: New acc-sm script and various test-framework enhancments.
617 Description: Previously landed patch was reverted.
621 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
625 Description: Add ldiskfs maxdirsize mount option.
629 Description: Fix mkfs.lustre support for 16TB LUNs.
633 Description: Print mmp_check_interval
634 Details : print mmp_check_interval and make it possible to abort
635 mount operation in case it takes too long.
639 Description: Add cascading_rw to lustre mpi tests.
643 Description: Store SOM data in EA.
644 Details : Also several SOM related cleanups.
648 Description: Add refcount to md_open_data
652 Description: Use refcount on lov_request_set to prevent memory corruption.
656 Description: Statahead: use parent dir's i_mutex for synchronization.
660 Description: loadgen improvements.
664 Description: Enable peer health detection in ksocklnd and ko2iblnd.
668 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
672 Description: Fix race on obd_devcie:md_stats
676 Description: Fix SLES 11 client builds.
677 Details : Resolved uninitalized pointer gcc warnings.
681 Description: Allow non-root access for "lfs check".
682 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
686 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
688 Severity : enhancement
690 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
691 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
695 Description: Sanity test_27p fails.
696 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
697 Sanity tests 27 were cleaned up and improved.
701 Description: Unify req format on client/servers
702 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
703 buffers and swabbers approach..
708 Description: drop join file stuff.
710 Severity : enhancement
712 Description: Add 2.6.27 and SLES11 (patchless client) support
716 Description: Fix in ptlrpc_expire_one_request() to print the signed time
719 Severity : enhancement
721 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
724 Severity : enhancement
726 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
728 Severity : enhancement
730 Description: correctly handle big reply message.
731 Details : send LNet event if reply is bigger than buffer and adjust this buffer
736 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
738 Severity : enhancement
740 Description: Add support for OEL5.
744 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
746 Severity : enhancement
748 Description: "weak-modules" support
749 Details : Implement "weak-modules" support which enables kernel modules
750 to be used with any kernel that implements the same kABI. In
751 order to achieve this modules are now installed in
752 /lib/modules/$(uname -r)/updates/kernel on all distributions.
756 Description: client could not reconnect to OST because of an active request.
757 Details : The client now retries to reconnect to the same server, if a
758 connect request failed with EBUSY or -EAGAIN. Test case 35b
759 conf-sanity.sh has been created.
763 Description: Don't allow make backward step on assignin osc next id.
764 Details : race between allocation next id and ll_sync thread can be cause
765 of set wrong osc next id and can be kill valid ost objects.
769 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
773 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
777 Description: Error handling in osc_statfs_interpret() has been improved.
778 Details : Check in osc_statfs_interpret() for EBADR.
782 Description: Do not send statfs() requests to OSTs disabled by administrator.
783 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
785 Severity : enhancement
787 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
789 Severity : enhancement
791 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
793 Severity : enhancement
795 Description: Enhancement for lfs(1) command to use numeric uid/gid.
799 Description: Improved cleanup procedure for conf-sanity test_46a.
803 Description: Do not start bulk transfer, if client tried to reconnect.
804 Details : Do not start bulk transfer, if client tried to reconnect,
805 but failed due to active rpcs. Check for exp_abort_active_req
806 has been implemented.
810 Description: Prevent inconsistences between linux and lustre mount structures.
811 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
812 Make the sleep interruptible.
814 Severity : enhancement
816 Description: provide server to client comms path
817 Details : server to client communications path via new LDLM_SET_INFO rpc
821 Description: truncate starts GFP_FS allocation under transaction causing deadlock
822 Details : ldiskfs_truncate calls grab_cache_page which may start page
823 allocation under an open transaction. This may lead to
824 calling prune_icache with consequent lustre reentrance.
828 Description: send correctly lsm on open replay
829 Details : MDS is trust to LSM size on replay open, but client can set wrong size
834 Description: enable client interface failover
835 Details : When a child reconnects from another NID, properly update export
836 nid hash position and ldlm reverse import.
838 Severity : enhancement
840 Description: Communicate OST degraded/readonly state via statfs to MDS
841 Details : Flags in the statfs returned from OSTs indicate whether the
842 OST is in a degraded RAID state, or if the filesystem has
843 turned read-only after a filesystem error is detected.
848 Description: don't panic if EPROTO was hit when reading symlink
849 Details : correctly handling request reference in error cases.
851 Severity : enhancement
853 Description: Add LustreNetLink, a kernel-userspace communcation path.
855 Severity : enhancement
857 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
862 Description: access to llog context before init.
863 Details : move handling CATALOGS file at osc layer and forbid access to llog
867 Frequency : rare, only when client is writing to OST on local node
869 Description: Avoid deadlock for local client writes
870 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
871 memory freeing context. This allows OST threads to set the
872 PF_MEMALLOC flag on task structures in order to allocate memory
873 from reserved pools and complete IO.
874 Use GFP_HIGHUSER for OST allocations for non-local client writes,
875 so that the OST threads generate memory pressure and allow
876 inactive pages to be reclaimed.
878 Severity : enhancement
880 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
885 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
886 Details : move ldlm namespace creation in setup phase to avoid grab
887 _lprocfs_lock with cli_sem held.
891 Description: Temporarily disable grant shrink.
892 Details : Disable the feature for debugging.
895 Frequency : only when formatting test filesystems
897 Description: Unable to run several mkfs.lustre on loop devices at the same time
898 Details : mkfs.lustre returns error 256 on the concurrent loop devices
899 formatting. The solution is to proper handle the error.
901 Severity : enhancement
903 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
905 Severity : enhancement
907 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
912 Description: resolve race between obd_disconnect and class_disconnect_exports
913 Details : if obd_disconnect will be called to already disconnected export he
914 forget release one reference and osc module can't unloaded.
916 Severity : enhancement
918 Description: Allow tuning service thread via /proc
919 Details : For each service a new
920 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
921 created that can be used to set min/max thread counts, and get the
922 current number of running threads.
927 Descriptoin: don't return error if have partially created objects for file.
928 Details : lov_update_create_set uses set->set_success as index for created
929 objects, so if some requests will be failed, they will have hole
930 at end of array and qos_shrink_lsm for allocate correct lsm.
932 Severity : enhancement
934 Description: Update OFED support to 1.4.1
939 Description: Reduce small size read RPC
940 Details : Set read-ahead limit for every file and only do read-ahead when
941 available read-ahead pages are bigger than 1M to avoid small size
944 Severity : enhancement
946 Description: move AT tunable parameters for more consistent usage
947 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
949 Severity : enhancement
951 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
952 Details : allow skip disconnected ost for send statfs request and hide error
958 Description: fix racy locking of mballoc block bitmaps causing BUG
959 Details : The locking of the mballoc buddy bitmap and the in-memory
960 block bitmap was using two different spin locks in some
961 cases. This made it possible to incorrectly access the
962 mballoc bitmap while another process was modifying it,
963 causing a sanity assertion to fail. While no on-disk corruption
964 was reported, there was some risk of this happening.
967 Frequency : rare, on llog test 6
969 Descriptoin: don't allow connect to already connected import
970 Details : allowing connect to already connected import is hide connecting problem.
975 Descriptoin: allow kill process which wait statahead result
976 Details : for some reasons 'ls' can stick in waiting result from statahead,
977 in this case need way for kill this process.
979 Severity : enhancement
981 Description: Add state history info file, enhance import info file
982 Details : Track import connection state changes in a new osc/mdc proc file;
983 add overview-type data to the osc/mdc import proc file.
985 Severity : enhancement
987 Description: MDS create should not wait for statfs RPC while holding DLM lock.
989 Severity : enhancement
991 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
996 Descriptoin: OST grant too much space to client even there are not enough space.
997 Details : Client will shrink its grant cache to OST if there are no write
998 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
999 this grant cache if there are already not enough avaible space
1000 (left_space < total_clients * 32M).
1003 Frequency : start MDS on uncleanly shutdowned MDS device
1005 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1006 Details : stay in waiting mds<>ost recovery finished produce random bugs
1007 due race between two ll_sync thread for one lov target. send
1008 ACTIVATE event only if connect realy finished and import have
1012 Frequency : rare, connect and disconnect target at same time
1014 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1015 Details : don't call obd_disconnect under lov_lock. this long time
1016 operation and can block ptlrpcd which answer to connect request.
1021 Descriptoin: don't lose wakeup for imp_recovery_waitq
1022 Details : recover_import_no_retry or invalidate_import and import_close can
1023 both sleep on imp_recovery_waitq, but we was send only one wakeup
1027 Frequency : always with long access acl
1029 Descriptoin: mds can't pack reply with long acl.
1030 Details : mds don't control size of acl but they limited by reint/getattr
1033 Severity : enhancement
1035 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1037 Severity : enhancement
1039 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1042 Frequency : start MDS on uncleanly shutdowned MDS device
1044 Descriptoin: aborting recovery hang on MDS
1045 Details : don't throttle destroy RPCs for the MDT.
1048 Frequency : on remount
1050 Description: external journal device not working after the remount
1051 Details : clear dev_rdonly flag for external journal devices in
1057 Description: shutdown vs evict race
1058 Details : client_disconnect_export vs connect request race.
1059 if client will evicted at this time - we start invalidate
1060 thread without referece to import and import can be freed
1064 Frequency : rare, need acl's on inode.
1066 Description: client can't handle ost additional correctly
1067 Details : if ost was added after client connected to mds client can have
1068 hit lnet_try_match_md ... to big messages to wide striped files.
1069 in this case need teach client to handle config events about add
1070 lov target and update client max ea size at that event.
1072 Severity : enhancement
1074 Description: Changelogs
1075 Details : Changelogs are a lightweight mechanism to track filesystem
1076 metadata and namespace changes. The changelog is recorded
1077 permanently on the MDTs, and is periodically "consumed" / purged
1078 when records are no longer needed.
1080 Severity : enhancement
1082 Description: compact fld format with extents
1083 Details : Store range of seq rather than every seq in FLD. Seq
1084 controller update FLD rather than clients. In Case of CMD, mdt0
1085 has FLD, all other metadata server act as non persistent proxy
1086 for FLD queries and cache fld entries in fld cache.
1091 Description: don't skip ost target if they assigned to file
1092 Details : Drop slow OSCs if we can, but not for requested start idx.
1093 This means "if OSC is slow and it is not the requested
1094 start OST, then it can be skipped, otherwise skip it only
1095 if it is inactive/recovering/out-of-space.
1099 Description: more cleanup in mds_lov
1100 Details : not send LOV EA under replay, we can't know about they size at this
1101 time. Don't allow client connect to mds before any ost connected,
1102 for avoid problems with LOV EA size and returning EIO to client.
1104 Severity : enhancement
1106 Description: Interoperability at server side (Disk interoperability)
1108 Severity : enhancement
1110 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1112 Severity : enhancement
1114 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1116 Severity : enhancement
1118 Description: New client IO stack (CLIO).
1120 Severity : enhancement
1122 Description: Commit on sharing. Eliminate inter-client dependencies between
1123 uncommitted transactions by doing transaction commits.
1124 Thereby clients may recovery independently.
1127 Frequency : Create a symlink file with a very long name
1129 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1130 Details : If there is no extra space in the request for early cancels,
1131 ldlm_req_handles_avail() returns 0 instead of a negative value.
1133 Severity : enhancement
1135 Description: Add /proc entry for import status
1136 Details : The mdc, osc, and mgc import directories now have
1137 an import directory that contains useful import data for debugging
1138 connection problems.
1140 Severity : enhancement
1142 Description: Re-disable certain /proc logging
1143 Details : Enable and disable client's offset_stats, extents_stats and
1144 extents_stats_per_process stats logging on the fly.
1147 Frequency : Only on FC kernels 2.6.22+
1149 Description: oops in statahead
1150 Details : Do not drop reference count for the dentry from VFS when lookup,
1151 VFS will do that by itself.
1153 Severity : enhancement
1155 Description: Generic /proc file permissions
1156 Details : Set /Proc file permissions in a more generic way to enable non-
1157 root users operate on some /proc files.
1161 Description: Hitting mdc_commit_close() ASSERTION
1162 Details : Properly handle request reference release in
1163 ll_release_openhandle().
1167 Description: quota recovery deadlock during mds failover
1168 Details : This patch includes att18982, att18236, att18237 in bz14840.
1170 1. fix osts hang when mds does failover with quotaon
1171 2. prevent watchdog storm when osts threads wait for the
1176 Frequency : only patchless client
1177 Description: add workaround for race between add/remove dentry from hash
1179 Severity : enhancement
1181 Description: Allow OST glimpses to return PW locks
1185 Description: LBUG when llog conf file is full
1186 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1190 Description: Prevent import from entering FULL state when server in recovery
1194 Description: service mount cannot take device name with ":"
1195 Details : Only when device name contains ":/" will mount treat it as
1201 Description: replace ptlrpcd with the statahead thread to interpret the async
1202 statahead RPC callback
1206 Frequency : on recovery
1207 Description: I/O failures after umount during fail back
1208 Details : if client reconnected to restarted server we need join to recovery
1209 instead of find server handler is changed and process self
1210 eviction with cancel all locks.
1212 Severity : enhancement
1214 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1216 Severity : enhancement
1218 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1220 Severity : enhancement
1222 Description: Upcall on Lustre log has been dumped
1223 Details : Allow for a user mode script to be called once a Lustre log has
1224 been dumped. It passes the filename of the dumped log to the
1225 script, the location of the script can be specified via
1226 /proc/sys/lnet/debug_log_upcall.
1231 Description: avoid idr_remove called for id which is not allocated.
1232 Details : Move assigment s_dev for clustered nfs to end of initialization,
1233 to avoid problem with error handling.
1238 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1239 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1240 into CONN_USED_HASH and this prodice warning when put connection
1241 again in unused hash.
1244 Severity : enhancement
1246 Description: File striping can now be set to use an arbitrary pool of OSTs.
1248 Severity : enhancement
1250 Description: Export bytes_read/bytes_write count on OSC/OST.
1254 Description: Early reply size mismatch, MGC loses connection
1255 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1256 the connect flags are properly negotiated.
1261 Description: quotas are not honored with O_DIRECT
1262 Details : all writes with the flag O_DIRECT will use grants which leads to
1263 this problem. Now using OBD_BRW_SYNC to guard this.
1267 Description: add quota statistics
1268 Details : 1. sort out quota proc entries and proc code.
1269 2. add quota statistics
1271 Severity : enhancement
1273 Description: enable quota support for HEAD.
1277 Description: Properly propagate oinfo flags from lov to osc for statfs
1278 Details : restore missing copy oi_flags to lov requests.
1280 Severity : enhancement
1282 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1283 : l_getgroups(8), lst(8), routerstat(8)
1284 Details : included man pages for llobdstat(8), llstat(8),
1285 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1287 Severity : enhancement
1289 Description: configure's --enable-quota should check the
1290 : kernel .config for CONFIG_QUOTA
1291 Details : configure is terminated if --enable-quota is passed but
1292 : no quota support is in kernel
1296 Description: Remove portals compatibility
1297 Details : Remove portals compatibility, not interoperable with releases
1302 Description: Resolve device initialization race
1303 Details : Prevent proc handler from accessing devices added to the
1304 obd_devs array but yet be intialized.
1306 Severity : enhancement
1308 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1310 Severity : enhancement
1312 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1317 Description: Using wrong pointer in osc_brw_prep_request
1318 Details : Access to array[-1] can produce panic if kernel compiled with
1319 CONFIG_PAGE_ALLOC enabled
1323 Description: Client runs out of low memory
1324 Details : Consider only lowmem when counting initial number of llap pages
1328 Description: *optional* service tags registration
1329 Details : if the "service tags" package is installed on a Lustre node
1330 When the filesystem is mounted, a local-node service tag will
1331 be created. See http://inventory.sun.com/ for more information
1332 about the Service Tags asset management system.
1336 Description: Kernel BUG tries to release flock
1337 Details : Lustre does not destroy flock lock before last reference goes
1338 away. So always drop flock locks when client is evicted and
1339 perform unlock regardless of successfulness of speaking to MDS.
1343 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1347 Description: sanity test 65a fails if stripecount of -1 is set
1348 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1353 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1354 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1357 Severity : enhancement
1359 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1363 Description: do not process already freed flock
1364 Details : flock can possibly be freed by another thread before it reaches
1365 to ldlm_flock_completion_ast.
1369 Description: LBUG during stress test
1370 Details : Need properly lock accesses the flock deadlock detection list.
1374 Description: oops in page fault handler
1375 Details : kernel page fault handler can return two special 'pages' in error
1376 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1380 Description: timeout with invalidate import.
1381 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1382 should be handled by ptlrpcd. This produce long age waiting and
1383 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1385 Severity : enhancement
1387 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1391 Description: enable MGS and MDT services start separately
1392 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1393 a co-located MGS without starting the MGS, which is a complement
1394 to 'nosvc' mount option.
1398 Frequency : after recovery
1399 Description: precreate to many object's after del orphan.
1400 Details : del orphan st in oscc last_id == next_id and this triger growing
1401 count of precreated objects. Set flag LOW to skip increase count
1402 of precreated objects.
1406 Frequency : rare, on clear nid stats
1407 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1408 Details : when clean nid stats sometimes try destroy live entry,
1409 and this produce panic in free.
1413 Description: Stack overflow during MDS log replay
1414 ease stack pressure by using a thread dealing llog_process.
1418 Description: wait until IO finished before start new when do lock cancel.
1419 Details : VM protocol want old IO finished before start new, in this case
1420 need wait until PG_writeback is cleared until check dirty flag and
1421 call writepages in lock cancel callback.
1423 Severity : enhancement
1425 Description: using special macro for print time and cleanup in includes.
1429 Description: mds_mfd_close() ASSERTION(rc == 0)
1430 Details : In mds_mfd_close(), we need protect inode's writecount change
1431 within its orphan write semaphore to prevent possible races.
1435 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1439 Frequency : rare, on shutdown ost
1440 Description: don't hit live lock with umount ost.
1441 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1442 use shrink_dcache_sb instead.
1446 Description: don't panic with use echo client
1447 Details : echo client pass NULL as client nid pointer and this produce null
1448 pointer dereference.
1452 Description: fix build on ppc32
1453 Details : compile code with -m64 flag produce wrong object file for ppc32.
1457 Description: add message levels for liblustreapi
1461 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1462 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1463 will result in an -EAGAIN and in the exit call path we call
1464 alloc_rr() which will with increasing aggressiveness attempt to
1465 aquire precreated objects on the minimum number of required OSCs.
1469 Description: Use old size assignment to avoid deadlock
1470 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1471 the scheduling while holding a spinlock. We do not need locking
1472 for size in ll_update_inode() because size is only updated from
1473 the MDS for directories or files without objects, so there is no
1474 other place to do the update, and concurrent access to such inodes
1475 are protected by the inode lock.
1479 Description: resolve "_IOWR redefined" build error on SLES10
1483 Description: dump the memory debugging after all modules are unloaded to
1484 suppress false negative in conf_sanity test 39
1486 Severity : enhancement
1488 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1493 Description: fix SLES kernel versioning
1494 Details : the kernel version for our SLES 10 kernel did not include a "-"
1495 before the "smp" at the end. while this was not a problem in
1496 general, it did mean that software trying to use the kernel
1497 version to try to detect a vendor specific kernel would fail.
1498 this was most evident by the OFED build scripts.
1502 Description: Don't update lov_desc members until making sure they are valid
1503 Details : When updating lov_desc members via proc fs, need fix their
1504 validities before doing the real update.
1508 Description: don't put request into delay list while invalidate in flight.
1509 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1510 import in flight. this produce timeout for invalidate and sometimes
1511 can cause stale data.
1515 Frequency : on ppc only
1516 Description: not convert ost objects for directory because it's not exist.
1517 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1521 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1523 Severity : enhancement
1525 Description: Update to support 2.6.22.14 vanilla kernel.
1529 Frequency : rare, on recovery
1530 Description: read procfs can produce deadlock in some situation
1531 Details : Holding lprocfs lock which send rpc can produce block for destroy
1532 obd objects and this also block reconnect with -EALREADY. This
1533 isn't fix all lprocfs bugs - but make it rare.
1535 Severity : enhancement
1537 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1540 Frequency : frequent on X2 node
1542 Description: mdc_set_open_replay_data LBUG
1543 Details : Set replay data for requests that are eligible for replay.
1547 Description: lustre_mgs: operation 101 on unconnected MGS
1548 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1549 MGC, and late on MGC cannot successfully connect to MGS and a lot
1550 of the error messages complaining that MGS is not connected.
1554 Frequency : on network error
1555 Description: panic with double free request if network error
1556 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1557 but it's true only for synchronus enqueue, for async enqueue
1558 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1561 Severity : enhancement
1563 Description: client-side metadata stat-ahead during readdir(directory readahead)
1564 Details : perform client-side metadata stat-ahead when the client detects
1565 readdir and sequential stat of dir entries therein
1568 Frequency : on start mds
1570 Description: Implement get_info(last_id) in obdfilter.
1573 Frequency : occasional
1575 Description: Correctly check stale fid, not start epoch if ost not support SOM
1576 Details : open with flag O_CREATE need set old fid in op_fid3 because
1577 op_fid2 was overwritten with new generated fid, but mds can answer
1578 with one of these two fids and both is not stale. Setattr
1579 incorrectly started an epoch and assume will be called
1580 done_writing, but without SOM done_writing ever being called.
1583 Frequency : rare, depends on device drivers and load
1585 Description: MDS or OSS nodes crash due to stack overflow
1586 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1587 In some cases, in conjunction with device drivers that use a lot
1588 of stack the MDS (or possibly OSS) service threads could overflow
1589 the stack. One change which was identified to consume additional
1590 stack has been reworked to avoid the extra stack usage.
1593 Frequency : occasional
1595 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1596 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1597 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1598 oscc_has_objects and see OSC has no objests and call
1599 oscc_internal_create to resend create request.
1601 Severity : enhancement
1603 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1605 Severity : enhancement
1607 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1610 Frequency : very rare
1612 Description: Fix replay if there is an un-replied request and open
1613 Details : In some cases, older replay request will revert the
1614 mcd->mcd_last_xid on MDS which is used to record the client's
1615 latest sent request.
1617 Severity : enhancement
1619 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1621 Severity : enhancement
1623 Description: Add rhel5 support to HEAD.
1625 Serverity : enhancement
1627 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1632 Description: Don't allow skipping OSTs if index has been specified.
1633 Details : Don't allow skipping OSTs if index has been specified, make
1634 locking in internal create lots better.
1638 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1639 Details : access to bitfield in structure is always rounded to long
1640 and this produce problem with not atomic change any bit.
1644 Description: Lustre make rpms failed.
1645 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1646 builds Lustre rpms from tarball.
1649 Frequency : rare on shutdown ost
1651 Description: If llog cancel was not send before clean_exports phase, this can
1652 produce deadlock in llog code.
1653 Details : If llog thread has last reference to obd and call class_import_put
1654 this produce deadlock because llog_cleanup_commit_master wait when
1655 last llog_commit_thread exited, but this never success because was
1656 called from llog_commit_thread.
1660 Description: allow userland application know is lost one of stripes.
1661 Details : fill lvb_blocks with error code on ost and return it to
1662 application if error flag found.
1666 Description: NULL lov_tgts causing MDS oops
1667 Details : more safe checks for NULL lov_tgts for avoid oops.
1669 Severity : enhancement
1671 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1675 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1676 Details : If an OST is down the MDS will hang indefinitely in
1677 obd_statfs_rqset() waiting for the statfs data. While for
1678 MDS QOS usage of statfs, it should not stuck in waiting.
1680 Severity : enhancement
1682 Description: remote_acl support
1683 Details : Support ACL-based permission check for remote user.
1684 Support setfacl/getfacl for remote user with the utils
1685 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1686 the system "{s,g}etfacl" utils.
1688 Severity : enhancement
1690 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1692 Severity : enhancement
1694 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1698 Description: make mgs_setparam() handle fsname containing dash
1699 Details : fsname containing a dash does not work with lctl conf_param
1701 Severity : enhancement
1703 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1705 Severity : enhancement
1707 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1711 Frequency : rare, at shutdown
1712 Description: access already free / zero obd_namespace.
1713 Details : if client_disconnect_export was called without force flag set,
1714 and exist connect request in flight, this can produce access to
1715 NULL pointer (or already free pointer) when connect_interpret
1716 store ocd flags in obd_namespace.
1720 Frequency : only at startup
1721 Description: not alloc memory with spinlock held.
1722 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1723 if any spinlock held.
1725 Severity : enhancement
1727 Description: make lustre randomly fail allocating memory
1728 Details : Make lustre randomly failed allocating memory for testing purpose.
1730 Severity : enhancement
1732 Description: lost problems with lov objid file
1733 Details : Fixes some scability and access to not inited memory problems
1734 in work with lov objdid file.
1739 Description: lfs find does not continue on file error
1740 Details : Continue other files processing when a file/dir is absent.
1744 Description: Inconsistent usage of lustre_pack_reply()
1745 Details : Standardize the usage of lustre_pack_reply() such that it
1746 always generate a CERROR on failure.
1749 Frequency : occasional
1751 Description: MDS hang or stay in waiting lock
1752 Details : If client receive lock with CBPENDING flag ldlm need send lock
1753 cancel as separate rpc, to avoid situation when cancel request
1754 can't processed due all i/o threads stay in wait lock.
1758 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1762 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1766 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1770 Description: Update to RHEL4 latest kernel.
1772 Severity : enhancement
1774 Description: Build SLES10 patchless client fails
1775 Details : The configure was broken by run ./configure with
1776 --with-linux-obj=.... argument for patchless client. When the
1777 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1778 can't search header adequately. Use absolute path such as
1779 -I($LINUX)/include instead.
1783 Description: interrupt oig_wait produce painc on resend.
1784 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1785 private set, and this produce situation when rq_ptlrpcd_data not
1786 copyed to new allocated request and triggered LBUG on assert
1787 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1788 ptlrpcd set if request is changed and can be safety changed to use
1793 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1794 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1795 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1796 these macros and replace them with their difinitions instead.
1800 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1801 Details : Update which_patch & target file for SLES10 latest kernel.
1804 Description: Remove client patches from SLES 10 kernel.
1805 Details : This causes SLES 10 clients to behave as patchless clients
1806 even on a Lustre-patched (server) kernel.
1808 Severity : enhancement
1810 Description: self-adjustable client's lru lists
1811 Details : use adaptive algorithm for managing client cached locks lru
1812 lists according to current server load, other client's work
1813 pattern, memory activities, etc. Both, server and client
1814 side namespaces provide number of proc tunables for controlling
1817 Severity : enhancement
1819 Description: light-weight GSS support
1820 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1821 and reduce performance overhead.
1823 Severity : enhancement
1825 Description: Linux keyring support
1826 Details : Support using service of Linux keyring for Lustre GSS internal
1827 context refresh/cache mechanism.
1831 Description: Fix errors in lfs documentation
1832 Details : Fixes man pages
1836 Description: don't use GFP_* in generic Lustre code.
1837 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1841 Description: obdclass is limited by single OBD_ALLOC(idarray)
1842 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1846 Description: warnings with build patchless client with vanila 2.6.19 and up
1847 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1848 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1852 Frequency : Only for SLES
1853 Description: sanity_quota fail test_1
1854 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1855 file, which makes TSTID[2] inunique.
1859 Frequency : Only for SLES9
1860 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1861 Details : Update target/ChangeLog/which_patch .
1865 Frequency : Only for test_10 in sanity-quota.sh
1866 Description: a bug in quota test code
1867 Details : checking a test flag in a wrong place
1871 Frequency : Only for test_13 in sanity-quota.sh
1872 Description: a bug in quota test script
1873 Details : shouldn't check overfull allocated quota
1877 Description: jbd statistics
1878 Details : Port older jbd statistics patch for sles10
1880 Severity : enhancement
1882 Description: Add printing inode info into message about error in writepage.
1887 Description: reply_lock_interpret crash due to race with it and lock cancel.
1888 Details : Do not replay locks that are being cancelled. Do not reference
1889 locks by their address during replay, just by their handle.
1893 Frequency : When flocks are used.
1894 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1895 Details : Flock locks might destroy just granted lock if it could be merged
1896 with another existing flock, this is done in completion handler,
1897 so teach ldlm_cli_enquque_fini that this is a valid case for
1903 Description: Oops in read and write path when failing to allocate lock.
1904 Details : Check if lock allocation failed and return error back.
1908 Description: lstripe command fails for valid OST index
1909 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1910 instead of lov->desc.ld_active_tgt_count.
1914 Description: sanity.sh failed test 103
1915 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1920 Description: df doesn't work properly if diskfs blocksize != 4K
1921 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1925 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1926 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1928 Severity : enhancement
1930 Description: Adaptive timeouts
1931 Details : RPC timeouts adapt to changing server load and network
1932 conditions to reduce resend attempts and improve recovery time.
1936 Description: llapi_file_create() does not allow some changes
1937 Details : add llapi_file_open() that allows specifying the mode and
1938 open flags, and also returns an open file handle.
1942 Description: merge and cleanup kernel patches.
1943 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1947 Description: Add journal checksum support.(Kernel part)
1948 Details : The journal checksum feature adds two new flags i.e
1949 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
1950 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
1951 indicates that the commit block contains the checksum for
1952 the blocks described by the descriptor blocks. Now commit
1953 record can be sent to disk without waiting for descriptor
1954 blocks to be written to disk. This behavior is controlled
1955 using JBD2_FEATURE_ASYNC_COMMIT flag.
1959 Description: OSS needs mutliple precreate threads
1960 Details : Add ability to start more than one create thread per OSS.
1964 Description: Sanity.sh test_65k failed: setstripe should have succeeded
1965 Details : execute lfs setstripe on client
1969 Description: mds_obd_create error creating tmp object
1970 Details : When the user sets quota on root, llog will be affected and can't
1971 create files and write files.
1974 Frequency : Always on ia64 patchless client, and possibly others.
1976 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1977 Details : This allows the patchless client to be loaded on architectures
1978 without this export.
1982 Description: RedHat Update kernel for RHEL5
1983 Details : Add the kernel config file for RHEL5.
1987 Description: RedHat Update kernel for RHEL5
1988 Details : Modify the kernel config file more closer RHEL5.
1992 Description: Build failure against Centos5 (RHEL5)
1993 Details : Define PAGE_SIZE when it isn't present.
1997 Description: test_7 on sanity-quota.sh can't be used on separate machines
1998 Details : add facet to handle it
2002 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2003 Details : with NFS, the anon dentry's parent was set to itself in
2004 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2005 corresponding dentry other than use rec->ur_name.
2007 Severity : enhancement
2009 Description: lfs setstripe enhancement
2010 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2014 Description: enable data checksumming by default
2015 Details : enable checksum by default, allow --disable-checksum
2016 configure option and "-o nochecksum" mount option. Checksums
2017 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2021 Description: lustre support for RHEL5
2022 Details : Add support for RHEL5.
2026 Description: Client eviction due to failover config
2027 Details : after a connection loss, the lustre client should attempt to
2028 reconnect to the last active server first before trying the
2029 other potential connections.
2033 Description: when mds and osts use different quota unit(32bit and 64bit),
2034 quota will be released repeatly.
2035 Details : void sending multiple quota reqs to mds, which will keep the
2036 status between the reqs.
2040 Description: osts not allocated evenly to files
2041 Details : change the condition to increase offset_idx
2044 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2046 Description: Data corruption for OSTs that are formatted larger than 2TB
2048 Details : When generating the bio request for lustre file writes the
2049 sector number would overflow a temporary variable before being
2050 used for the IO. The data reads correctly from Lustre (which
2051 will overflow in a similar manner) but other file data or
2052 filesystem metadata may be corrupted in some cases.
2056 Description: Tune the kernel for good SCSI performance.
2057 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2058 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2063 Description: rewrite ext2-derived code in obdclass/uuid.c
2064 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2065 from scratch preserving functionality.
2069 Description: "lfs find -obd UUID" prints directories
2070 Details : "lfs find -obd UUID" will return all directory names instead
2071 of just file names. It is incorrect because the directories
2072 do not reside on the OSTs.
2076 Description: use i_size_read and i_size_write in 2.6 port
2077 Details : replace inode->i_size access with i_size_read/write()
2080 Frequency : when removing large files
2082 Description: scheduling issue during removal of large Lustre files
2083 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2084 It causes scheduling issues when removing large files (17TB in the
2088 Frequency : only with liblustre clients on XT3
2090 Description: evictions taking too long
2091 Details : allow llrd to evict clients directly on OSTs
2094 Frequency : only on ppc
2096 Description: /proc/fs/lustre/devices broken on ppc
2097 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2098 We should make sure the type of 'index' is loff_t and then cast
2099 explicitly as needed below. Do not assign an explicitly cast
2103 Frequency : only for rhel5
2105 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2106 Details : Modify the target file & which_kernel.
2108 Severity : enhancement
2110 Description: omit set fsid for export NFS
2111 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2112 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2116 Frequency : Always, for kernels after 2.6.16
2117 Description: Fix warning idr_remove called for id=.. which is not allocated.
2118 Details : Last kernels save old s_dev before kill super and not allow
2119 to restore from callback - restore it before call kill_anon_super.
2123 Description: buffer overruns could theoretically occur
2124 Details : llapi_semantic_traverse() modifies the "path" argument by
2125 appending values to the end of the origin string, and a
2126 overrun may occur. Adding buffer overrun check in liblustreapi.
2130 Description: Fix error on 'ls .' at the top of the Lustre mount.
2131 Details : Don't revalidate dentry if it is a root dentry.
2135 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2136 Details : Modify vm-tunables-rhel4.patch.
2140 Description: Kernel config for 2.6.18-vanilla.
2141 Details : Modify targets/2.6-vanilla.target.in.
2142 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2143 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2144 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2145 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2149 Description: improve handling recoverable errors
2150 Details : if request processig with error which can be recoverable on server
2151 request should be resend, otherwise page released from cache and
2156 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2157 Details : Modify target file & which_patch.
2158 A flaw was found in the IA32 system call emulation provided
2159 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2160 value could be stored in the %RAX register, which could trigger an
2161 out-of-bounds system call table access. An untrusted local user
2162 could exploit this flaw to run code in the kernel
2163 (ie a root privilege escalation). (CVE-2007-4573).
2167 Description: O_DIRECT bypasses client statistics.
2168 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2169 read_ahead_stats were updated. copied the stats section from
2170 osc_send_oap_rpc() into async_internal().
2174 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2178 Description: change order of libsysio includes
2179 Details : '#include sysio.h' should always come before '#include xtio.h'
2183 Description: handle "serious error: objid * already exists" more gracefully
2184 Details : If LAST_ID value on disk is smaller than the objects existing in
2185 the O/0/d* directories, it indicates disk corruption and causes an
2186 LBUG(). If the object is 0-length, then we should use the existing
2187 object. This will help to avoid a full fsck in most cases.
2189 Severity : enhancement
2191 Description: adapt the lustre_config script to support the upgrade case
2192 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2193 server targets to 1.6 in parallel.
2197 Description: To avoid grant space > avaible space when the disk is almost
2198 full. Without this patch you might see the error "grant XXXX >
2199 available" or some LBUG about grant, when the disk is almost
2201 Details : In filter_check_grant, for non_grant cache write, we should
2202 check the left space by if (*left > ungranted + bytes), instead
2203 of (*left > ungranted), because only we are sure the left space
2204 is enough for another "bytes", then the ungrant space should be
2205 increase. In client, we should update cl_avail_grant only there
2206 is OBD_MD_FLGRANT in the reply.
2210 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2211 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2212 described in CVE-2007-4573. This problem could allow untrusted
2213 local users to gain root access.
2216 Frequency : when using O_DIRECT and quotas
2218 Description: Incorrect file ownership on O_DIRECT output files
2219 Details : block usage reported by 'lfs quota' does not take into account
2220 files that have been written with O_DIRECT.
2225 Description: touch file failed when fs is not full
2226 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2227 otherwise we can get ENOSP while fs is not full.
2231 Description: parallel lock callbacks
2232 Details : Instead of sending blocking and completion callbacks as separated
2233 requests, adding them to a set and sending in parallel.
2236 Frequency : only for Cray XT3
2237 Bugzilla : 12829/13455
2238 Description: Changing primary group doesn't change the group lustre assigns to
2240 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2241 overridden with the primary group provided by the group upcall,
2242 whereas the supplied fsgid can be trusted if it is in the list of
2243 supplementary groups returned by the group upcall.
2245 Severity : enhancement
2247 Description: Allow masking D_WARNING, D_ERROR messages from console
2248 Details : Console messages can now be disabled via lnet.printk.
2252 Description: User code with malformed file open parameter crashes client node
2253 Details : Before packing join_file req, all the related reference should be
2254 checked carefully in case some malformed flags cause fake
2255 join_file req on client.
2259 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2260 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2261 mode is equal to the granted mode. However, it should also check
2262 that LDLM_FL_CP_REQD is not set.
2266 Description: Heavy nfs access might result in deadlocks
2267 Details : After ELC code landed, it is now improper to enqueue any mds
2268 locks under och_sem, because enqueue might want to decide to
2269 cancel open locks for same inode we are holding och_sem for.
2273 Description: Client eviction while running blogbench
2274 Details : A lot of unlink operations with concurrent I/O can lead to a
2275 deadlock causing evictions. To address the problem, the number of
2276 oustanding OST_DESTROY requests is now throttled to
2277 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2278 callbacks are processed in priority.
2282 Description: enable ACLs on MDS by default
2283 Details : ACLs must be enabled on MDS by default.
2286 Frequency : PPC/PPC64 only
2288 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2289 Details : fix duplicated definitions between asm-ppc64/types.h and
2290 lustre_types.h on PPC.
2293 Frequency : PPC/PPC64 only
2295 Description: asm-ppc/segment.h does not exist
2296 Details : fix compile issue on PPC.
2300 Description: better handle error messages in extents code
2303 Frequency : RHEL4 only
2305 Description: mkfs is very slow on IA64/RHEL4
2306 Details : A performance regression has been discovered in the MPT Fusion
2307 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2308 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2309 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2311 Severity : enhancement
2313 Description: SNMP support enhancement
2314 Details : Adding total number of sampled request for an MDS node in snmp
2317 Severity : enhancement
2319 Description: Optimize ldlm waiting list processing for PR extent locks
2320 Details : When processing waiting list for read extent lock and meeting read
2321 lock that is same or wider to it that is not contended, skip
2322 processing rest of the list and immediatelly return current
2323 status of conflictness, since we are guaranteed there are no
2324 conflicting locks in the rest of the list.
2328 Description: Time out and refuse to reconnect
2329 Details : When the failover node is the primary node, it is possible
2330 to have two identical connections in imp_conn_list. We must
2331 compare not conn's pointers but NIDs, otherwise we can defeat
2332 connection throttling.
2336 Description: port llog fixes from b1_6 into HEAD
2337 Details : Port llog reference couting and some llog cleanups from b1_6
2338 (bug 10800) into HEAD, for protect from panic and access to already
2339 free llog structures.
2343 Description: Detect stride IO mode in read-ahead
2344 Details : When a client does stride read, read-ahead should detect that and
2345 read-ahead pages according to the detected stride pattern.
2349 Description: data checksumming impacts single node performance
2350 Details : add support for several checksum algorithm. Currently, only CRC32
2351 and Adler-32 are supported. The checksum type can be changed on
2352 the fly via /proc/fs/lustre/osc/*/checksum_type.
2356 Description: use adler32 for page checksums
2357 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2362 Description: build for x2 fails
2363 Details : fix compile issue on Cray systems.
2367 Description: Properly match for duplicate locks
2368 Details : Due to different lock order from skiplists code, we need to
2369 traverse entire list for now
2372 Frequency : only on PPC/SLES10
2374 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2375 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2376 be defined. Add a hack in mkfs_lustre.c to work around this compile
2381 Description: LASSERT on MDS when client holding flock lock dies
2382 Details : ldlm pool logic depends on number of granted locks equal to
2383 number of released locks which is not true for flock locks, so
2384 just exclude such locks from consideration.
2388 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2389 Details : Use fsfilt_sync() for both the whole filesystem sync and
2390 individual file sync to eliminate dangerous inode locking
2391 with I_LOCK that can lead to a deadlock.
2395 Description: performance in 1.6.3
2396 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2397 sure that 1MB requests are not fragmented by the block layer.
2399 Severity : enhancement
2401 Description: organize the server-side client stats on per-nid basis
2402 Details : Change the structure of stats under obdfilter and mds to
2410 The "uuid"s file would list the uuids of _active_ exports.
2411 And the clear entry is to clear all stats and stale nids.
2413 Severity : enhancement
2415 Description: eliminate client locks in face of contention
2416 Details : file contention detection and lockless i/o implementation
2417 for contended files.
2421 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2424 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2426 Description: MDS may lose file striping (and hence file data) in some cases
2427 Details : If there are additional extended attributes stored on the MDS,
2428 in particular ACLs, SELinux, or user attributes (if user_xattr
2429 is specified for the client mount options) then there is a risk
2430 of attribute loss. Additionally, the Lustre file striping
2431 needs to be larger than default (e.g. striped over all OSTs),
2432 and an additional attribute must be stored initially in the
2433 inode and then increase in size enough to be moved to the
2434 external attribute block (e.g. ACL growing in size) for file
2439 Description: skiplist implementation simplification
2440 Details : skiplists are used to group compatible locks on granted list
2441 that was implemented as tracking first and last lock of each
2442 lock group the patch changes that to using doubly linked lists
2446 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2447 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2448 because LASSERT on both the data supplied by a client, and the
2449 data on disk is dangerous and incorrect.
2451 Severity : enhancement
2453 Description: Slow truncate/writes to huge files at high offsets.
2454 Details : Directly associate cached pages to lock that protect those pages,
2455 this allows us to quickly find what pages to write and remove
2456 once lock callback is received.
2460 Description: more ldlm soft lockups
2461 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2462 starve other threads from the resource lock for a long time in
2463 case of long waiting queue, so change the debug level from
2464 D_OTHER to the less frequently used D_INFO.
2466 Severity : enhancement
2468 Description: add -gid, -group, -uid, -user options to lfs find
2472 Description: Hung threads in invalidate_inode_pages2_range
2473 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2474 once one is completed. As a result, some RPCs are stuck in the
2475 queue and are never sent.
2479 Description: filter threads hungs on waiting journal commit
2480 Details : Cleanup filter group llog code, then only filter group llog will
2481 be only created in the MDS/OST syncing process.
2485 Description: Procfs and llog threads access destoryed import sometimes.
2486 Details : Sync the import destoryed process with procfs and llog threads by
2487 the import refcount and semaphore.
2489 Severity : enhancement
2491 Description: openlock cache of b1_6 port to HEAD
2496 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2497 Details : If insertion of an extent fails, then discard the inode
2498 preallocation and free data blocks else it can lead to duplicate
2503 Description: don't always update ctime in ext3_xattr_set_handle()
2504 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2505 In some cases the ctime should not be updated, for example for
2506 2.0->1.8 compatibility it is necessary to delete an xattr and it
2507 should not update the ctime.
2511 Bugzilla : 15713/16362
2512 Description: Assertion in iopen_connect_dentry in 1.6.3
2513 Details : looking up an inode via iopen with the wrong generation number can
2514 populate the dcache with a disconneced dentry while the inode
2515 number is in the process of being reallocated. This causes an
2516 assertion failure in iopen since the inode's dentry list contains
2517 both a connected and disconnected dentry.
2521 Description: assertion failure in ldlm_handle2lock()
2522 Details : fix a race between class_handle_unhash() and class_handle2object()
2523 introduced in lustre 1.6.5 by bug 13622.
2528 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2529 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2534 Description: oops in sd_iostats_seq_show()
2535 Details : unloading/reloading the scsi low level driver triggers a kernel
2536 bug when trying to access the sd iostat file.
2541 Description: Kernel panics during QLogic driver reload
2542 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2543 patch, causing memory corruption.
2548 Description: journal_dev option does not work in b1_6
2549 Details : pass mount option during pre-mount.
2551 Severity : enhancement
2553 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2554 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2555 extent information of a file. It can be used to map logical blocks
2556 in a file to physical blocks in the block device.
2560 Description: LDLM soft lockups - improvement
2561 Details : It is be possible to send the lock handle along with each read
2562 or write request because the client is already doing a lock match
2563 itself so there isn't any reason the OST should have to re-do that
2567 Frequency : only on Cray X2
2569 Description: X2 build failures
2570 Details : fix build failures on Cray X2.
2574 Description: xid & resent requests
2575 Details : Initialize RPC XID from clock at startup (randomly if clock is
2578 Severity : enhancement
2580 Description: Add lustre_start utility to start or stop multiple Lustre servers
2585 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2586 Details : In case of memory pressure, list_del() can be called twice on
2587 req->rq_history_list, causing a kernel oops.
2591 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2592 Details : fix stack overflow in the distributed lock manager by defering
2593 export eviction after a failed AST to the elt thread instead of
2594 handling it in the dlm interpret routine.
2598 Description: Convert some comments to new format.
2599 Details : Update documenting comments to match doxygen conventions.
2603 Description: Grammar fixes.
2604 Details : A couple of trivial spelling fixes.
2608 Description: OSD_COUNTERS-mandatory
2609 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2610 to simplify the code.
2614 Description: simplify cmm_device freeing logic.
2615 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2619 Description: Add lockdep support to dt_object_operations locking interface.
2620 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2621 parameter indicating lock ordering. Update mdd code to use new
2626 Description: Introduce failloc constants for lockless IO tests.
2627 Details : Add two new failloc constants to test lockless IO. Only one of
2628 them in implemented---another is checked in yet to be landed
2633 Description: Add lockdep support for inode mutex.
2634 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2635 in the situations where Lustre has to lock more than one inode
2640 Description: Add optional invariants checking support.
2641 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2642 configure switch. This macro is to be used for consistency and
2643 sanity checks that are too expensive to be left in `production'
2648 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2649 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2654 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2655 Details : Use spin_lock_nested() in (the only) situation where more than
2656 one ldlm_lock is locked simultaneously. Also, fix possible
2657 dead-lock in ldlm_lock_change_resource() by enforcing particular
2662 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2663 Details : Instead of specifying each ldlm_lock call-back through separate
2664 parameter, wrap them into struct ldlm_callback_suite.
2668 Description: Kill join_lru obd method and its callers.
2669 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2670 in memory---a requirement implemented through ->o_join_lru() obd
2671 method in HEAD. Remove this method and its users.
2675 Description: Add asynchronous ldlm ENQUEUE completion handler.
2676 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2677 case that stock ldlm_completion_ast() cannot handle as it waits
2678 until lock is granted. Introduce new ldlm_completion_ast_async()
2679 for this. Also comment ldlm_completion_ast().
2683 Description: ldlm_error <-> errno conversion.
2684 Details : Add functions to map (rather arbitrary) between LDLM error codes
2685 and standard errno values. CLIO needs this to prevent LDLM specific
2686 constants from escaping ldlm and osc.
2690 Description: Kill unused ldlm_handle2lock_ns() function.
2691 Details : Kill unused ldlm_handle2lock_ns() function.
2695 Description: Add lu_ref support to ldlm_lock
2696 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2697 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2698 and ldlm_resource. LDLM interface has to be changed, because code that
2699 releases a reference on a lock, has to "know" what reference this is.
2700 In the most frequent case
2702 lock = ldlm_handle2lock(handle);
2704 LDLM_LOCK_PUT(lock);
2706 no changes are required. When any other reference (received _not_ from
2707 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2708 instead of LDLM_LOCK_PUT().
2710 Arguably, changes are pervasive, and interface requires some discipline
2711 for proper use. On the other hand, it was very instrumental in finding
2712 a few leaked lock references.
2716 Description: Add ldlm_lock_addref_try().
2717 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2718 attempts to addref a lock that might be being canceled
2723 Description: Add ldlm_weigh_callback().
2724 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2725 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2726 hard-coded `number of pages' logic.
2730 Description: Add lockdep annotations to llog code.
2731 Details : Use appropriately tagged _nested() locking calls in the places
2732 where llog takes more than one ->lgh_lock lock.
2736 Description: Add loi_kms_set().
2737 Details : Wrap kms updates into a helper function.
2741 Description: Constify instances of struct lsm_operations.
2742 Details : Constify instances of struct lsm_operations.
2746 Description: lu_conf support.
2747 Details : On a server, a file system object is uniquely identified
2748 by a fid, which is sufficient to locate and load all object
2749 state (inode). On a client, on the other hand, more data are
2750 necessary instantiate an object. Change lu_object_find() and
2751 friends to take additional `lu_conf' argument describing object.
2752 Typically this includes layout information.
2756 Description: lu_context fixes.
2757 Details : Introduce new lu_context functions that are needed on the client
2758 side, where some system threads (ptlrpcd) are shared by multiple
2759 modules, and so cannot be stopped during module shutdown.
2763 Description: Add start and stop methods to lu_device_type_operations.
2764 Details : Introduce two new methods in lu_device_type_operations, that are
2765 invoked when first instance of a given type is created and last one
2766 is destroyed respectively. This is need by CLIO.
2770 Description: Add lu_ref support to struct lu_device.
2771 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2772 track leaked references.
2776 Description: Introduce lu_kmem_descr.
2777 Details : lu_kmem_descr and its companion interface allow to create
2778 and destroy a number of kmem caches at once.
2782 Description: Fix lu_object finalization race.
2783 Details : Fix a race between lu_object_find() finding an object and its
2784 concurrent finalization. This race is (most likely) not possible
2785 on the server, but might happen on the client.
2789 Description: Introduce lu_ref interface.
2790 Details : lu_ref is a debugging module allowing to track references to
2791 a given object. It is quite cpu expensive, and has to be
2792 explicitly enabled with --enable-lu_ref. See usage description
2797 Description: Factor lu_site procfs stats into a separate function.
2798 Details : Separate lu_site stats printing code into a separate function
2799 that can be reused on a client.
2803 Description: Constify instances of struct {lu,dt,md}_device_operations.
2804 Details : Constify instances of struct {lu,dt,md}_device_operations.
2808 Description: Introduce struct md_site and move meta-data specific parts of
2809 struct lu_site here.
2810 Details : Move md-specific fields out of struct lu_site into special struct
2811 md_site, so that lu_site can be used on a client.
2815 Description: Kill mdd_lov_destroy().
2816 Details : Remove unused mdd code.
2820 Description: Add st_block checking to multistat.c.
2821 Details : Add st_block checking to multistat.c.
2825 Description: Add lu_ref support to struct obd_device.
2826 Details : Add lu_ref tracking to obd_device.
2830 Description: Kill obd_set_fail_loc().
2831 Details : Remove unused code.
2835 Description: Add special type for ptlrpc_request interpret functions.
2836 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2837 there. Actual usage will be in CLIO.
2841 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2842 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2843 doesn't work with lockdep.
2847 Description: Add rwv.c test program.
2848 Details : New testing program exercising readv(2) and writev(2) (Qian).
2852 Description: Add sendfile.c test program.
2853 Details : New testing program exercising sendfile(2) (Jay).
2857 Description: Ratelimit a message that can be very frequent.
2858 Details : Ratelimit a memory allocation failure message that can
2863 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2864 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2868 Description: Liblustre build fixes.
2869 Details : Liblustre build fixes.
2873 Description: libcfs: add cfs_{need,cond}_resched() interface.
2874 Details : libcfs: add cfs_{need,cond}_resched() definition and
2875 implementations for Linux, NT, and liblustre.
2877 Severity : enhancement
2879 Description: More exported tunables for mballoc
2880 Details : Add support for tunable preallocation window and new tunables for
2881 large/small requests
2885 Description: Detect corruption of block bitmap and checking for preallocations
2886 Details : Checks validity of on-disk block bitmap. Also it does better
2887 checking of number of applied preallocations. When corruption is
2888 found, it turns filesystem readonly to prevent further corruptions.
2892 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2893 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2894 window to 0. If it does hit cache miss, and read-pattern is still
2895 stride-io mode, does not reset the stride window, but also does not
2896 increase the stride window length in this case.
2900 Frequency : only for big-endian servers
2901 Description: Check if system is big-endian while mounting fs with extents feature
2902 Details : Mounting a filesystem with extents feature will fail on big-endian
2903 systems since ext3-based ldiskfs is not supported on big-endian
2904 systems. This can be over-riden with "bigendian_extents" mount option.
2906 Severity : enhancement
2908 Description: The root squash functionality
2909 Details : A security feature, which is to prevent users from being able
2910 to mount lustre on their desktop, run as root, and delete
2911 all of the files in the filesystem. The goal is accomplished by
2912 remapping user id (UID) and group id (GID) of the root user to
2913 a UID and GID specified by the system administartor via Lustre
2914 configuration management server (MGS). The functionality also
2915 allows to specify sets of clients for which the remapping does
2920 Description: Excessive recovery window
2921 Details : With AT enabled, the recovery window can be excessively long (6000+
2922 seconds). To address this problem, we no longer use
2923 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2924 timeout no longer depends on the service time, it is set to
2925 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2926 time via pb_service_time.
2930 Description: Watchdog triggered on MDS failover
2931 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2932 the OSTs know that the MDS "UUID" can be reused for the same export
2933 from a different NID, so we do not need to wait for the export to be
2937 Frequency : rare, only if using MMP with Linux RAID
2939 Description: MMP doesn't work with Linux RAID
2940 Details : While using HA for Lustre servers with Linux RAID, it is possible
2941 that MMP will not detect multiple mounts. To make this work we
2942 need to unplug the device queue in RAID when the MMP block is being
2943 written. Also while reading the MMP block, we should read it from
2944 disk and not the cached one.
2946 Severity : enhancement
2948 Description: open file using fid
2949 Details : A file can be opened using just its fid, like
2950 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
2953 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
2956 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
2957 Details : add proper locking for creation of jbd_4k slab cache
2961 Description: MMP check in ext3_remount() fails without displaying any error
2962 Details : When multiple mount protection fails during remount, proper error
2965 Severity : enhancement
2967 Description: Allow stripe size to be up to 4G-64k
2968 Details : Fix math logic to allow large stripe sizes.
2972 Description: add check for >8TB ldiskfs filesystems
2973 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
2974 Don't allow >8TB ldiskfs filesystems to be mounted without
2975 force_over_8tb mount option
2980 Descriptoin: OSS complains "too many blocks" and crashes
2981 Details : In filter_commitrw_write():
2982 1. lquota_chkquota();
2983 2. obdo_from_inode();
2984 3. lquota_pending_commit();
2985 "2." has changed uid and gid in this case, then caused
2986 lquota_chkquota() and lquota_pending_commit() to use
2987 different uid and gid. That is the root of the bug.
2991 Descriptoin: getstripe doesn't report stripe count = -1 sanely
2992 Details : Detect and print out the 'stripe count = -1' case properly.
2996 Description: Out or order replies might be lost on replay
2997 Details : In ptlrpc_retain_replayable_request if we cannot find retained
2998 request with tid smaller than one currently being added, add it
2999 to the start, not end of the list.
3001 --------------------------------------------------------------------------------
3003 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3005 * Support for kernels:
3006 2.6.5-7.283 (SLES 9),
3007 2.6.9-55.EL (RHEL 4),
3008 2.6.16.46-0.14 (SLES 10),
3009 2.6.18.8 vanilla (kernel.org)
3010 * Client support for unpatched kernels:
3011 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3012 2.6.16 - 2.6.22 vanilla (kernel.org)
3013 * Due to recently discovered recovery problems, we do not recommend
3014 using patchless RHEL 4 clients with this or any earlier release.
3015 * Recommended e2fsprogs version: 1.39.cfs8
3016 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3017 * Starting with this release, the ldiskfs backing filesystem required
3018 by Lustre is now in its own package, lustre-ldiskfs. This package
3019 should be installed. It is versioned separately from Lustre and
3020 may be released separately in future.
3024 Description: block reactivating mgc import until all deactivates complete
3025 Details : Fix race when failing back MDT/MGS to itself (testing)
3028 Frequency : at statup only
3030 Description: mds_lov_synchronize race leads to various problems
3031 Details : simultaneous MDT->OST connections at startup can cause the
3032 sync to abort, leaving the OSC in a bad state.
3034 Severity : enhancement
3036 Description: add optional extra BUILD_VERSION info
3037 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3038 to override the lustre version.
3041 Frequency : 2.6.18 servers only
3043 Description: ll_kern_mount() doesn't release the module reference
3044 Details : The ldiskfs module reference count never drops down to 0
3045 because ll_kern_mount() doesn't release the module reference.
3050 Description: server LBUG when using old ost_num_threads parameter
3051 Details : Accept the old ost_num_threads parameter but warn that it
3052 is deprecated, and fix an off-by-one error that caused an LBUG.
3057 Description: Transient SCSI error results in persistent IO issue
3058 Details : iobuf->dr_error is not reinitialized to 0 between two
3062 Frequency : sometimes when underlying device returns I/O errors
3064 Description: OSTs not going read-only during write failures
3065 Details : OSTs are not remounted read-only when the journal commit threads
3066 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3067 instead of the ext3 wrappers.
3071 Description: poor connect scaling with increasing client count
3072 Details : Don't run filter_grant_sanity_check for more than 100 exports
3073 to improve scaling for large numbers of clients.
3076 Frequency : SLES10 only
3078 Description: sanity-quota.sh quotacheck failed: rc = -22
3079 Details : Quotas cannot be enabled on SLES10.
3082 Frequency : liblustre clients only
3084 Description: getdirentries does not give error when run on compute nodes
3085 Details : getdirentries does not fail when the size specified as an argument
3086 is too small to contain at least one entry
3088 Severity : enhancement
3090 Description: Add LNET router traceability for debug purposes
3091 Details : If a checksum failure occurs with a router as part of the
3092 IO path, the NID of the last router that forwarded the bulk data
3093 is printed so it can be identified.
3098 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3099 Details : Due to a race condition, liblustre clients were occasionally
3100 evicted incorrectly.
3102 Severity : enhancement
3104 Description: lfs setstripe use optional parameters instead of postional
3107 Severity : enhancement
3109 Description: Nanosecond timestamp support for ldiskfs
3110 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3111 resolution timestamps. There is not yet support for this at
3112 the Lustre filesystem level.
3115 Frequency : during server recovery
3117 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3118 Details : request with rq_no_resend flag not awake l_wait_event if they get
3122 Frequency : nfs export on patchless client
3124 Description: connectathon hang when test nfs export over patchless client
3125 Details : Disconnected dentry cannot be found with lookup, so we do not need
3126 to unhash it or make it invalid
3129 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3130 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3131 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3136 Description: open req refcounting wrong on reconnect
3137 Details : If reconnect happened between getting open reply from server and
3138 call to mdc_set_replay_data in ll_file_open, we will schedule
3139 replay for unreferenced request that we are about to free.
3140 Subsequent close will crash in variety of ways.
3141 Check that request is still eligible for replay in
3142 mdc_set_replay_data().
3147 Description: disable writes to filesystem when reading health_check file
3148 Details : the default for reading the health_check proc file has changed
3149 to NOT do a journal transaction and write to disk, because this
3150 can cause reads of the /proc file to hang and block HA state
3151 checking on a healthy but otherwise heavily loaded system. It
3152 is possible to return to the previous behaviour during configure
3153 with --enable-health-write.
3155 Severity : enhancement
3157 Description: 64-bit inode version
3158 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3159 to the inode. This will be required for version-based recovery.
3164 Description: MDS fails to start if a duplicate client export is detected
3165 Details : in some rare cases it was possible for a client to connect to
3166 an MDS multiple times. Upon recovery the MDS would detect this
3167 and fail during startup. Handle this more gracefully.
3169 Severity : enhancement
3171 Description: Add -o localflock option to simulate old noflock
3173 Details : This will achieve local-only flock/fcntl locks
3179 Description: log_commit_thread vs filter_destroy race leads to crash
3180 Details : Take import reference before releasing llog record semaphore
3185 Description: Wrong request locking in request set processing
3186 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3187 imp_delayed_list, in this place should be used imp_lock.
3190 Frequency : when reconnection
3192 Description: Grant Leak when osc reconnect to OST
3193 Details : When osc reconnect ost, OST(filter) should check whether it
3194 should grant more space to client by comparing fed_grant and
3195 cl_avail_grant, and return the granted space to client instead
3196 of "new granted" space, because client will call osc_init_grant
3197 to update the client grant space info.
3200 Frequency : when client reconnect to OST
3202 Description: Grant Leak when osc do resend and replay bulk write
3203 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3204 bulk write request, because the grant info while be sync between
3205 OSC and OST when reconnect, and we should ignore the grant info
3206 these of resend/replay write req.
3211 Description: Grant space more than available left space sometimes.
3212 Details : When then OST is about to be full, if two bulk writing from
3213 different clients came to OST. Accord the avaliable space of the
3214 OST, the first req should be permitted, and the second one
3215 should be denied by ENOSPC. But if the seconde arrived before
3216 the first one is committed. The OST might wrongly permit second
3217 writing, which will cause grant space > available space.
3220 Frequency : when client is evicted
3222 Description: Grant might be wrongly erased when osc is evicted by OST
3223 Details : when the import is evicted by server, it will fork another
3224 thread ptlrpc_invalidate_import_thread to invalidate the
3225 import, where the grant will be set to 0. While the original
3226 thread will update the grant it got when connecting. So if
3227 the former happened latter, the grant will be wrongly errased
3228 because of this race.
3233 Description: Checking Stale with correct fid
3234 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3235 to check whether it is stale, because sometimes, we want the
3236 enqueue happened anyway, and op_data.fid2 will not be initialized.
3238 Severity : enhancement
3240 Description: update patchless client
3241 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3244 Frequency : only with 2.4 kernel
3246 Description: random memory corruption
3247 Details : size of struct ll_inode_info is to big for union inode.u and this
3248 can be cause of random memory corruption.
3253 Description: Memory leak in recovery
3254 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3255 It should also check obd_fail before fsfilt_start, otherwise if
3256 fsfilt_start return -EROFS,(failover mds during mds recovery).
3257 then the req will return with repmsg->transno = 0 and rc = EROFS.
3258 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3259 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3260 be freed no matter whether fsfilt_commit success or not.
3263 Frequency : only with huge count clients
3265 Description: Prevents from taking the superblock lock in llap_from_page for
3267 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3268 for llap_from_page call in ll_removepage prevents from taking the
3269 superblock lock for a soon died page.
3274 Description: Not check open intent error before release open handle
3275 Details : in some rare cases, the open intent error is not checked before
3276 release open handle, which may cause
3277 ASSERTION(open_req->rq_transno != 0), because it tries to release
3278 the failed open handle.
3283 Description: Set cat log bitmap only after create log success.
3284 Details : in some rare cases, the cat log bitmap is set too early. and it
3285 should be set only after create log success.
3289 Description: Accessing a block bevice can re-enable I/O when Lustre is
3290 tearing down a device.
3291 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3296 Description: service threads may hog cpus when there are a lot of requests
3297 Details : Insert cond_resched to give other threads a chance to use some CPU
3302 Description: the cat log was not initialized in recovery
3303 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3304 unlink log on mds will not be initialized until mds post
3305 recovery. And also in mds post recovery, the unlink log will
3306 initialization will be done asynchronausly, so there will be race
3307 between add unlink log and unlink log initialization.
3311 Description: brw_stats were being printed incorrectly
3312 Details : brw_stats were being printed as log2 but all of them were not
3313 recorded as log2. Also remove some code duplication arising from
3314 filter_tally_{read,write}.
3318 Frequency : rare, only in recovery.
3319 Description: ASSERTION(req->rq_type != LI_POISON) failed
3320 Details : imp_lock should be held while iterating over imp_sending_list for
3321 prevent destroy request after get timeout in ptlrpc_queue_wait.
3325 Description: replay-single.sh test 52 fails
3326 Details : A lock's skiplist need to be cleanup when it being unlinked
3327 from its resource list.
3331 Description: Short directio read returns full requested size rather than
3333 Details : Direct I/O operations should return actual amount of bytes
3334 transferred rather than requested size.
3336 Severity : enhancement
3338 Description: metadata RPC reduction (e.g. for rm performance)
3339 Details : decrease the amount of synchronous RPC between clients and servers
3340 by canceling conflicing lock before the operation on the client
3341 and packing thier handles into the main operation RPC to server.
3343 Severity : enhancement
3345 Description: Async OSC create to avoid the blocking unnecessarily.
3346 Details : If an OST has no remaining object, system will block on the
3347 creation when it needs to create a new object on this OST. Now,
3348 ways use pre-created objects when available, instead of blocking on
3349 an empty osc while others are not empty. If we must block, we
3350 block for the shortest possible period of time.
3354 Description: improve handling recoverable errors
3355 Details : if request processig with error which can be recoverable on server
3356 request should be resend, otherwise page released from cache and
3359 Severity : enhancement
3361 Description: refine locking for avoid write wrong info into lov_objid file
3362 Details : fix possible races with add new target and write/update data in
3365 --------------------------------------------------------------------------------
3367 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3372 Frequency : on some architectures
3374 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3375 Details : Uninitialized flags sometimes cause configuration commands to
3379 Frequency : patchless clients only
3381 Description: missing __iget() symbol export
3382 Details : The __iget() symbol export is missing. To avoid the need for
3383 this on patchless clients the deathrow inode reaper is turned
3384 off, and we depend on the VM to clean up old inodes. This
3385 dependency was during via the fix for bug 12181.
3387 --------------------------------------------------------------------------------
3389 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3391 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3392 INTEROPERATE with older versions automatically. In many cases a
3393 special upgrade step is needed. Please read the
3394 user documentation before upgrading any part of a live system.
3395 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3396 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3397 * WARNING: Lustre configuration and startup changes are required with
3398 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3402 Severity : enhancement
3404 Description: MountConf
3405 Details : Lustre configuration is now managed via mkfs and mount
3406 commands instead of lmc and lconf. New obd types (MGS, MGC)
3407 are added for dynamic configuration management. See
3408 https://mail.clusterfs.com/wikis/lustre/MountConf for
3411 Severity : enhancement
3413 Description: dynamic OST addition
3414 Details : OSTs can now be added to a live filesystem
3416 Severity : enhancement
3418 Description: startup order invariance
3419 Details : MDTs and OSTs can be started in any order. Clients only
3420 require the MDT to complete startup.
3422 Severity : enhancement
3424 Description: parallel, asynchronous orphan cleanup
3425 Details : orphan cleanup is now performed in separate threads for each
3426 OST, allowing parallel non-blocking operation.
3428 Severity : enhancement
3430 Description: optimized stripe assignment
3431 Details : stripe assignments are now made based on ost space available,
3432 ost previous usage, and OSS previous usage, in order to try
3433 to optimize storage space and networking resources.
3435 Severity : enhancement
3437 Description: Permanently set tunables
3438 Details : All writable /proc/fs/lustre tunables can now be permanently
3439 set on a per-server basis, at mkfs time or on a live system.
3441 Severity : enhancement
3443 Description: Lustre message v2
3444 Details : Add lustre message format v2.
3446 Severity : enhancement
3448 Description: client OST exclusion list
3449 Details : Clients can be started with a list of OSTs that should be
3450 declared "inactive" for known non-responsive OSTs.
3454 Description: SPEC SFS validation failure on NFS v2 over lustre.
3455 Details : Changes the blocksize for regular files to be 2x RPC size,
3456 and not depend on stripe size.
3458 Severity : enhancement
3460 Description: Multiple MD RPCs in flight.
3461 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3462 To avoid overly-overloading MDS, introduce a limit on number of
3463 MDS RPCs in flight for a single client and add /proc controls
3464 to adjust this limit.
3466 Severity : enhancement
3468 Description: client read/write statistics
3469 Details : Add client read/write call usage stats for performance
3470 analysis of user processes.
3471 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3472 file access. extents_stats shows chunk size distribution.
3473 extents_stats_per_process show chunk size distribution per
3476 Severity : enhancement
3478 Description: mds statistics
3479 Details : Add detailed mds operations statistics in
3480 /proc/fs/lustre/mds/*/stats.
3484 Description: Failure of copying files with lustre special EAs.
3485 Details : Client side always return success for setxattr call for lustre
3486 special xattr (currently only "trusted.lov").
3491 Description: Refcount LNET uuids
3492 Details : The global LNET uuid list grew linearly with every startup;
3493 refcount repeated list entries instead of always adding to
3496 Severity : enhancement
3498 Description: Dynamic service threads
3499 Details : Within a small range, start extra service threads
3500 automatically when the request queue builds up.
3503 Frequency : mixed-endian client/server environments
3505 Description: mixed-endian crashes
3506 Details : The new msg_v2 system had some failures in mixed-endian
3509 Severity : enhancement
3511 Description: Easy OST removal
3512 Details : OSTs can be permanently deactivated with e.g. 'lctl
3513 conf_param lustre-OST0001.osc.active=0'
3515 Severity : enhancement
3517 Description: MGS proc entries
3518 Details : Added basic proc entries for the MGS showing what filesystems
3521 Severity : enhancement
3523 Description: provide MGS failover
3524 Details : Added config lock reacquisition after MGS server failover.
3526 Severity : enhancement
3528 Description: add Linux 2.4 support
3529 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3533 Description: a large application tries to do I/O to the same resource and dies
3534 in the middle of it.
3535 Details : Check the req->rq_arrival time after the call to
3536 ost_brw_lock_get(), but before we do anything about
3537 processing it & sending the BULK transfer request. This
3538 should help move old stale pending locks off the queue as
3539 quickly as obd_timeout.
3542 Frequency : when an incorrect nid is specified during startup
3544 Description: ptlrpc connect to non-existant node causes kernel crash
3545 Details : LNET can't be re-entered from an event callback, which
3546 happened when we expire a message after the export has been
3547 cleaned up. Instead, hand the zombie cleanup off to another
3550 Severity : enhancement
3552 Description: plain/inodebits lock performance improvement
3553 Details : Grouping plain/inodebits in granted list by their request modes
3554 and bits policy, thus improving the performance of search through
3558 Frequency : only if OST filesystem is corrupted
3560 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3561 Details : for a short time RPCs with bulk IO are in the replay list,
3562 but replay of bulk IOs is unimplemented. If the OST filesystem
3563 is corrupted due to disk cache incoherency and then replay is
3564 started it is possible to trip an assertion. Avoid putting
3565 committed RPCs into the replay list at all to avoid this issue.
3568 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3570 Description: System hang on startup
3571 Details : This bug allowed the liblustre (e.g. catamount) client to
3572 return to the app before handling all startup RPCs. This
3573 could leave the node unresponsive to lustre network traffic
3574 and manifested as a server ptllnd timeout.
3576 Severity : enhancement
3578 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3579 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3580 boolean to control whether to print peer debug info when a
3581 client's RPC times out.
3584 Frequency : only for kernels with patches from Lustre below 1.4.3
3586 Description: Remove old rdonly API
3587 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3590 Frequency : only for devices with external journals
3592 Description: Set external device read-only also
3593 Details : During a commanded failover stop, we set the disk device
3594 read-only while the server shuts down. We now also set any
3595 external journal device read-only at the same time.
3598 Frequency : when upgrading from 1.4 while trying to change parameters
3600 Description: The wrong (new) MDC name was used when setting parameters for
3601 upgraded MDT's. Also allows changing of OSC (and MDC)
3602 parameters if --writeconf is specified at tunefs upgrade time.
3605 Frequency : when setting specific ost indicies
3607 Description: QOS code breaks on skipped indicies
3608 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3609 created with --index need not be sequential.
3611 Severity : enhancement
3613 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3614 Details : The uninit_groups feature works in conjunction with the kernel
3615 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3616 up the pass1 processing of e2fsck. This is a read-only feature
3617 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3618 mount filesystems that have had this feature enabled.
3620 Severity : enhancement
3622 Description: Improve multi-block allocation algorithm to avoid fragmentation
3623 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3624 allocation locality and avoid filesystem fragmentation.
3626 ------------------------------------------------------------------------------
3628 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3630 * Support for kernels:
3631 2.4.21-47.0.1.EL (RHEL 3)
3632 2.6.5-7.283 (SLES 9)
3633 2.6.9-42.0.10.EL (RHEL 4)
3634 2.6.12.6 vanilla (kernel.org)
3635 2.6.16.27-0.9 (SLES 10)
3636 * Recommended e2fsprogs version: 1.39.cfs5
3638 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3642 Frequency : occasional, depends on client load and configuration
3643 Bugzilla : 12181, 12203
3644 Description: data loss for recently-modified files
3646 Details : In some cases it is possible that recently written or created
3647 files may not be written to disk in a timely manner (this should
3648 normally be within 30s unless client IO load is very high).
3649 The problem appears as zero-length files or files that are a
3650 multiple of 1MB in size after a client crash or client eviction
3651 that are missing data at the end of the file.
3653 This problem is more likely to be hit on clients where files are
3654 repeatedly created and unlinked in the same directory, clients
3655 have a large amount of RAM, have many CPUs, the filesystem has
3656 many OSTs, the clients are rebooted frequently, and/or the files
3657 are not accessed by other nodes after being written.
3659 The presence of the problem can be detected by looking at
3660 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3661 smaller than the second (nr_unused) then dirty files will not
3662 be flushed automatically to disk. "sync; sleep 10" should be
3663 run several times on the node before unmounting it to update
3664 Lustre (this is also safe to run on nodes without this problem).
3666 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3667 that can cause this same problem, so customers using that kernel
3668 also need to update the kernel in addition to Lustre. In order
3669 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3671 It is normal that files written just before a client crash (less
3672 than 30s) may not yet have been flushed to disk, even for local
3676 Frequency : frequent on thin XT3 nodes
3678 Description: UUID collision on thin XT3 Linux nodes
3679 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3680 randomly, since we relied on an insufficiently-seeded PRNG.
3685 Description: OSS hangs after "All ost request buffers busy"
3686 Details : A deadlock between quota and journal operations caused OSS
3687 hangs after printing "All ost request buffers busy."
3690 Frequency : always on liblustre builds
3692 Description: Cleanup compiler warnings on liblustre
3695 Frequency : always on liblustre builds on XT3
3697 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3701 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3702 > 32000 subdirectories
3703 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3704 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3705 subdirectory count crosses 32000. This will aid e2fsck to
3706 correctly handle more than 32000 subdirectories.
3709 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3711 Description: System hang on startup
3712 Details : This bug allowed the liblustre (e.g. catamount) client to
3713 return to the app before handling all startup RPCs. This
3714 could leave the node unresponsive to lustre network traffic
3715 and manifested as a server ptllnd timeout.
3717 Severity : enhancement
3719 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3720 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3721 boolean to control whether to print peer debug info when a
3722 client's RPC times out.
3727 Description: make O_SYNC working on 2.6 kernels
3728 Details : 2.6 kernels use different method for mark pages for write,
3729 so need add a code to lustre for O_SYNC work.
3734 Description: Failure to close file and release space on NFS
3735 Details : Put inode details into lock acquired in ll_intent_file_open.
3736 Use mdc_intent_lock in ll_intent_open to properly
3737 detect all kind of errors unhandled by mdc_enqueue.
3742 Description: proc file read during shutdown sometimes raced obd removal,
3744 Details : Add lock to prevent obd access after proc file removal.
3747 Frequency : Only for files larger than 4GB on 32-bit clients.
3749 Description: improperly doing page alignment of locks
3750 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3751 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3756 Description: RPCs being resent when they shouldn't be
3757 Details : Some RPCs that should not be resent are being resent. This
3758 can cause inconsistencies in the RPC state machine. Do not
3759 resend such requests.
3762 Frequency : rare, only with NFS export
3764 Description: Crash on NFS re-export node
3765 Details : under very unusual load conditions an assertion is hit in
3766 ll_intent_file_open()
3769 Frequency : only if OST filesystem is corrupted
3771 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3772 Details : for a short time RPCs with bulk IO are in the replay list,
3773 but replay of bulk IOs is unimplemented. If the OST filesystem
3774 is corrupted due to disk cache incoherency and then replay is
3775 started it is possible to trip an assertion. Avoid putting
3776 committed RPCs into the replay list at all to avoid this issue.
3781 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3782 Details : Large single O_DIRECT read and write calls can fail to allocate
3783 a sufficiently large buffer to process the request. In case of
3784 allocation failure the allocation is retried with a smaller
3785 buffer and broken into smaller requests.
3787 Severity : enhancement
3789 Description: Add -o localflock option to simulate old noflock behaviour.
3790 Details : This will achieve local-only flock/fcntl locks coherentness.
3795 Description: versioning check is incomplete
3796 Details : Checking the version difference of client vs. server, report
3797 error if the gap is too big.
3802 Description: add support PG_writeback bit
3803 Details : add support for PG_writeback bit for Lustre, for more carefull
3804 work with page cache in 2.6 kernel. This also fix some deadlocks
3805 and remove hack for work O_SYNC with 2.6 kernel.
3807 Severity : enhancement
3809 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3810 Details : The uninit_groups feature works in conjunction with the kernel
3811 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3812 up the pass1 processing of e2fsck. This is a read-only feature
3813 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3814 mount filesystems that have had this feature enabled.
3816 Severity : enhancement
3818 Description: Improve multi-block allocation algorithm to avoid fragmentation
3819 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3820 allocation locality and avoid filesystem fragmentation.
3822 ------------------------------------------------------------------------------
3824 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3826 * Support for kernels:
3827 2.6.9-42.0.3.EL (RHEL 4)
3828 2.6.5-7.276 (SLES 9)
3829 2.4.21-47.0.1.EL (RHEL 3)
3830 2.6.12.6 vanilla (kernel.org)
3831 2.6.16.21-0.8 (SLES10)
3832 * Recommended e2fsprogs version: 1.39.cfs2-0
3834 * The backwards-compatible /proc/sys/portals symlink has been removed
3835 in this release. Before upgrading, please ensure that you change
3836 any configuration scripts or /etc/sysctl.conf files that access
3837 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3838 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3839 in advance of the upgrade on any system running Lustre 1.4.6 or
3840 newer, since /proc/sys/lnet was added in that version.
3841 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3845 Frequency : only when quota is used
3847 Description: avoid scanning export list for quota master
3848 Details : Change the algorithms to avoid scanning export list in order
3849 to improve the efficiency.
3852 Frequency : MDS failover only, very rarely
3854 Description: "went back in time" messages on mds failover
3855 Details : The greatest transno may be lost when the current operation
3856 finishes with an error (transno==0) and the client's last_rcvd
3857 record is over-written. Save the greatest transno in the
3858 mds_last_transno for this case.
3861 Frequency : always for specific kernels and striping counts
3863 Description: client may get "Matching packet too big" without ACL support
3864 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3865 when trying to access files in certain configurations. The
3866 clients should in fact be denied when mounting because they do
3867 not understand ACLs.
3870 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3872 Description: many clients connecting with IO in progress causes connect timeouts
3873 Details : Avoid synchronous journal commits to avoid delays caused by many
3874 clients connecting/disconnecting when bulk IO is in progress.
3875 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3876 OST_IO_PORTAL to avoid delays behind potentially many pending
3880 Frequency : occasionally with multiple writers to a single file
3882 Description: shared writes to file may result in wrong size reported by stat()
3883 Details : Allow growing of kms when extent lock is cancelled
3886 Frequency : always with random mmap IO to multi-striped file
3888 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3889 Details : Only if the hole is at the end of OST object so that kms is too
3890 small. Fix is to increase kms accordingly in ll_nopage.
3893 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3895 Description: writes to a missing object would leak memory on the OST
3896 Details : If there is an inconsistency between the MDS and OST filesystems,
3897 such that the MDS references an object that doesn't exist, writes
3898 to that object will leak memory due to incorrect cleanup in the
3899 error handling path, eventually running out of memory on the OST.
3904 Description: Creating too long symlink causes lustre errors
3905 Details : Check symlink and name lengths before sending requests to MDS.
3908 Frequency : only if flock is enabled (not on by default)
3910 Description: posix locks not released on fd closure on 2.6.9+
3911 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3912 kernels, this caused such locks not to be released on fd close and
3913 then assertions on fs unmount about still used locks.
3916 Frequency : MDS failover only, very rarely
3918 Description: clients may get ASSERTION(granted_lock != NULL)
3919 Details : When request was taking a long time, and a client was resending
3920 a getattr by name lock request. The were multiple lock requests
3921 with the same client lock handle and
3922 mds_getattr_name->fixup_handle_for_resent_request found one of the
3923 lock handles but later failed with ASSERTION(granted_lock != NULL).
3928 Description: handle->h_buffer_credits > 0, assertion failure
3929 Details : h_buffer_credits is zero after truncate, causing assertion
3930 failure. This patch extends the transaction or creates a new
3934 Frequency : NFS re-export or patchless client
3935 Bugzilla : 11179, 10796
3936 Description: Crash on NFS re-export node (__d_move)
3937 Details : We do not want to hash the dentry if we don't have a lock.
3938 But if this dentry is later used in d_move, we'd hit uninitialised
3939 list head d_hash, so we just do this to init d_hash field but
3940 leave dentry unhashed.
3943 Frequency : NFS re-export or patchless client
3945 Description: NFS exports has problem with symbolic link
3946 Details : lustre client didn't properly install dentry when re-exported
3947 to NFS or running patchless client.
3950 Frequency : NFS re-export or patchless client
3952 Description: Various nfs/patchless fixes.
3953 Details : fixes reuse disconected alias for lookup process - this fixes
3954 warning "find_exported_dentry: npd != pd",
3955 fix permission error with open files at nfs.
3956 fix apply umask when do revalidate.
3959 Frequency : occasional
3961 Description: Crash on NFS re-export node
3962 Details : calling clear_page() on the wrong pointer triggered oops in
3963 generic_mapping_read().
3966 Frequency : rarely, using O_DIRECT IO
3968 Description: unaligned directio crashes client with LASSERT
3969 Details : check for unaligned buffers before trying any requests.
3972 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3974 Description: stale data returned from RAID cache
3975 Details : If only a small amount of IO is done to the RAID device before
3976 reading it again it is possible to get stale data from the RAID
3977 cache instead of reading it from disk.
3980 Frequency : always for sles10 kernel
3982 Description: sles10 support
3983 Details : ll_follow_link: compile fixes and using of nd_set_link
3984 under newer kernels.
3987 Frequency : depends on arch, kernel and compiler version, always on sles10
3990 Description: recursive or deep enough symlinks cause stack overflow
3991 Details : getting rid of large stack-allocated variable in
3995 Frequency : depends on hardware
3997 Description: lustre write performance loss in the SLES10 kernel
3998 Details : the performance loss is caused by using of write barriers in the
3999 ext3 code. The SLES10 kernel turns barrier support on by
4000 default. The fix is to undo that change for ldiskfs.
4002 ------------------------------------------------------------------------------
4004 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4006 * Support for kernels:
4007 2.6.9-42.0.3EL (RHEL 4)
4008 2.6.5-7.276 (SLES 9)
4009 2.4.21-47.0.1.EL (RHEL 3)
4010 2.6.12.6 vanilla (kernel.org)
4014 Frequency : quota enabled and large files being deleted
4016 Description: releasing more than 4GB of quota at once hangs OST
4017 Details : If a user deletes more than 4GB of files on a single OST it
4018 will cause the OST to spin in an infinite loop. Release
4019 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4024 Description: statfs data retrieved from /proc may be stale or zero
4025 Details : When reading per-device statfs data from /proc, in the
4026 {kbytes,files}_{total,free,avail} files, it may appear
4027 as zero or be out of date.
4030 Frequency : systems with MD RAID1 external journal devices
4032 Description: lconf's call to blkid is confused by RAID1 journal devices
4033 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4034 of returning all block devices that match the journal UUID.
4037 Frequency : always, for aggregate stripe size over 4GB
4039 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4040 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4041 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4044 Frequency : always if "lfs find" used on a local file/directory
4046 Description: "lfs find" segfaults if used on a local file/directory
4047 Details : The case where a directory component was not specified wasn't
4048 handled correctly. Handle this properly.
4051 Frequency : always on ppc64
4053 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4054 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4055 perform endianness conversion.
4058 Frequency : rarely (truncate to non-zero file size after write under load)
4059 Bugzilla : 10730, 10687
4060 Description: Files padded with zeros to next 4K multiple
4061 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4062 it is possible that files that are truncated to a non-zero size
4063 immediately after being written are filled with zero bytes beyond
4064 the truncated size. No file data is lost.
4066 Severity : enhancement
4068 Description: Allow recovery/failover for liblustre clients.
4069 Details : liblustre clients were unaware of failover configurations until
4072 Severity : enhancement
4074 Description: user file locks should fail when not mounting with flock option
4075 Details : Set up an error-returning stub in ll_file_operations.lock field
4076 to prevent incorrect behaviour when client is mounted without
4077 flock option. Also, set up properly f_op->flock field for
4081 Frequency : always on ia64
4083 Description: "lfs df" loops on printing out MDS statfs information
4084 Details : The obd_ioctl_data was not initialized and in some systems
4085 this caused a failure during the ioctl that did not return
4086 an error. Initialize the struct and return an error on failure.
4089 Frequency : SLES 9 only
4091 Description: Error of copying files with lustre special EAs as root
4092 Details : Client side always return success for setxattr call for lustre
4093 special xattr (currently only "trusted.lov").
4096 Frequency : rarely on clusters with both ia64+i386 clients
4098 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4099 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4100 clients are doing concurrent writes to the same file it is
4101 possible that the ia64 clients may LASSERT because the OST
4102 extent locks are not PAGE_SIZE aligned. Ensure that grown
4103 locks are always aligned on the request boundary.
4106 Frequency : specific use, occasional
4108 Description: Overwriting in use executable truncates on-disk binary image
4109 Details : If one node attempts to overwrite an executable in use by
4110 another node, we now correctly return ETXTBSY instead of
4111 truncating the file.
4113 Severity : enhancement
4115 Description: Async OSC create to avoid the blocking unnecessarily.
4116 Details : If an OST has no remaining object, system will block on the creating
4117 when need to create a new object on this OST. Now, ways use
4118 pre-created objects when available, instead of blocking on an
4119 empty osc while others are not empty. If we must block, we block
4120 for the shortest possible period of time.
4125 Description: chmod on Lustre root is propagated to other clients
4126 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4127 invalid by the follow_mount time.
4130 Frequency : liblustre clients only
4132 Description: Race in 'instant cancel' lock handling could lead to such locks
4133 never to be granted in case of SMP MDS
4134 Details : Do not destroy not yet granted but cbpending locks in
4138 Frequency : replay/resend of open
4140 Description: non null lock assetion failure in mds_intent_policy
4141 Details : Trying to replay/resend lockless open requests resulted in
4142 mds_open() returning 0 with no lock. Now it sets a flag if
4143 a lock is going to be returned.
4145 Severity : enhancement
4147 Description: Checksum enhancements
4148 Details : New checksum enhancements allow for resending RPCs that failed
4151 Severity : enhancement
4153 Description: Tunables on number of dirty pages in cacche
4154 Details : Allow to set limit on number of dirty pages cached.
4159 Description: client crash on unmount - lock still has references
4160 Details : In some error handling cases it was possible to leak a lock
4161 reference on a client while accessing a file. This was not
4162 harmful to the client during operation, but would cause the
4163 client to crash when the filesystem is unmounted.
4166 Frequency : specific case, rare
4168 Description: ETXTBSY on mds though file not in use by client
4169 Details : ETXTBSY is no longer incorrectly returned when attempting to
4170 chmod or chown a directory that the user previously tried to
4171 execute or a currently-executing binary.
4174 Frequency : extremely rare except on liblustre-based clients
4176 Description: Lustre space not freed when files are deleted
4177 Details : Clean up open-unlinked files after client eviction. Previously
4178 the unlink was skipped and the files remained as orphans.
4183 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4184 Details : In some cases it was possible to send a blocking callback to
4185 a client doing a glimpse, even though that client didn't get
4186 a lock granted. When the glimpse lock is cancelled on the OST
4187 the freed lock is left on the waiting list and corrupted the list.
4190 Frequency : all core dumps
4192 Description: Broke core dumps to lustre
4193 Details : Negative dentry may be unhashed if parent does not have UPDATE
4194 lock, but some callers, e.g. do_coredump, expect dentry to be
4195 hashed after successful create, hash it in ll_create_it.
4197 ------------------------------------------------------------------------------
4199 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4201 * Support for kernels:
4202 2.6.9-42.0.2.EL (RHEL 4)
4203 2.6.5-7.276 (SLES 9)
4204 2.4.21-40.EL (RHEL 3)
4205 2.6.12.6 vanilla (kernel.org)
4209 Frequency : always on RHEL 3
4211 Description: Number of open files grows over time
4212 Details : The number of open files grows over time, whether or not
4213 Lustre is started. This was due to a filp leak introduced
4214 by one of our kernel patches.
4216 ------------------------------------------------------------------------------
4218 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4220 * Support for kernels:
4221 2.6.9-42.EL (RHEL 4)
4222 2.6.5-7.276 (SLES 9)
4223 2.4.21-40.EL (RHEL 3)
4224 2.6.12.6 vanilla (kernel.org)
4229 Bugzilla : 5719, 9635, 9792, 9684
4230 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4231 Details : If a server is under heavy load and cannot reply to new
4232 connection requests before the client resends the (re)connect,
4233 the connection handling code can behave badly if two service
4234 threads are concurrently handing separate (re)connections from
4235 the same client. Add better locking to the connection handling
4236 code, and ensure that only a single connection will be processed
4237 for a given client UUID, even if the lock is dropped.
4239 Severity : enhancement
4241 Description: add TCP zero-copy support to kernel
4242 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4243 sends if the hardware supports scatter-gather and checksumming.
4244 This allows socklnd to do client-write and server-read more
4245 efficiently and reduce CPU utilization from skbuf copying.
4248 Frequency : only if NFS exporting from client
4250 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4251 Details : If setting attributes on a file created under NFS that had
4252 never been opened it would be possible to oops the client
4253 if the file had no objects.
4256 Frequency : always for liblustre
4258 Description: liblustre client does MDS+OSTs setattr RPC for each write
4259 Details : When doing a write from a liblustre client, the client
4260 incorrectly issued an RPC to the MDS and each OST the file was
4261 striped over in order to update the timestamps. When writing
4262 with small chunks and many clients this could overwhelm the MDS
4263 with RPCs. In all cases it would slow down the write because
4264 these RPCs are unnecessary.
4266 Severity : enhancement
4268 Description: allow number of MDS service threads to be changed at module load
4269 Details : It is now possible to change the number of MDS service threads
4270 running. Adding "options mds mds_num_threads={N}" to the MDS's
4271 /etc/modprobe.conf will set the number of threads for the next
4272 time Lustre is restarted (assuming the "mds" module is also
4273 reloaded at that time). The default number of threads will
4274 stay the same, 32 for most systems.
4279 Description: OST crash if filesystem is unformatted or corrupt
4280 Details : If an OST is started on a device that has never been formatted
4281 or if the filesystem is corrupt and cannot even mount then the
4282 error handling cleanup routines would dereference a NULL pointer.
4287 Description: NULL pointer deref in llap_from_page.
4288 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4289 incorrect) mapping if the page was truncated/reclaimed while it was
4290 searched for. Check for this condition and skip such pages when
4291 doing readahead. Introduce extra check to llap_from_page() to
4292 verify page->mapping->host is non-NULL (so page is not anonymous).
4295 Frequency : Sometimes when using sys_sendfile
4297 Description: "page not covered by a lock" warnings from ll_readpage
4298 Details : sendfile called ll_readpage without right page locks present.
4299 Now we introduced ll_file_sendfile that does necessary locking
4300 around call to generic_file_sendfile() much like we do in
4304 Frequency : with certain MDS communication failures at client mount time
4306 Description: NULL pointer deref after failed client mount
4307 Details : a client connection request may delayed by the network layer
4308 and not be sent until after the PTLRPC layer has timed out the
4309 request. If the client fails the mount immediately it will try
4310 to clean up before the network times out the request. Add a
4311 reference from the request import to the obd device and delay
4312 the cleanup until the network drops the request.
4315 Frequency : occasionally during client (re)connect
4317 Description: assertion failure during client (re)connect
4318 Details : processing a client connection request may be delayed by the
4319 client or server longer than the client connect timeout. This
4320 causes the client to resend the connection request. If the
4321 original connection request is replied in this interval, the
4322 client may trip an assertion failure in ptlrpc_connect_interpret()
4323 which thought it would be the only running connect process.
4326 Frequency : only with obd_echo servers and clients that are rebooted
4328 Description: kernel BUG accessing uninitialized data structure
4329 Details : When running an obd_echo server it did not start the ping_evictor
4330 thread, and when a client was evicted an uninitialized data
4331 structure was accessed. Start the ping_evictor in the RPC
4332 service startup instead of the OBD startup.
4334 Severity : enhancement
4335 Bugzilla : 10193 (patchless)
4336 Description: Remove dependency on various unexported kernel interfaces.
4337 Details : No longer need reparent_to_init, exit_mm, exit_files,
4338 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4341 Frequency : rare (only users of deprecated and unsupported LDAP config)
4343 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4344 Details : LDAP apparently contains 'lustreName' attributes instead of
4345 'name'. A simple remapping of the name is sufficient.
4348 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4350 Description: waiting_locks_callback trips kernel BUG if client is evicted
4351 Details : Running with the dump_on_timeout debug flag turned on makes
4352 it possible that the waiting_locks_callback() can try to dump
4353 the Lustre kernel debug logs from an interrupt handler. Defer
4354 this log dumping to the expired_lock_main() thread.
4356 Severity : enhancement
4358 Description: Support NFS exporting on 2.6 kernels.
4359 Details : Implement non-rawops metadata methods for NFS server to use without
4360 changing NFS server code.
4363 Frequency : very rare (synthetic metadata workload only)
4365 Description: two racing renames might cause an MDS thread to deadlock
4366 Details : Running the "racer" program may cause one MDS thread to rename
4367 a file from being the source of a rename to being the target of
4368 a rename at exactly the same time that another thread is doing
4369 so, and the second thread has already enqueued these locks after
4370 doing a lookup of the target and is trying to relock them in
4371 order. Ensure that we don't try to re-lock the same resource.
4374 Frequency : only very large systems with liblustre clients
4376 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4377 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4378 all OSTs in parallel. This allows the eviction of stale liblustre
4379 clients to proceed much faster than if they were done in series,
4380 and also offers similar improvements for other set_info RPCs.
4385 Description: excessive CPU usage during initial read phase on client
4386 Details : During the initial read phase on a client, it would agressively
4387 retry readahead on the file, consuming too much CPU and impacting
4388 performance (since 1.4.5.8). Improve the readahead algorithm
4389 to avoid this, and also improve some other common cases (read
4390 of small files in particular, where "small" is files smaller than
4391 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4396 Description: MDS crash when receiving packet with unknown intent.
4397 Details : Do not LBUG in unknown intent case, just return -EFAULT
4399 Severity : enhancement
4400 Bugzilla : 9293, 9385
4401 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4402 Details : Do not serialize getattr (non-intent version) and statfs.
4405 Frequency : occasional, when OST network is overloaded/intermittent
4407 Description: client evicted by OST after bulk IO timeout
4408 Details : If a client sends a bulk IO request (read or write) the OST
4409 may evict the client if it is unresposive to its data GET/PUT
4410 request. This is incorrect if the network is overloaded (takes
4411 too long to transfer the RPC data) or dropped the OST GET/PUT
4412 request. There is no need to evict the client at all, since
4413 the pinger and/or lock callbacks will handle this, and the
4414 client can restart the bulk request.
4417 Frequency : Always when mmapping file with no objects
4419 Description: client crashes when mmapping file with no objects
4420 Details : Check that we actually have objects in a file before doing any
4421 operations on objects in ll_vm_open, ll_vm_close and
4427 Description: Request leak when working with deleted CWD
4428 Details : Introduce advanced request refcount tracking for requests
4429 referenced from lustre intent.
4431 Severity : Enhancement
4433 Description: Cache open file handles on client.
4434 Details : MDS now will return special lock along with openhandle, if
4435 requested and client is allowed to hold openhandle, even if unused,
4436 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4437 closing files for every read/write openration.
4439 Severity : Enhancement
4441 Description: Cache open negative dentries on client when possible.
4442 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4443 negative dentries on lock revocation.
4448 Description: Remounting a client read-only wasn't possible with a zconf mount
4449 Details : It wasn't possible to remount a client read-only with llmount.
4451 Severity : enhancement
4452 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4453 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4454 MPICH is included to add Lustre-specific ADIO interfaces.
4455 This is based closely on the UFS ADIO layer and only differs
4456 in file creation, in order to allow the OST striping to be set.
4457 This is user-contributed code and not supported by CFS.
4462 Description: extended inode attributes (immutable, append-only) work improperly
4463 when 2.4 and 2.6 kernels are used on client/server or vice versa
4464 Details : Introduce kernel-independent values for these flags.
4466 Severity : enhancement
4469 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4470 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4471 granularity. Now, Fractional MB granularity is be supported,
4472 this is very useful for low memory system.
4474 Severity : enhancement
4476 Description: Getattr by fid
4477 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4478 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4479 have locks and avoids getattr rpc requests for it on every stat.
4482 Frequency : Always, for filesystems larger than 2TB
4484 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4485 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4486 filesystems larger than 2TB to be mounted with mballoc enabled.
4489 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4491 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4492 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4493 filesystems larger than 2TB is mounted then the kernel will
4494 silently corrupt the start of the filesystem. CONFIG_LBD is
4495 enabled for all CFS-supported kernels, but the possibility of
4496 this happening with a modified kernel config exists.
4498 Severity : enhancement
4500 Description: add client O_DIRECT support for 2.6 kernels
4501 Details : It is now possible to do O_DIRECT reads and writes to files
4502 in the Lustre client mountpoint on 2.6 kernel clients.
4504 Severity : enhancement
4506 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4507 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4508 parallel, not waiting for response from every OST before sending
4509 a rpc to the next OST.
4514 Description: setattr vs write race when updating file timestamps
4515 Details : Client processes that update a file timestamp into the past
4516 right after writing to the file (e.g. tar) it is possible that
4517 the updated file modification time can be reset to the current
4518 time due to a race between processing the setattr and write RPC.
4520 Severity : enhancement
4522 Description: Bring 'lfs find' closer in line with regular Linux find.
4523 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4524 -print0 options and obtains all the needed info through the lustre
4527 Severity : enhancement
4529 Description: support up to 1024 configured devices on one node
4530 Details : change obd_dev array from statically allocated to dynamically
4531 allocated structs as they are first used to reduce memory usage
4536 Description: Flush dirty partially truncated pages during truncate
4537 Details : Immediatelly flush partially truncated pages in filter_setattr,
4538 this way we completely avoid having any pages in page cache on OST
4539 and can retire ugly workarounds during writes to flush such pages.
4544 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4545 Details : protect inode from truncation within vfs_unlink() context
4546 just take a reference before calling vfs_unlink() and release it
4547 when parent's i_sem is free.
4552 Description: last_id value checked outside lock on OST caused LASSERT failure
4553 Details : If there were multiple MDS->OST object precreate requests in
4554 flight, it was possible that the OST's last object id was checked
4555 outside a lock and incorrectly tripped an assertion. Move checks
4556 inside locks, and discard old precreate requests.
4559 Frequency : always, if extents are used on OSTs
4561 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4562 Details : OSTs using the extents format would not zero the high 16 bits of
4563 the index physical block number. This is not a problem for any
4564 OST filesystems smaller than 16TB, and no kernels support ext3
4565 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4566 new/modified files) and can be fixed for existing filesystems
4567 with e2fsprogs-1.39-cfs1.
4572 Description: import connection selection may be incorrect if timer wraps
4573 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4574 import connections to be ignored if the 32-bit jiffies counter
4575 wraps. Use a 64-bit jiffies counter.
4578 Frequency : very large clusters immediately after boot
4580 Description: LNET request buffers exhausted under heavy short-term load
4581 Details : If a large number of client requests are generated on a service
4582 that has previously never seen so many requests it is possible
4583 that the request buffer growth cannot keep up with the spike in
4584 demand. Instead of dropping incoming requests, they are held in
4585 the LND until the RPC service can accept more requests.
4588 Frequency : Sometimes during replay
4590 Description: Assertion failure in ll_local_open after replay.
4591 Details : If replay happened on an open request reply before we were able
4592 to set replay handler, reply will become not swabbed tripping the
4593 assertion in ll_local_open. Now we set the handler right after
4594 recognising of open request
4597 Frequency : very rare
4599 Description: kernel reports "badness in vsnprintf"
4600 Details : Reading from the "recovery_status" /proc file in small chunks
4601 may cause a negative length in lprocfs_obd_rd_recovery_status()
4602 call to vsnprintf() (which is otherwise harmless). Exit early
4603 if there is no more space in the output buffer.
4605 Severity : enhancement
4607 Description: clear OBD RPC statistics by writing to them
4608 Details : It is now possible to clear the OBD RPC statistics by writing
4609 to the "stats" file.
4614 Description: Client mtime is not the same on different clients after utimes
4615 Details : In some cases, the client was using the utimes() syscall on
4616 a file cached on another node. The clients now validate the
4617 ctime from the MDS + OSTs to determine which one is right.
4622 Description: Inability to activate failout mode
4623 Details : lconf script incorrectly assumed that in pythong string's numeric
4624 value is used in comparisons.
4627 Frequency : always with multiple stripes per file
4629 Description: Inefficient object allocation for mutli-stripe files
4630 Details : When selecting which OSTs to stripe files over, for files with
4631 a stripe count that divides evenly into the number of OSTs,
4632 the MDS is always picking the same starting OST for each file.
4633 Return the OST selection heuristic to the original design.
4638 Description: mount failures may take full timeout to return an error
4639 Details : Under some heavy load conditions it is possible that a
4640 failed mount can wait for the full obd_timeout interval,
4641 possibly several minutes, before reporting an error.
4642 Instead return an error as soon as the status is known.
4644 Frequency : quota enabled and large files being deleted
4646 Description: releasing more than 4GB of quota at once hangs OST
4647 Details : If a user deletes more than 4GB of files on a single OST it
4648 will cause the OST to spin in an infinite loop. Release
4649 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4654 Description: statfs data retrieved from /proc may be stale or zero
4655 Details : When reading per-device statfs data from /proc, in the
4656 {kbytes,files}_{total,free,avail} files, it may appear
4657 as zero or be out of date.
4660 Frequency : systems with MD RAID1 external journal devices
4662 Description: lconf's call to blkid is confused by RAID1 journal devices
4663 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4664 of returning all block devices that match the journal UUID.
4667 Frequency : always, for aggregate stripe size over 4GB
4669 Description: assertion fails when trying to use 4GB stripe size
4670 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4671 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4674 Frequency : always on ppc64
4676 Description: the first write on an ext3 filesystem with mballoc got stuck
4677 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4678 perform endianness conversion.
4680 ------------------------------------------------------------------------------
4682 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4684 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4685 INTEROPERATE with older versions automatically. Please read the
4686 user documentation before upgrading any part of a live system.
4687 * WARNING: Lustre networking configuration changes are required with
4688 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4691 * Support for kernels:
4692 2.6.9-22.0.2.EL (RHEL 4)
4693 2.6.5-7.244 (SLES 9)
4694 2.6.12.6 vanilla (kernel.org)
4697 Severity : enhancement
4698 Bugzilla : 7981/8208
4699 Description: Introduced Lustre Networking (LNET)
4700 Details : LNET is new networking infrastructure for Lustre, it includes
4701 a reorganized network configuration mode (see the user
4702 documentation for full details) as well as support for routing
4703 between different network fabrics. Lustre Networking Devices
4704 (LNDs) for the supported network fabrics have also been
4705 created for this new infrastructure.
4707 Severity : enhancement
4708 Description: Introduced Access control lists
4709 Details : clients can set ACLs on files and directories in order to have
4710 more fine-grained permissions than the standard Unix UGO+RWX.
4711 The MDS must be started with the "-o acl" mount option.
4713 Severity : enhancement
4714 Description: Introduced filesystem quotas
4715 Details : Administrators may now establish per-user quotas on the
4718 Severity : enhancement
4720 Description: Configuration change for the XT3
4721 The PTLLND is now used to run Lustre over Portals on the XT3
4722 The configure option(s) --with-cray-portals are no longer used.
4723 Rather --with-portals=<path-to-portals-includes> is used to
4724 enable building on the XT3. In addition to enable XT3 specific
4725 features the option --enable-cray-xt3 must be used.
4730 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4731 Details : A race between allocating a new llcd and re-getting the llcd_lock
4732 allowed another thread to grab newly-allocated llcd.
4734 Severity : enhancement
4736 Description: 2.6 OST async journal commit and locking fix to improve performance
4737 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4738 2.6 kernels are now async as they already were in 2.4 kernels so
4739 that they can commit concurrently with the network bulk transfer.
4740 For block-allocated files the filter allocation semaphore is held
4741 to avoid filesystem fragmentation during allocation. BKL lock
4742 removed for 2.6 xattr operations where it is no longer needed.
4747 Description: lconf incorrectly determined whether two IP networks could talk
4748 Details : In some more complicated routing and multiple-network
4749 configurations, lconf will avoid trying to make a network
4750 connection to a disjoint part of the IP space. It was doing the
4751 math incorrectly for one set of cases.
4756 Description: Fix for potential infinite loop processing records in an llog.
4757 Details : If an llog record is corrupted/zeroed, it is possible to loop
4758 forever in llog_process(). Validate the llog record length
4759 and skip the remainder of the block on error.
4762 Frequency : occasional (liblustre only)
4764 Description: liblustre could not open files whose last component is a symlink
4765 Details : sysio_path_walk() would incorrectly pass the open intent to
4766 intermediate path components.
4769 Frequency : rare (liblustre only with non-standard tuning)
4770 Bugzilla : 7201 (7350)
4771 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4772 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4773 after it has been released, only do this for referenced locks
4775 Severity : enhancement
4777 Description: specifying an (invalid) directory default stripe_size of -1
4778 would reset the directory default striping
4779 Details : stripe_size -1 was used internally to signal directory stripe
4780 removal, now use "all default" to signal dir stripe removal
4781 as a directory striping of "all default" is not useful
4784 Frequency : common for large clusters running liblustre clients
4786 Description: doing an ls when liblustre clients are running is slow
4787 Details : sending a glimpse AST to a liblustre client waits for every AST
4788 to time out, as liblustre clients will not respond. Since they
4789 cannot cache data we refresh the OST lock LVB from disk instead.
4791 Severity : enhancement
4793 Description: doing an ls at the same time as file IO can be slow
4794 Details : enqueue and other "small" requests can be blocked behind many
4795 large IO requests. Create a new OST IO portal for non-IO
4796 requests so they can be processed faster.
4799 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4801 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4802 created file with mode 000
4803 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4804 allow an ftruncate() as a non-root user to a file with mode 000.
4805 Lustre now respects this flag to disable mode checks when
4806 truncating a file owned by the user
4809 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4812 Description: Revoking locks from clients that went dead or catatonic might take
4814 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4815 cancellation of such locks instant on servers without waiting for
4816 any reply from clients. Clients drops these locks when cancel
4817 notification from server is received without replying.
4820 Frequency : liblustre-only, when liblustre client dies or becomes busy
4822 Description: Doing ls on Linux clients can take a long time with active
4824 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4825 granting such locks to it in the first place if possible. Locks
4826 are taken by proxy on the OST during the read or write and
4827 dropped immediately afterward. Add connect flags handling, do
4828 not grant locks to liblustre clients for glimpse ASTs.
4830 Severity : enhancement
4832 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4833 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4834 sequential reads and behaves badly with applications doing random
4835 reads. Improve it by reading ahead at least read region, and
4836 avoiding excessive large RPC for small reads.
4838 Severity : enhancement
4840 Description: Creating more than 1000 files for a single job may cause a load
4841 imbalance on the OSTs if there are also a large number of OSTs.
4842 Details : qos_prep_create() uses an OST index reseed value that is an
4843 even multiple of the number of available OSTs so that if the
4844 reseed happens in the middle of the object allocation it will
4845 still utilize the OSTs as uniformly as possible.
4850 Description: OST or MDS may oops in ping_evictor_main()
4851 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4852 but doesn't restart at beginning of obd_exports_timed list
4855 Severity : enhancement
4857 Description: improve by-nid export eviction on the MDS and OST
4858 Details : allow multiple exports with the same NID to be evicted at one
4859 time without re-searching the exports list.
4862 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4864 Description: MDS may oops in groups_free()
4865 Details : in rare race conditions a newly allocated group_info struct is
4866 freed again, and this can be NULL. The 2.4 compatibility code
4867 for groups_free() checked for a NULL pointer, but 2.6 did not.
4870 Frequency : common for liblustre clients doing little filesystem IO
4871 Bugzilla : 9352, 7313
4872 Description: server may evict liblustre clients accessing contended locks
4873 Details : if a client is granted a lock or receives a completion AST
4874 with a blocking AST already set it would not reply to the AST
4875 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4878 Frequency : lfs setstripe, only systems with more than 160 OSTs
4880 Description: unable to set striping with a starting offset beyond OST 160
4881 Details : llapi_create_file() incorrectly limited the starting stripe
4882 index to the maximum single-file stripe count.
4885 Frequency : LDAP users only
4887 Description: lconf did not handle in-kernel recovery with LDAP properly
4888 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4890 Severity : enhancement
4892 Description: bind OST threads to NUMA nodes to improve performance
4893 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4894 node and do their allocations there to localize memory access
4896 Severity : enhancement
4898 Description: llmount can determine client NID directly from Myrinet (GM)
4899 Details : the client NID code from gmnalnid was moved directly into
4900 llmount, removing the need to use this or specifying the
4901 client NID explicitly when mounting GM clients with zeroconf
4904 Frequency : if client is started with down MDS
4906 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4907 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4908 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4909 ptlrpc_import_delayed_req() because we didn't check if the
4910 request was interrupted, and we also didn't break out of the
4911 event loop if there was no timeout
4916 Description: data loss during non-page-aligned writes to a single file from
4917 both multiple nodes and multiple threads on one node at same time
4918 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4919 inconsistency led to false short-reads, that were cached and later
4920 used by ->prepare_write() to fill in partially written page,
4921 leading to data loss.
4924 Frequency : always, if lconf --abort_recovery used
4926 Description: lconf --abort_recovery fails with 'Operation not supported'
4927 Details : lconf was attempting to abort recovery on the MDT device and not
4930 Severity : enhancement
4932 Description: remove cleanup logs
4933 Details : replace lconf-generated cleanup logs with lustre internal
4934 cleanup routines. Eliminates the need for client-cleanup and
4937 Severity : enhancement
4939 Description: add support for EAs (user and system) on lustre filesystems
4940 Details : it is now possible to store extended attributes in the Lustre
4941 client filesystem, and with the user_xattr mount option it
4942 is possible to allow users to store EAs on their files also
4944 Severity : enhancement
4946 Description: Add possibility (config option) to show minimal available OST free
4948 Details : When compiled with --enable-mindf configure option, statfs(2)
4949 (and so, df) will return least minimal free space available from
4950 all OSTs as amount of free space on FS, instead of summary of
4951 free spaces of all OSTs.
4953 Severity : enhancement
4955 Description: do not expand extent locks acquired on OST-side
4956 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4957 by server: they are not cached anyway.
4960 Frequency : when mmap is used/binaries executed from Lustre
4962 Description: Unmmap pages before throwing them away from read cache.
4963 Details : llap_shrink cache now attempts to unmap pages before discarding
4964 them (if unmapping failed - do not discard). SLES9 kernel has
4965 extra checks that trigger if this unmapping is not done first.
4970 Description: lconf didn't resolve symlinks before checking to see whether a
4971 given mountpoint was already in use
4974 Frequency : when migrating failover services
4975 Bugzilla : 6395, 9514
4976 Description: When migrating a subset of services from a node (e.g. failback
4977 from a failover service node) the remaining services would
4978 time out and evict clients.
4979 Details : lconf --force (implied by --failover) sets the global obd_timeout
4980 to 5 seconds in order to quickly disconnect, but this caused
4981 other RPCs to time out too quickly. Do not change the global
4982 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4984 Severity : enhancement
4985 Frequency : if MDS is started with down OST
4986 Bugzilla : 9439,5706
4987 Description: Allow startup/shutdown of an MDS without depending on the
4988 availability of the OSTs.
4989 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4990 Add appropriate locking and lov-osc refcounts for safe
4991 cleaning. Add osc abort_inflight calls in case the
4992 synchronize never started.
4995 Frequency : occasional (Cray XT3 only)
4997 Description: root not authorized to access files in CRAY_PORTALS environment
4998 Details : The client process capabilities were not honoured on the MDS in
4999 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5000 been accessed by an authorized user then root was able to access
5001 the file on the local client also. The root user capabilities
5002 are now allowed on the MDS, as this environment has secure UID.
5005 Frequency : occasional
5007 Description: ldiskfs "too long searching" message happens too often
5008 Details : A debugging message (otherwise harmless) prints too often on
5009 the OST console. This has been reduced to only happen when
5010 there are fragmentation problems on the filesystem.
5015 Description: Division by zero in statfs when all OSCs are inactive
5016 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5017 lov_statfs divides by value returned by lov_get_stripecnt().
5021 Bugzilla : 9489, 3273
5022 Description: First write from each client to each OST was only 4kB in size,
5023 to initialize client writeback cache, which caused sub-optimal
5024 RPCs and poor layout on disk for the first written file.
5025 Details : Clients now request an initial cache grant at (re)connect time
5026 and so that they can start streaming writes to the cache right
5027 away and always do full-sized RPCs if there is enough data.
5028 If the OST is rebooted the client also re-establishes its grant
5029 so that client cached writes will be honoured under the grant.
5034 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5035 Details : Now I/O RPCs go to different portal number and (presumably) fast
5036 lock requests (and glimses) and other RPCs get their own service
5037 threads pool that should be able to service those RPCs
5040 Severity : enhancement
5042 Description: Ability to exchange lustre version between client and servers and
5043 issue warnings at client side if client is too old. Also for
5044 liblustre clients there is ability to refuse connection of too old
5046 Details : New 'version' field is added to connect data structure that is
5047 filled with version info. That info is later checked by server and
5051 Frequency : rare, liblustre only.
5052 Bugzilla : 9296, 9581
5053 Description: Two simultaneous writes from liblustre at offset within same page
5054 might proceed at the same time overwriting eachother with stale
5056 Details : I/O lock withing llu_file_prwv was released too early, before data
5057 actually was hitting the wire. Extended lock-holding time until
5058 server acknowledges receiving data.
5061 Frequency : extremely rare. Never observed in practice.
5063 Description: avoid generating lustre_handle cookie of 0.
5064 Details : class_handle_hash() generates handle cookies by incrementing
5065 global counter, and can hit 0 occasionaly (this is unlikely, but
5066 not impossible, because initial value of cookie counter is
5067 selected randonly). Value of 0 is used as a sentinel meaning
5068 "unassigned handle" --- avoid it. Also coalesce two critical
5069 sections in this function into one.
5071 Severity : enhancement
5073 Description: allow liblustre clients to delegate truncate locking to OST
5074 Details : To avoid overhead of locking, liblustre client instructs OST to
5075 take extent lock in ost_punch() on client's behalf. New connection
5076 flag is added to handle backward compatibility.
5078 Severity : enhancement
5079 Bugzilla : 4928, 7341, 9758
5080 Description: allow number of OST service threads to be specified
5081 Details : a module parameter allows the number of OST service threads
5082 to be specified via "options ost ost_num_threads={N}" in the
5083 OSS's /etc/modules.conf or /etc/modprobe.conf.
5087 Bugzilla : 6146, 9635, 9895
5088 Description: servers crash with bad pointer in target_handle_connect()
5089 Details : In rare cases when a client is reconnecting it was possible that
5090 the connection request was the last reference for that export.
5091 We would temporarily drop the export reference and get a new
5092 one, but this may have been the last reference and the export
5093 was just destroyed. Get new reference before dropping old one.
5095 Severity : enhancement
5096 Frequency : if client is started with failover MDS
5098 Description: Allow multiple MDS hostnames in the mount command
5099 Details : Try to read the configuration from all specified MDS
5100 hostnames during a client mount in case the "primary"
5103 Severity : enhancement
5105 Description: Stop sending data to evicted clients as soon as possible.
5106 Details : Check if the client we are about to send or are sending data to
5107 was evicted already. (Check is done every second of waiting,
5108 for which l_wait_event interface was extended to allow checking
5109 of exit condition at specified intervals).
5112 Frequency : rare, normally only when NFS exporting is done from client
5114 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5116 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5119 Frequency : always, if srand() is called before liblustre initialization
5121 Description: Liblustre uses system PRNG disturbing its usage by user application
5122 Details : Introduce internal to lustre fast and high-quality PRNG for
5123 lustre usage and make liblustre and some other places in generic
5124 lustre code to use it.
5126 Severity : enhancement
5127 Bugzilla : 9477, 9557, 9870
5128 Description: Verify that the MDS configuration logs are updated when xml is
5129 Details : Check if the .xml configuration logs are newer than the config
5130 logs stored on the MDS and report an error if this is the case.
5131 Request --write-conf, or allow starting with --old_conf.
5133 Severity : enhancement
5135 Description: Handle symlinks in the path when checking if Lustre is mounted.
5136 Details : Resolve intermediate symlinks when checking if a client has
5137 mounted a filesystem to avoid duplicate client mounts.
5142 Description: lconf can hit an error exception but still return success.
5143 Details : The lconf command catches the Command error exception at the top
5144 level script context and will exit with the associated exit
5145 status, but doesn't ensure that this exit status is non-zero.
5150 Description: failure of ptlrpc thread startup can cause oops
5151 Details : Starting a ptlrpc service thread can fail if there are a large
5152 number of threads or the server memory is very fragmented.
5153 Handle this without oopsing.
5156 Frequency : always, only if liblustre and non-default acceptor port was used
5158 Description: liblustre cannot connect to servers with non-default acceptor port
5159 Details : tcpnal_set_default_params() was not called and was therefore
5160 ignoring the environment varaible TCPNAL_PORT, as well as other
5161 TCPNAL_ environment variables
5166 Description: two objects could be created on the same OST for a single file
5167 Details : If an OST is down, in some cases it was possible to create two
5168 objects on a single OST for a single file. No problems other
5169 than potential performance impact and spurious error messages.
5173 Bugzilla : 5681, 9562
5174 Description: Client may oops in ll_unhash_aliases
5175 Details : Client dcache may become inconsistent in race condition.
5176 In some cases "getcwd" can fail if the current directory is
5182 Description: Inode refcounting problems in NFS export code
5183 Details : link_raw functions used to call d_instantiate without obtaining
5184 extra inode reference first.
5188 Bugzilla : 9942, 9903
5189 Description: Referencing freed requests leading to crash, memleaks with NFS.
5190 Details : We used to require that call to ll_revalidate_it was always
5191 followed by ll_lookup_it. Also with revalidate_special() it is
5192 possible to call ll_revalidate_it() twice for the same dentry
5193 even if first occurence returned success. This fix changes semantic
5194 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5195 reference on a request referred from the intent.
5196 ll_intent_release() then releases such a request.
5199 Frequency : rare, normally benchmark loads only
5201 Description: unlinked inodes were kept in memory on the client
5202 Details : If a client is repeatedly creating and unlinking files it
5203 can accumulate a lot of stale inodes in the inode slab cache.
5204 If there is no other client load running this can cause the
5205 client node to run out of memory. Instead flush old inodes
5206 from client cache that have the same inode number as a new inode.
5209 Frequency : SLES9 2.6.5 kernel and long filenames only
5210 Bugzilla : 9969, 10379
5211 Description: utime reports stale NFS file handle
5212 Details : SLES9 uses out-of-dentry names in some cases, which confused
5213 the lustre dentry revalidation. Change it to always use the
5217 Frequency : rare, unless heavy write-truncate concurrency is continuous
5218 Bugzilla : 4180, 6984, 7171, 9963, 9331
5219 Description: OST becomes very slow and/or deadlocked during object unlink
5220 Details : filter_destroy() was holding onto the parent directory lock
5221 while truncating+unlinking objects. For very large objects this
5222 may block other threads for a long time and slow overall OST
5223 responsiveness. It may also be possible to get a lock ordering
5224 deadlock in this case, or run out of journal credits because of
5225 the combined truncate+unlink. Solution is to do object truncate
5226 first in one transaction without parent lock, and then do the
5227 final unlink in a new transaction with the parent lock. This
5228 reduces the lock hold time dramatically.
5231 Frequency : rare, 2.4 kernels only
5233 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5234 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5235 before unmounting their filesystem in order to avoid deadlock.
5236 The kernel_locked() function in 2.4 kernels only checks whether
5237 the kernel lock is held, not whether it is this process that is
5238 holding it as 2.6 kernels do.
5243 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5244 Details : The client ptlrpc code may be trying to reconnect to a down
5245 server before a previous connection attempt has timed out.
5246 Increase the reconnect interval to be longer than the connection
5247 timeout interval to avoid sending duplicate connections to
5251 Frequency : echo_client brw_test command
5253 Description: fix echo_client to work with OST preallocated code
5254 Details : OST preallocation code (5137) didn't take echo_client IO path
5255 into account: echo_client calls filter methods outside of any
5256 OST thread and, hence, there is no per-thread preallocated
5257 pages and buffers to use. Solution: hijack pga pages for IO. As
5258 a byproduct, this avoids unnecessary data copying.
5262 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5263 Description: Client can oops in mdc_commit_close() after open replay
5264 Details : It was possible for the MDS to return an open request with no
5265 transaction number in mds_finish_transno() if the client was
5266 evicted, but without actually returning an error. Clients
5267 would later try to replay that open and may trip an assertion
5268 Simplify the client close codepath, and always return an error
5269 from the MDS in case the open is not successful.
5272 Frequency : rare, 2.6 OSTs only
5274 Description: OST may deadlock under high load on fragmented files
5275 Details : If there was a heavy load and highly-fragmented OST filesystems
5276 it was possible to have all the OST threads deadlock waiting on
5277 allocation of biovecs, because the biovecs were not released
5278 until the entire RPC IO was completed. Instead, release biovecs
5279 as soon as they are complete to ensure forward IO progress.
5281 Severity : enhancement
5283 Description: Support for specifying external journal device at mount
5284 Details : If an OST or MDS device is formatted with an external journal
5285 device, this device major/minor is stored in the ext3 superblock
5286 and may not be valid for failover. Allow detecting and
5287 specifying the external journal at mount time.
5292 Description: Mounting an MDS with pending unlinked files may cause oops
5293 Details : target_finish_recovery() calls mds_postrecov() which returned
5294 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5295 considers this an error and immediately begins cleaning up the
5296 lov, just after starting the mds_lov process
5298 Severity : enhancement
5300 Description: Implement 'lfs df' to report actual free space on per-OST basis
5301 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5302 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5303 usage of objects; '-h' to report in human readable format.
5305 ------------------------------------------------------------------------------
5307 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5314 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5315 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5316 at startup during orphan unlinking. Instead, skip buddy update
5317 before setup, as it will be regenerated after recovery is complete.
5322 Description: If an OST is inactive, its locks might reference stale inodes.
5323 Details : lov_change_cbdata() must iterate over all namespaces, even if
5324 they are inactive to clear inode references from the lock.
5326 Severity : enhancement
5327 Frequency : occasional, if non-standard max_dirty_mb used
5329 Description: Client will block write RPCs if not enough grant
5330 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5331 then the client will block writes while waiting for another RPC
5332 to complete instead of consuming its dirty limit. With change
5333 we get improved performance when max_dirty_mb is small.
5335 Severity : enhancement
5336 Bugzilla : 3389, 6253
5337 Description: Add support for supplementary groups on the MDS.
5338 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5339 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5340 lookups for user supplementary groups into a cache.
5344 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5345 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5346 for liblustre O_EXCL is enforced by MDS.
5350 Description: readdir never returns NULL in liblustre.
5351 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5352 dentry in struct dirent.
5356 Frequency : liblustre only, depends on application IO pattern
5357 Description: liblustre clients evicted if not contacting servers
5358 Details : Don't put liblustre clients into the ping_evictor list, so
5359 they will not be evicted by the pinger ever.
5361 Severity : enhancement
5363 Description: Add ability to evict clients by NID from MDS.
5364 Details : By echoing "nid:$NID" string into
5365 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5366 $NID would be instantly evicted from this MDS and from all active
5367 OSTs connected to it.
5371 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5372 Details : lookup_it_finish() used to query file size from OSTs that was not
5377 Description: service threads change working directory to that of init
5378 Details : Starting lustre service threads may pin the working directory
5379 of the parent thread, making that filesystem busy. Threads
5380 now change to the working directory of init to avoid this.
5384 Frequency : during shutdown only
5385 Description: shutdown with a failed MDS or OST can cause unmount to hang
5386 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5391 Frequency : 2.6 only
5392 Description: chmod/chown may include an extra supplementary group
5393 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5394 supplementary group and if none is specified this is used.
5397 Bugzilla : 5479 (6816)
5399 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5400 Details : If lookup is in progress on a file that is unlinked we might try
5401 to revalidate the inode and fail in revalidate after lookup is
5402 complete and ll_file_open() enqueues the open again but
5403 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5406 Frequency : always, if lconf --abort_recovery used
5408 Description: lconf --abort_recovery fails with 'Operation not supported'
5409 Details : lconf was attempting to abort recovery on the MDT device and not
5412 ------------------------------------------------------------------------------
5414 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5419 Frequency : rare (only unsupported configurations with a node running as an
5421 Bugzilla : 6514, 5137
5422 Description: Mounting a Lustre file system on a node running as an OST could
5424 Details : OSTs now preallocates memory needed to write out data at
5425 startup, instead of when needed, to avoid having to
5426 allocate memory in possibly low memory situations.
5427 Specifically, if the file system is mounted on on OST,
5428 memory pressure could force it to try to write out data,
5429 which it needed to allocate memory to do. Due to the low
5430 memory, it would be unable to do so and the node would
5431 become unresponsive.
5433 Severity : enhancement
5435 Description: Addition of lconf --service command line option
5436 Details : lconf now accepts a '--service <arg>' option, which is
5437 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5439 Severity : enhancement
5441 Description: Failover mode is now the default for OSTs.
5442 Details : By default, OSTs will now run in failover mode. To return to
5443 the old behaviour, add '--failout' to the lmc line for OSTs.
5445 Severity : enhancement
5447 Description: Health checks are now provided for MDS and OSTs
5448 Details : Additional detailed health check information on MSD and OSTs
5449 is now provided through the procfs health_check value.
5452 Frequency : occasional, depends on IO load
5454 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5455 after numerous create/delete cycles
5456 Details : The ext3 inode allocation policy would not allocate new inodes
5457 very well on the OSTs because there are no new directories
5458 being created. Instead we look for groups with free space if
5459 the parent directories are nearly full.
5464 Description: Network or server problems during mount may cause partially
5465 mounted clients instead of returning an error.
5466 Details : The config llog parsing code may overwrite the error return
5467 code during mount error handling, returning success instead
5473 Description: MDS can fail to allocate large reply buffers
5474 Details : After long uptimes the MDS can fail to allocate large reply
5475 buffers (e.g. zconf client mount config records) due to memory
5476 fragmentation or consumption by the buffer cache. Preallocate
5477 some large reply buffers so that these replies can be sent even
5478 under memory pressure.
5482 Frequency : rare (liblustre)
5483 Description: fsx running with liblustre complained that using truncate() to
5484 extend the file doesn't work. This patch corrects that issue.
5485 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5486 ATTR_SIZE and lsm use in llu_setattr_raw.
5490 Frequency : rare, only 2.6 kernels
5491 Description: Unusual file access patterns on the MDS may result in inode
5492 data being lost in very rare circumstances.
5493 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5494 code in the 2.6 kernel caused the inode and/or EA data not to
5495 be read from disk, causing single-file corruption.
5499 Frequency : rare, only 2.6 filesystems using extents
5500 Description: Heavy concurrent write and delete load may cause data corruption.
5501 Details : It was possible under high-load situations to have an extent
5502 metadata block in the block device cache from a just-unlinked
5503 file overwrite a newly-allocated data block. We now unmap any
5504 metadata buffers that alias just-allocated data blocks.
5508 Frequency : filesystems with default stripe_count larger than 77
5509 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5510 Details : lconf specifies an inode size of 4096 bytes when the default
5511 stripe_count is larger than 77. This conflicts with the default
5512 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5513 this case to avoid pinning too much memory for large EAs.
5515 ------------------------------------------------------------------------------
5517 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5522 Frequency : rare (extremely heavy IO load with hundreds of clients)
5524 Description: Client is evicted, gets IO error writing to file
5525 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5526 caused clients to be evicted for AST timeouts. The fixes in
5527 bug 5192 mean we no longer need to have such short AST timeouts
5528 so ldlm_timeout has been increased.
5531 Frequency : occasional during --force or --failover shutdown under load
5532 Bugzilla : 5949, 4834
5533 Description: Server oops/LBUG if stopped with --force or --failover under load
5534 Details : a collection of import/export refcount and cleanup ordering
5535 issues fixed for safer force cleanup
5538 Frequency : only filesystems larger than 120 OSTs
5539 Bugzilla : 5990, 6223
5540 Description: lfs getstripe would oops on a very large filesystem
5541 Details : lov_getconfig used kfree on vmalloc'd memory
5544 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5545 Bugzilla : 6242, 6243
5546 Description: reading from files that had been truncated to a non-zero size
5547 but never opened returned no data
5548 Details : ll_file_read() reads zeros from no-object files to EOF
5553 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5554 Details : There's little harm in aborting MDS/OSS recovery and letting it
5555 try again, so I removed the LASSERT and return an error instead.
5557 Severity : enhancement
5559 Description: New debugging infrastructure for tracking down data corruption
5560 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5561 (b) cover more of the client-side code path, and (c) try to narrow
5562 down where problems occurred
5566 Bugzilla : 3819, 4364, 4397, 6313
5567 Description: Racing close and eviction MDS could cause assertion in mds_close
5568 Details : It was possible to get multiple mfd references during close and
5569 client eviction, leading to one thread referencing a freed mfd.
5571 Severity: : enhancement
5572 Bugzilla : 3262, 6359
5573 Description: Attempts to reconnect to servers are now more aggressive.
5574 Details : This builds on the enhanced upcall-less recovery that was added
5575 in 1.4.2. When trying to reconnect to servers, clients will
5576 now try each server in the failover group every 10 seconds. By
5577 default, clients would previously try one server every 25 seconds.
5582 Description: After recovery, certain operations trigger a failed
5583 assertion on a client.
5584 Details : Failing over an mds, using lconf -d --failover, while a
5585 client was doing a readdir() call would cause the client to
5586 LBUG after recovery completed and the readdir() was resent.
5588 Severity : enhancement
5590 Description: Default groups are now added by lconf
5591 Details : You can now run lconf --group <servicename> without having to
5592 manually add groups with lmc.
5595 Frequency : occasional
5597 Description: Nodes with an elan id of 0 trigger a failed assertion
5600 Frequency : always when accessing e.g. tty/console device nodes
5602 Description: tty and some other devices nodes cannot be used on lustre
5603 Details : file's private_data field is used by device data and lustre
5604 values in there got lost. New field was added to struct file to
5605 store fs-specific private data.
5608 Frequency : when exporting Lustre via NFS
5610 Description: NFSD failed occasionally when looking up a path component
5611 Details : NFSD is looking up ".." which was broken in ext3 directories
5612 that had grown large enough to become hashed.
5615 Frequency : Clusters with multiple interfaces not on the same subnet
5617 Description: Nodes will repeatedly try to reconnect to an interface which it
5618 cannot reach and report an error to the log.
5619 Details : Extra peer list entries will be created by lconf with some peers
5620 unreachable. lconf now validates the peer before adding it.
5623 Frequency : Only if a default stripe is set on the filesystem root.
5625 Description: Setting a default stripe on the filesystem root prevented the
5626 filesystem from being remounted.
5627 Details : The client was sending extra request flags in the root getattr
5628 request and did not allocate a reply buffer for the dir EA.
5631 Frequency : occasional, higher if lots of files are accessed by one client
5632 Bugzilla : 6159, 6097
5633 Description: Client trips assertion regarding lsm mismatch/magic
5634 Details : While revalidating inodes the VFS looks up inodes with ifind()
5635 and in rare cases can find an inode that is being freed.
5636 The ll_test_inode() code will free the lsm during ifind()
5637 when it finds an existing inode and then the VFS later attaches
5638 this free lsm to a new inode.
5642 Bugzilla : 6422, 7030
5643 Description: MDS deadlock between mkdir and client eviction
5644 Details : Creating a new file via mkdir or mknod (starting a transaction
5645 and getting the ns lock) can deadlock with client eviction
5646 (gets ns lock and trying to finish a synchronous transaction).
5649 Frequency : occasional
5650 Description: While starting a server, the fsfilt_ext3 module could not be
5652 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5653 kernels. Previously, lconf would still use the ext3 name
5654 when trying to load modules. Now, it will correctly use
5655 ext3 on 2.4 and ldiskfs on 2.6.
5657 Severity : enhancement
5658 Description: The default stripe count has been changed to 1
5659 Details : The interpretation of the default stripe count (0, to lfs
5660 or lmc) has been changed to mean striping across a single
5661 OST, rather than all available. For general usage we have
5662 found a stripe count of 1 or 2 works best.
5664 Severity : enhancement
5665 Description: Add support for compiling against Cray portals.
5666 Details : Conditional compiling for some areas that are different
5670 Frequency : occasional
5671 Bugzilla : 6409, 6834
5672 Description: Creating files with an explicit stripe count may lead to
5673 a failed assertion on the MDS
5674 Details : If some OSTs are full or unavailable, creating files may
5675 trigger a failed assertion on the MDS. Now, Lustre will
5676 try to use other servers or return an error to the
5680 Frequency : occasional
5682 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5683 caused lock timeout during readahead (since 1.4.2).
5684 Details : Processes doing readahead might match a lock that hasn't been
5685 granted yet if there are overlapping and conflicting lock
5686 requests. The readahead process waits on ungranted lock
5687 (original lock is CBPENDING), while OST waits for that process
5688 to cancel CBPENDING read lock and eventually evicts client.
5690 Severity : enhancement
5692 Description: Initial enabling of flock support for clients
5693 Details : Implements fcntl advisory locking and file status functions.
5694 This feature is provided as an optional mount flag (default
5695 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5696 locking are implemented yet, and those that are are not guaranteed
5697 to be completely correct in production environments.
5698 mount -t lustre -o [flock|noflock] ...
5701 Frequency : occasional
5703 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5704 assertion in the ext3 JBD (journaling) layer.
5705 Details : The b_committed_data struct is protected by the big kernel lock
5706 in 2.4 kernels, serializing journal_commit_transaction() and
5707 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5708 access to this struct. In 2.6 kernels there is finer grained
5709 locking to improve SMP performance of the JBD layer.
5713 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5718 Description: merge and cleanup kernel patches.
5721 -----------------------------------------------------------------------------
5723 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5725 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5726 versions of Lustre. Please update all servers and clients to
5727 version 1.4.2 or later at the same time. You must also run
5728 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5729 to update the configuration logs.
5731 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5732 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5733 - swab llog records of type '0' so we get proper header size/idx (5861)
5734 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5735 - fix rename of one directory over another leaking an inode (5953)
5736 - avoid SetPageDirty on 2.6 (5981)
5737 - don't re-add just-being-destroyed locks to the waiting list (5653)
5738 - when creating new directories, inherit the parent's custom
5739 striping settings if present parent (3048)
5740 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5741 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5742 - don't hold client locks on temporary worklist from l_lru (5666)
5743 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5744 - automatically evict dead clients (5921)
5745 - Update file size properly in create+truncate+fstat case (6196)
5746 - Do not unhash mountpoint dentries, do not allow removal of
5748 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5749 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5750 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5751 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5752 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5753 - Correctly handle reads of files with no objects (6243)
5754 - lctl recover will also mark a device active if deactivate used (5933)
5756 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5757 - allow --write-conf on an MDS with different nettype than client (5619)
5758 - don't write config llogs to MDS for mounts not from that MDS (5617)
5759 - lconf should create multiple TCP connections from a client (5201)
5760 - init scripts are now turned off by default; run chkconfig --on
5761 lustre and chkconfig --on lustrefs to use them
5762 - upcalls are no longer needed for clients to recover to failover
5764 - add --abort-recovery option to lconf to abort recovery on device
5766 - add support for an arbitrary number of OSTs (3026)
5767 - Quota support protocol changes.
5768 - forward compatibility changes to wire structs (6007)
5769 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5770 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5771 - improved "lustre status" script
5772 - initialize blocksize for non-regular files (6062)
5773 - added --disable-server and --disable-client configure options (5782)
5774 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5775 - Vanilla 2.4.29 support
5776 - increase maximum number of obd devices to 520 (6242)
5777 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5778 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5779 SLES 9 2.6.5 kernels
5780 - limit stripes per file to 160 (the maximum EA size) (6093)
5782 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5785 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5786 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5787 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5788 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5789 - small ext3 extents cleanups and fixes (5733)
5790 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5791 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5792 - avoid needless client->OST connect, fix handle mismatch (5317)
5793 - fix DLM error path that led to out-of-sync client, long delays (5779)
5794 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5795 - fix several locking issues related to i_size (5492,5624,5654,5672)
5796 - don't move pending lock onto export if it is already evicted (5683)
5797 - fix kernel oops when creating .foo in unlinked directory (5548)
5798 - fix deadlock in obdfilter statistics vs. object create (5811)
5799 - use time_{before,after} to avoid timer jiffies wrap (5882)
5800 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5801 - Do not leak request if server was not able to process it (5154)
5802 - If mds_open unable to find parent dir, make that negative lookup(5154)
5803 - don't create new directories with extent-mapping (5909, 5936)
5805 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5806 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5807 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5808 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5809 - support for RHEL4 kernel on i686 (5773)
5810 - provide error messages when incompatible logs are encountered (5898)
5812 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5813 * version 1.4.0.10 (1.4.1 release candidate 1)
5815 - don't keep a lock reference when lock is not granted (4238)
5816 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5817 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5818 - drop import inflight refcount on signal_completed_replay error (5255)
5819 - unlock page after async write error during send (3677)
5820 - handle missing objects in filter_preprw_read properly (5265)
5821 - no transno return for symlink open, don't save no-trasno open (3440)
5822 - don't try to complete elan receive that already failed (4012)
5823 - free RPC server reply state on error (5406)
5824 - clean up thread from ptlrpc_start_thread() on error (5160)
5825 - readahead could read extra page into cache that wasn't ejected (5388)
5826 - prevent races in class_attach/setup/cleanup/detach (5260)
5827 - don't dereference de->d_inode after l_dput of de (5458)
5828 - use "int" for stripe value returned from lock_to_stripe (5544)
5829 - mballoc allocation and error-checking fixes in 2.6 (5504)
5830 - block device patches to fix I/O request sizes in 2.6 (5482)
5831 - look up hostnames for IB nals (5602)
5832 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5833 - don't start multiple acceptors for the same port (5277)
5834 - fix incorrect LASSERT in mds_getattr_name (5635)
5835 - export a proc file for general "ping" checking (5628)
5836 - fix "lfs check" to not block when the MDS is down (5628)
5838 - service request history (4965)
5839 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5840 - create an "evict_client" /proc entry on OSTs, like the MDS has
5841 - fix mount usage message, return errors per mount(8) (5168)
5842 - change grep [] to grep "[]" in tests so they work in more UMLs
5843 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5844 - remove ancient LOV_MAGIC_V0 EA support (5047)
5845 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5846 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5847 - allow readahead allocations to fail when low on memory (5383)
5848 - mmap locking landed again, after considerable improvement (2828)
5849 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5851 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5854 - send OST transaction number in read/write reply to free req (4966)
5855 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5856 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5857 - the watchdog thread now runs as interruptible (5246)
5858 - flock/lockf fixes (but it's still disabled, pending 5135)
5859 - don't use EXT3 constants in llite code (5094)
5860 - memory shortage at startup could cause assertion (5176)
5862 - reorganization of lov code
5863 - single portals codebase
5865 - add extents/mballoc support (5025)
5866 - direct I/O reads in the obdfilter (4048)
5867 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5869 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5872 - send OST transaction number in read/write reply to free req (4966)
5873 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5874 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5875 - the watchdog thread now runs as interruptible (5246)
5876 - handle missing objects in filter_preprw_read properly (5265)
5877 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5878 - drop import inflight refcount on signal_completed_replay error (5255)
5879 - unlock page after async write error during send (3677)
5880 - return original error code on reconstructed replies (3761)
5881 - no transno return for symlink open, don't save no-trasno open (3440)
5883 - add pid to ldlm debugging output (4922)
5884 - bump the watchdog timeouts -- we can't handle 30sec yet
5885 - extra debugging for orphan dentry/inode bug (5259)
5887 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5890 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5891 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5892 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5893 - fix NULL dereference in /proc/sys/portals/routes (4827)
5894 - allow failed mdc_close() operations to be interrupted (4561)
5895 - stop precreate on OST before MDS would time out on it (4778)
5896 - don't send partial-page writes before EOF from client (4410)
5897 - discard client grant for sub-page writes on large-page clients (4520)
5898 - don't free dentries not owned by NFS code, check generation (4806)
5899 - fix lsm leak if mds_create_objects() fails (4801)
5900 - limit debug_daemon file size, always print CERROR messages (4789)
5901 - use transno after validating reply (3892)
5902 - process timed out requests if import state changes (3754)
5903 - update mtime on OST during writes, return in glimpse (4829)
5904 - add mkfsoptions to LDAP (4679)
5905 - use ->max_readahead method instead of zapping global ra (5039)
5906 - don't interrupt __l_wait_event() during strace
5908 - add software watchdogs to catch hung threads quickly (4941)
5909 - make lustrefs init script start after nfs is mounted
5910 - fix CWARN/ERROR duplication (4930)
5911 - return async write errors to application if possible (2248)
5912 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5913 - print NAL number in %x format (4645)
5914 - update barely-supported suse-2.4.21-171 series (4842)
5915 - support for sles 9 %post scripts
5916 - support for building 2.6 kernel-source packages
5917 - support for sles km_* packages
5919 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5922 - ignore -ENOENT errors in osc_destroy (3639)
5923 - notify osc create thread that OSC is being cleaned up (4600)
5924 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5925 - reconstruct ost_handle() like mds_handle() (4657)
5926 - create a new thread to do import eviction to avoid deadlock (3969)
5927 - let lconf resolve symlinked-to devices (4629)
5928 - don't unlink "objects" from directory with default EA (4554)
5929 - hold socknal file ref over connect in case target is down (4394)
5930 - allow more than 32000 subdirectories in a single directory (3244)
5931 - fix blocks count for O_DIRECT writes (3751)
5932 - OST returns ENOSPC from object create when no space left (4539)
5933 - don't send truncate RPC if file size isn't changing (4410)
5934 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5935 - bind to privileged port in socknal and tcpnal (3689)
5937 - rate limit CERROR/CWARN console message to avoid overload (4519)
5938 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5939 - basic mmap support (3918)
5940 - kernel patch series update from b1_4 (4711)
5942 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5945 - avoid crash during MDS cleanup with OST shut down (2775)
5946 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5947 - don't use bad inodes on the MDS (3744)
5948 - dynamic object preallocation to improve recovery speed (4236)
5949 - don't hold spinlock over lock dumping or change debug flags (4401)
5950 - don't zero obd_dev when it is force cleaned (3651)
5951 - print grants to console if they go negative (4431)
5952 - "lctl deactivate" will stop automatic recovery attempts (3406)
5953 - look for existing locks in ldlm_handle_enqueue() (3764)
5954 - don't resolve lock handle twice in recovery avoiding race (4401)
5955 - revalidate should check working dir is a directory (4134)
5957 - don't always mark "slow" obdfilter messages as errors (4418)
5959 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5962 - don't close LustreDB during write_conf until it is done (3860)
5963 - fix typo in lconf for_each_profile (3821)
5964 - allow dumping logs from multiple threads at one time (3820)
5965 - don't allow multiple threads in OSC recovery (3812)
5966 - fix debug_size parameters (3864)
5967 - fix mds_postrecov to initialize import for llog ctxt (3121)
5968 - replace config semaphore with spinlock (3306)
5969 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5970 - don't allow enqueue to complete on a destroyed export (3822)
5971 - down write_lock before checking llog header bitmap (3825)
5972 - recover from lock replay timeout (3764)
5973 - up llog sem before sending rpc (3652)
5974 - reduce ns lock hold times when setting kms (3267)
5975 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5976 - fix NULL deref and obd_dev leak on setup error (3312)
5977 - replace some LBUG about llog ops with error handling (3841)
5978 - don't match INVALID dentries from d_lookup and spin (3784)
5979 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5980 - fix invalid assertion in ptlrpc_set_wait (3880)
5982 - add libwrap support for the TCP acceptor (3996)
5983 - add /proc/sys/portals/routes for non-root route listing (3994)
5984 - allow setting MDS UUID in .xml (2580)
5985 - print the stack of a process that LBUGs (4228)
5987 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5990 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5991 - make sure to unset replay flag from failed open requests (3440)
5992 - if default stripe count is 0, use OST count for inode size (3636)
5993 - update parent mtime/ctime on client for create/unlink (2611)
5994 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5995 - free recovery state on server during a forced cleanup (3571)
5996 - unregister_reply for resent reqs (3063)
5997 - loop back devices mounting and status check on 2.6 (3563)
5998 - fix resource-creation race that can provoke i_size == 0 (3513)
5999 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6000 - more debugging for page-accounting assertion (3746)
6001 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6002 - avoid deadlock after precreation failure (3758)
6003 - fix race and lock order deadlock in orphan handling (3450, 3750)
6004 - add validity checks when grabbing inodes from l_ast_data (3599)
6006 - add /proc/.../recovery_status to obdfilter (3428)
6007 - lightweight CDEBUG infrastructure, debug daemon (3668)
6008 - change default OSC RPC parameters to be better on small clusters
6009 - turn off OST read cache for files smaller than 32MB
6010 - install man pages and include them in rpms (3100)
6011 - add new init script for (un)mounting lustre filesystems (2593)
6012 - run chkconfig in %post for init scripts (3701)
6013 - drop scimac NAL (unmaintained)
6015 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6018 - clean kiobufs before and after use (3485)
6019 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6020 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6021 - change init script to fail more clearly if not run as root (1528)
6022 - allow clients to reconnect during replay (1742)
6023 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6024 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6025 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6026 - fix cond_resched (3554)
6027 - only evict unfinished clients after recovery (3515)
6028 - allow bulk resend, prevent data loss (3570)
6029 - dynamic ptlrpc request buffer allocation (2102)
6030 - don't allow unlinking open directory if it isn't empty (2904)
6031 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6032 - remove extraneous obd dereference causing LASSERT failure (3334)
6033 - don't use get_cycles() when creating temp. files on the mds (3156)
6034 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6035 - handle EEXIST for set-stripe, set proper directory name (3336)
6037 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6038 - fix ksocknal_fmb_callback() error messages (2918)
6040 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6043 - don't copy lvb into (possibly NULL) reply on error (2983)
6044 - don't deref dentry after dput, don't free lvb on error (2922)
6045 - use the kms to determine writeback rpc length (2947)
6046 - increment oti_logcookies when osc is inactive (2948)
6047 - update client's i_blocks count via lvb messages (2543)
6048 - handle intent open/close of special files properly (1557)
6049 - mount MDS with errors=remount-ro, like obdfilter (2009)
6050 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6051 - don't use cancelling-locks' kms values (2947)
6052 - use highest lock extent for kms, not last one (2925)
6053 - don't dereference ERR_PTR() dentry in error handling path (3107)
6054 - fix thread race in portals_debug_dumplog() (3122)
6055 - create lprocfs device entries at setup instead of at attach (1519)
6056 - common AST error handler, don't evict client on completion race (3145)
6057 - zero nameidata in detach_mnt in 2.6 (3118)
6058 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6059 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6060 - initialize RPC timeout timer earlier for 2.6 (3219)
6061 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6062 - print nal/nid for unknown nid (3258)
6063 - additional checks for oscc recovery before doing precreate (3284)
6064 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6065 - don't crash in mdc_close for bad permissions on open (3285)
6066 - zero i_rdev for non-device files (3147)
6067 - clear page->private before handing to FS, better assertion (3119)
6068 - tune the read pipeline (3236)
6069 - fix incorrect decref of invalidated dentry (2350)
6070 - provide read-ahead stats and refine rpc in flight stats (3328)
6071 - don't hold journal transaction open across create RPC (3313)
6072 - update atime on MDS at close time (3265)
6073 - close LDAP connection when recovering to avoid server load (3315)
6074 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6075 - don't leak open file on MDS after open resend (3325)
6076 - serialize filter_precreate and filter_destroy_precreated (3329)
6077 - loop device shouldn't call sync_dev() for nul device (3092)
6078 - clear page cache after eviction (2766)
6079 - resynchronize MDS->OST in background (2824)
6080 - refuse to mount the same filesystem twice on same mountpoint (3394)
6081 - allow llmount to create routes for mounting behind routers (3320)
6082 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6083 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6084 - fix rare problem with rename on htree directories (3417)
6086 - allow default OST striping configuration per directory (1414)
6087 - fix compilation for qswnal for 2.6 kernels (3125)
6088 - increase maximum number of MDS request buffers for large systems
6089 - change liblustreapi to be useful for external progs like lfsck (3098)
6090 - increase local configuration timeout for slow disks (3353)
6091 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6093 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6096 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6097 - don't overwrite extent policy data in reply if lock was blocked (2901)
6098 - drop filter export grants atomically with removal from device (2663)
6099 - del obd_self_export from work_list in class_disconnect_exports (2908)
6100 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6101 - swab reply message in mdc_close, other PPC fixes (2464)
6102 - fix destroying of named logs (2325)
6103 - overwrite old logs when running lconf --write_conf (2264)
6104 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6105 - fix race in target_handle_connect (2898)
6106 - mds_reint_create() should take same inode create lock (2926)
6107 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6108 - don't close files for self_export to avoid uninitialized obd (2936)
6109 - allow MDS with the same name as client node (2939)
6110 - hold dentry reference for closed log files for unlink (2325)
6111 - reserve space for all logs during transactions (2059)
6112 - don't evict page beyond end of stripe extent (2925)
6113 - don't oops on a deleted current working directory (2399)
6114 - handle hard links to targets without a parent properly (2517)
6115 - don't dereference NULL lock when racing during eviction (2867)
6116 - don't grow lock extents when lots of conflicting locks (2919)
6118 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6121 - account for cache space usage on clients to avoid data loss (974)
6122 - lfsck support in lustre kernel code (2349)
6123 - reduce journal credits needed for BRW writes (2370)
6124 - orphan handling to avoid losing space on client/server crashes
6125 - ptlrpcd can be blocked, stopping ALL progress (2477)
6126 - use lock value blocks to assist in proper KMS, faster stat (1021)
6127 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6128 - recovery for initial connections (2355)
6129 - fixes for mds_cleanup_orphans (1934)
6130 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6131 - block all file creations until orphan recovery completes (1901)
6132 - client remove rq_connection from request struct (2423)
6133 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6134 - recovery timer race (2670)
6135 - mdc_close recovey bug (2532)
6136 - ptlrpc cleanup bug (2710)
6137 - mds timeout on local locks (2588)
6138 - namespace lock held during RPCs (2431)
6139 - handle interrupted sync write properly (2503)
6140 - don't try to handle a message that hasn't been replied to (2699)
6141 - client assert failure during cleanup after abort recovery (2701)
6142 - leak mdc device after failed mount (2712)
6143 - ptlrpc_check_set allows timedout requests to complete (2714)
6144 - wait for inflight reqs when ptlrpcd finishes (2710)
6145 - make sure unregistered services are removed from the srv_list
6146 - reset bulk XID's when resending them (caught by 1138 test)
6147 - unregister_bulk after timeout
6148 - fix lconf error (2694)
6149 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6150 - readahead locks pages, leaves pending causing memory pressure (2673)
6151 - increase OST request buffers to 4096 on large machines (2729)
6152 - fix up permission of existing directories in simple_mkdir (2661)
6153 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6154 - don't assign transno to errored transactions (2742)
6155 - don't delete objects on OST if given a bogus objid from MDS (2751)
6156 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6157 - if rq_no_resend, then timeout request after recovery (2432)
6158 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6159 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6160 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6161 - MDS recovery shouldn't delete objects that it has given out (2730)
6162 - if enqueue arrives after completion, don't clobber LVB (2819)
6163 - don't unlock pages twice when trigger_group_io returns error (2814)
6164 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6165 - don't write pages to disk if there was an error (1450)
6166 - don't ping imports that have recovery disabled (2676)
6167 - take buffered bytes into account when balancing socknal conn (2817)
6168 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6169 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6170 - remove little-endian swabbing of llog records (1987)
6171 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6172 - retry reposting request buffers if they fail (1191)
6173 - grow extent at grant time to avoid granting a revoked lock (2809)
6174 - lock revoke doesn't evict page if covered by a second lock (2765)
6175 - disable VM readahead to avoid reading outside lock extents (2805)
6177 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6178 - updated kernel patches for hp-2.4.20 kernel (2681)
6180 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6183 - fix truncated write corruption (2366)
6184 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6186 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6187 - ensure len(uuid) < 37 in lmc (1171)
6188 - fix ia64 OOPS in llog_test (2255)
6189 - zero end of page at obdfilter for partial page writes (2648)
6190 - don't leave stale dentries around after renames (bug 2428)
6191 - fix timeouts when evicting a client with a single lock held (2642)
6192 - set deadline for the initial HELLO message to drain (2634)
6193 - print out dotted-quad IP addresses in the socknal (2302)
6195 - additional debugging for MDS client eviction problem (2443)
6196 - fix mkfsoptions support for osts (2603, 2604)
6198 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6201 - add series for the vanilla 2.6.0 kernel
6202 - add series for the vanilla 2.4.24 kernel
6203 - add series for a cray x86/64 UL kernel drop
6204 - fix xattr patches for the vanilla 2.4.19 series
6206 - generate true UUIDs in lmc (1171)
6207 - have portals stack dumping break in UML (2466)
6208 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6209 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6210 - wake callback waiting threads on client eviction (2460)
6211 - Add --ptldebug and --subsystem to lmc (1719)
6212 - update assertion to allow safe interrupt allocation
6213 - set rq_no_resend for cancel requests (2432)
6214 - recalculate ptlrpcd timeout after resend (2494)
6215 - call vfs_rmdir when removing pending directories (2368)
6216 - fix renaming a file to itself (2429)
6217 - lmc creates a default one-stripe lov (2454)
6218 - expand procfs space to handle large clusters (2326)
6219 - increase UML stack to avoid overflow
6220 - update lconf's list of debug and subsystem masks
6221 - fix lfs find --obd (2510)
6222 - /proc tunable for disabling filter read caching (2591)
6223 - stop rpm packages from altering slapd.conf (2301)
6224 - disable nagle in the socknal under 0conf (2578)
6225 - choose mds inode size based on stripe count (2572)
6226 - fix kernel-source rpm problems (2516)
6228 - add --disable-doc to avoid pdf generation (2421)
6229 - update documentation, tests, type-os, comments
6230 - avoid format warnings on ia64
6231 - remove the TOE NAL
6232 - tiny code cleanups by removing unused fields
6234 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6237 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6238 - remove the most blatant lies from BUILDING file (2371)
6239 - change default debug level to reasonable production setting
6240 - reduce client side cache size to reduce cache flush time
6241 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6242 - make TCP zerocopy and pinger support enabled by default (2476)
6243 - sync writes completed after process exits caused crashes (2319)
6244 - maintain correct mount count on the MDS (2356)
6245 - backout 1557, because 2316 wasn't really fixed
6246 - better file I/O statistics gathering in /proc
6247 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6248 - another kernel patch to fix zero-copy TCP function export
6249 - don't take duplicate lock when processing re-sent getattr (2420)
6250 - lctl uses obd_self_export instead of creating new conn (2353)
6251 - MDS/OST recovery case which requires object creation asserted (2425)
6252 - move lfs from /usr/sbin to /usr/bin in packages
6253 - fix race between mds_client_add and mds_client_free (2417)
6254 - use kmalloc instead of slabs in portals (2430)
6255 - don't create duplicate records when a failover MDS is present (2442)
6256 - remove unnecessary mount age check (2332)
6257 - don't remove directory inodes from locks prematurely (2451)
6258 - don't break if MDS service name is the same as hostname (2103)
6259 - fix races in client write RPC generation when cache full (2482)
6261 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6264 - remove now-unused request->rq_obd (278)
6265 - if an allocation fails, print out how much memory we've used (1933)
6266 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6267 - add the "configurable stack size" patch to most series files (1256)
6268 - ability to write large log records, for 100+ OST configs (2306)
6269 - fix NULL deref when filter_prep fails (2314)
6270 - fix operator precedence error in filter_sync
6271 - dynamic allocation of socknal TX descriptors (2315)
6272 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6273 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6274 - ensure that CWARN messages in Portals always get to the syslog
6275 - __init/__exit are not for prototype decls (ldlm_init/exit)
6276 - x86-64 compile warning fixes
6277 - fix gateway LMC keyword conflict (2318)
6278 - fix MDS lock inversions in getattr/reint paths (1844)
6279 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6280 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6281 - DLM race condition prevented some lock evictions (2328)
6282 - ENOMEM detection and retry on socknal sends (2230)
6283 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6284 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6285 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6286 - better error messages when a client is rejected during recovery (1505)
6287 - avoid cancelling locks which were never granted, after failure (2330)
6288 - fix i_sem/journal inversion in mds_client_add (2333)
6289 - fix truncate/getattr lock cycle deadlock (2334)
6290 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6291 - fix two rare exit paths which could leak an l_lock() ref (2321)
6292 - fencepost error in MDS/OST orphan recovery (2226)
6293 - make log record alignment 8 bytes (1988)
6294 - lstripe now fails when requested offset > ost_count (2237)
6295 - ensure that all kernel series have a complete list.h (1607)
6296 - fix crashes in special-file operations (2316)
6297 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6299 - allow configurable automake binary, for testing new versions
6300 - small update to the lfs documentation
6302 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6304 * fix negative export reference count in fsfilt_sync (2312)
6306 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6307 * release candidate 0.9.1
6309 - orphans are moved into the PENDING directory for possible recovery
6310 - replayed opens now open by fid for orphan/rename safety (1042)
6311 - last close of an orphan inode generates a transno (683)
6312 - chdir() and mount() now pin the directory entry (1020)
6313 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6314 - discard very old requests without processing them (1502)
6315 - remove some common, well-understood CERRORs (1505)
6316 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6317 - clear "grant" flags in OST replies until OST grant code lands (1644)
6318 - fix read performance by not clobbering i_blksize on client (1598)
6319 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6320 - make LRU size a /proc tunable, clears locks when reduced (707)
6321 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6322 - server locks take references on exports now (1558)
6323 - build fixes for 2.4.20-rh trees (1663)
6324 - return an error from lov_create if all OSCs are inactive (1751)
6325 - fix import levels when a reconnect happens without a timeout (1597)
6326 - exit early from mds_open if we get a lookup error (1749)
6327 - partial page read at EOF wouldn't wait for disk before sending (1642)
6328 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6329 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6330 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6331 - write/truncate lock inversion (1639)
6332 - Don't auto-load obdclass, portals modules during cleanup (1495)
6333 - fix timestamps from jumping to "now" (1763)
6334 - extra journal assertions (1648)
6335 - add an extra multiunlink test (1771)
6336 - fix read_record/write_record API (1776)
6337 - fix leak of offset_extent, possible incorrect i_size later (1772)
6338 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6339 - Debugging for the kqswnal_get_idle_tx problems (1820)
6340 - Allow recovery to be attempted multiple times (1536)
6341 - Write out MDS last_rcvd file after it is first created (1600)
6342 - Fix tx_descriptor leak in failed transmit situations (1827)
6343 - ext3 journaling fixes for assertion failure after IO error (1871)
6344 - class_export_put() on freed export after completion AST error (1896)
6345 - Fix revalidate looping in VFS (1322)
6346 - Don't access a freed export during MDS_REINT timeout (1521)
6347 - Add open-unlink recovery support on the MDS (1673,1764)
6348 - Return an error if no MDS data was read from last_rcvd (1946)
6349 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6350 - Don't setup a disk device that doesn't match exported UUID (317)
6351 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6352 - avoid committing NULL handle in force close
6353 - local.sh is now a one-stripe LOV configuration
6354 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6355 - fix ext3 htree duplicate directory entry corruption (1516)
6356 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6357 - update to new LOV EA format (2097)
6358 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6360 2003-06-15 Phil Schwan <phil@clusterfs.com>
6363 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6364 - per-import/export recovery handling (958, 931, 959)
6365 - multiple last-rcvd slots, for serving multiple FSes (949)
6366 - connections are again shared between multiple imp/exports (963, 964)
6367 - "umount -f" would hang if any requests needed to be sent (393, 978)
6368 - avoid pinning large req buffer by copying for queued messages (989)
6369 - add "uuid" to "lctl device" command to help upcalls (991)
6370 - "open" RPCs with transnos would confuse recovery counters (1037)
6371 - do proper endian conversion of all wire messages (288, 340, 891)
6372 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6373 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6374 - fix ptlrpc_connection leak in target_handle_connect (1174)
6375 - fix import refcounting bug in OST and MDS cleanup (1134)
6376 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6377 - fix very unlikely obd_types race condition (501)
6378 - remove osc_open hack for echo_client (1187)
6379 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6380 - a failure in read_inode2 leads to deadlock (1139)
6381 - cancel ack-locks as soon as transaction is committed (1072)
6382 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6383 - make sure to commitrw after any preprw to avoid deadlock (1162)
6384 - failing to execute a file in a lustre FS would lock inode (1203)
6385 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6386 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6387 - obd_uuid2tgt would walk off the end of the list (1255)
6388 - on IA64 the osc would give portals incorrect bulk size (1258)
6389 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6390 - fix lock inversion caused by new llite matching code (1282)
6391 - limit the number of dirty pages on a client to 10MB (1286)
6392 - timed out locks were not being corrected cancelled (1289)
6393 - fix O_DIRECT above 4GB on IA-32 (1292)
6394 * major user-visible changes
6395 - fail out/fail over policy now controlled by the upcall (993)
6397 - add OBD_PING to check server availability and failure (954)
6398 - lustre messages are now sent in sending host order (288, 340, 891)
6399 - add eadatalen to MDS getattr reply (340)
6400 - OST read replies may contain second buffer, with per-page status (593)
6402 2003-03-11 Phil Schwan <phil@clusterfs.com>
6405 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6406 - failed open()s could cause deadlock; fixed (867, 869)
6407 - stop cancelling OST locks when files are closed (481)
6408 - overlapping XID spaces caused network corruption (851, 853)
6409 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6410 - setattr_raw added, to do single-RPC, server-side setattrs
6411 - lmc/lconf syntax change for OST UUIDs
6412 - fix crashy race condition between ptlrpc_free_req and osc_close
6413 - don't use request in mdc_enqueue if we hit a timeout (889)
6414 - don't set the inode i_size for regular files from the MDS (896)
6415 - handle out of order completion AST (842)
6416 - don't LBUG if a lock request times out after receiving AST (913)
6417 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6418 - if a bad lock AST arrives, send an error instead of dropping entirely
6419 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6420 - fix leak in bulk IO when only partially completed (899, 900, 926)
6421 - fix O_DIRECT for ia64 (55)
6422 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6423 - C-z after timeout could hang a process forever; fixed (977)
6425 - client-side I/O cache (678, 924, 929, 941, 970)
6427 - READPAGE and SETATTRs which don't take server-side locks get
6430 2003-02-11 Phil Schwan <phil@clusterfs.com>
6433 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6434 - Fix fsfilt_extN_readpage() to read a full page of directory
6435 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6436 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6437 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6438 - Most intent operations were removed, in favour of a new RPC mode
6439 that does a single RPC to the server and bypasses most of the VFS
6440 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6441 - Aggressively cancel local locks on DLM servers
6442 - mds_reint_unlink sends EA to the client if it's the last nlink.
6443 client uses that EA to unlink OST objects.
6444 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6445 - recursive symlinks were fixed (439)
6446 - fixed NULL deref in DEBUG_REQ
6447 - filter_update_lastobjid no longer calls sync, which annoyed extN
6448 - fixed multi-client small-writes to a single file problem (445)
6449 - fixed mtime updates during file writes (607)
6450 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6451 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6452 - fixed Linux OST size reporting problem (444, 656)
6453 - OST now updates object mtime with writes or setattr (607, 619)
6454 - client verifies file size before zeroing page past EOF (445)
6455 - OST now writes last allocated objid to disk with allocation (108)
6456 - LOV on echo now works (409)
6458 - mds_reint_unlink sends a new buffer, with the EA included. this
6459 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6460 set if a regular file was being unlinked, and it was the last link
6461 - use PtlGet from the target for bulk writes (315)
6462 - OST now updates object mtime with writes or setattr (607, 619)
6463 - LDLM now has a grant-time callback to revalidate locked items, if
6465 - Many MDS operations were reorganized to combat race conditions
6467 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6468 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6470 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6473 - Fully reactivate OST imports after reconnection (512, others)
6474 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6475 - More graceful error handling for truncating on dead OST (515)
6476 - Don't error out unless we're actually accessing dead stripes (474)
6477 - Fix garbage sizes when stripes are missing (410)
6478 - LRU counters were broken, causing constant lock purge (433, 432)
6479 - garbage on read from stripes with failed OSTs (441)
6480 - mark OSCs as active before reconnecting during recovery (438)
6481 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6482 - lfind did not preserve OST order in output (443)
6483 - symlinks cause hung clients, incorrect data (439)
6484 - stop dereferencing request after dropping refcount (457)
6485 - don't LASSERT(spin_is_locked) on non-SMP (455)
6486 - fixes for many rename() bugs
6487 - fstat didn't correctly synchronize attributes (399)
6488 - server must handle lock cancellation during blocking AST prep (487)
6489 - bulk descriptors were free()d too soon (511)
6490 - fix paths in lconf, which would load incorrect modules (451, 507)
6491 - fix confusing lconf 'host not found' error message (386)
6492 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6493 - fix race condition in mdc_blocking_ast() for inode access (526)
6494 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6495 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6496 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6497 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6498 - fix lstripe in several ways (kernel side) (527)
6499 - fix request leak in ldlm_cli_enqueue (262)
6500 - incorrect OSC was marked inactive after OST failure
6501 - call mds_fs_cleanup before unmounting filesystem (524)
6502 - fix races between taking ns_lock and ldlm_lock_change_resource
6503 - fix races updating LOV export open file list
6504 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6505 - fix recovery NULL deref in ldlm_cli_cancel_unused
6506 - fix some DLM races by using new hash table for lock handles (419)
6507 - permit the client to specify desired inodes, at replay
6508 - duplicate requests when we queue them for replay reintegration
6509 - fix last_rcvd offset calculation
6510 - sync after each recovered transaction, so we always make progress
6511 - never, not always, ERESTART requests without transnos
6512 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6514 - skip replay if the MDS says that the client is already connected
6515 - don't check for a recovery-enabled export to match lctl's UUID
6516 - don't INC_USE_COUNT for phantom exports
6517 - don't crash when cleaning up phantom exports (567)
6518 - don't double-finish or set replay data for errored mdc_open requests
6519 - abort requests when they time out, so we don't get old replies
6520 - send/receive replies for AST messages again
6521 - if the client says that it doesn't have the lock, cancel it on the
6523 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6524 mark it as destroyed, it will all work out when decref is called
6525 - fix module use counts (22, 581)
6527 - ASTs now expect a reply (server cancels lock on error reply)
6529 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6532 - fix many simultaneous client startup (392)
6533 - fix dentry->d_it clobbering
6534 - credentials weren't being shipped for readdir/getattr operations
6535 - remove invalid assertions triggered during some concurrent MD
6537 - proper Lustre versions added (336, 389)
6538 - fix memory leak for create error case (398)
6539 - fix LOV locking bug that would get cli/srv out of sync
6540 - fix echo client over LOV (409)
6541 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6542 - fix double-O_EXCL intent crash (424)
6543 - avoid sending multiple lock CANCELs (352)
6545 - MDS can do multi-client recovery (modulo bugs in new code)
6547 - many updates, edits, cleanups
6549 2002-11-18 Phil Schwan <phil@clusterfs.com>
6552 - fix null d_it dereference (346)
6553 - fix full OST/dbench hang (333)
6554 - fix permission problem with file removal (286)
6555 - fix removal of OSCs from LOV when they fail
6556 - fix NULL deref during bulk timeout (214)
6557 - fix problems related to multiple filesystems on one MDS (241)
6558 - fixed serious subtle metadata locking bugs
6559 - free locks on clients when inodes are removed due to memory
6561 - fix inode pointer in lock data (285)
6562 - partial support for multiple MDS on a single host (241)
6563 - data locks weren't cancelled at clear_inode time (290, 311)
6564 - intent locks could lead to unbounded lock growth (205)
6565 - added a maximum lock count, an LRU list, and a flusher
6566 - fix multiple rename (365)
6567 - properly abstracted the echo client
6568 - OSC locked 1 byte too many; fixed
6569 - rewrote brw callback code:
6570 - fixed recovery bugs related to LOVs (306)
6571 - fixed too-many-pages-in-one-write crash (191)
6572 - fixed (again) crash in sync_io_timeout (214)
6573 - probably fixed callback-related race (385)
6575 - Add capability to MDS protocol
6576 - LDLM cancellations and callbacks on different portals
6578 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6581 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6582 - defer open object destruction to close time (601981, 138)
6583 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6584 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6585 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6586 - don't call obd_setattr in ll_file_release for destroyed objects
6587 * protocol change to lustre_msg: move |version| and add |flags|
6588 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6589 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6590 * added replay of create, unlink, link and rename operations during
6591 MDS failover; recovery should be much more robust now
6592 * remove failed OSCs from LOVs (only lov_create uses this so far)
6593 * the lustre-HOWTO was brought (more) up to date (582544)
6595 2002-10-23 Phil Schwan <phil@clusterfs.com>
6598 - in-use dentries weren't being reused properly (617851)
6599 - prevent multiple LDLM setup (599178)
6600 - fix LOV size calculations for truncate (617853)
6601 - fix client handling of MDS intent errors (POSIX)
6602 - fix permission bug in lovstripe.c test (624321)
6603 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6604 - truncate past end of file could corrupt data
6605 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6606 - a race in recovery could return ETIMEDOUT to apps (623947)
6607 - building outside the source directory was fixed
6608 * the lustre-HOWTO was brought (more) up to date (582544)
6609 * major progress was made on recovery functionality
6611 2002-10-10 Phil Schwan <phil@clusterfs.com>
6614 - recovery deadlock fix
6615 - rm -rf causes LBUG fix (617817)
6616 - file open by multiple tasks fix (618962)
6617 - directory permissions bugs (602707 and 620007)
6618 - journal_stop fixed with locking (611313)
6619 - O_APPEND failures resolved (618273, perhaps 614459)
6620 - lconf PATH fix (619770)
6621 - IA64 build fix (621450)
6622 - RPC buffer sizes scale with amount of memory
6624 2002-10-01 Phil Schwan <phil@clusterfs.com>
6627 - locks would be cancelled without throwing away data pages,
6628 resulting in inconsistent data (605627)
6629 - inode attributes were not always being refreshed (605627, 612449)
6630 - lconf now continues to cleanup after lctl reports an error
6631 - MDS now enforces user permissions (602707)
6632 - lprocfs cleanup fixed, but not yet enabled (614157)
6633 - fixed infinite server hang, should a client not respond to an AST
6634 - avoid going into recovery if user calls readlink() with a buffer
6635 that's too small (613941)
6636 - AST RPCs no longer require replies (614867) -- this may be changed
6637 - don't crash server if client sends an IOV that's too big (611336)
6638 - fixed lock conversion deadlock (611892)
6639 - fixed the following of symlinks (614622)
6640 * recovery: the server can remove locks from a client that dies, other
6641 clients can make progress
6642 * more extN patch fixes
6643 * compile-time configurable ptlrpc buffer allocations
6645 - collaborative read cache document
6646 - Lustre Lite Performance CDR document-in-progress
6648 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6651 - fix typo in patch-2.4.18
6653 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6656 - clear ptlrpc request each time in handle_incoming_request()
6657 - unlink of files now destroys the object on the OST
6659 2002-09-19 Peter Braam <braam@clusterfs.com>
6661 * add hard link support
6662 * change obdfile creation method
6663 * kernel patch changed
6665 2002-09-19 Peter Braam <braam@clusterfs.com>
6668 - stack overflow bug in extN fixed
6670 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6672 * documentation updates
6673 - add man pages for config tools
6674 - update tests/README to describe testing with new config tools
6675 - finish metadata API descriptions
6676 * bug fixes and cleanups
6677 - statfs workaround for 16TB limit
6678 - LOV stripe allocation improved, can stripe on subset of OSTs
6679 - LOV file size/IO offset was wrong for files > 4GB in size
6680 - object EA data was being dropped, caused files to be unreadable
6681 - memory overflow with non-LOV OST caused memory corruption
6682 - fixed regression tests to work with new config tools, obdfilter
6683 - fixed bug when directory size became larger than 1 block
6684 - fixed bug (for single client case) when PWD was deleted
6685 - invalidate local directory pages when doing intent-based ops
6686 - avoid LDLM oops when lock callback contained bad data
6688 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6690 * documentation updates
6691 * bug fixes and cleanups
6692 - configuration tools
6695 - 64-bit compile warnings
6696 - 64-bit internal statfs data
6698 * test_brw on persistent OST devices
6700 * lprocfs (disabled)
6702 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6704 * documentation updates
6705 * bug fixes and cleanups
6706 * configuration tools
6708 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6711 * many small fixes to 0_5_4
6712 * io/network handling
6713 * thinkos in MDS operations
6715 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6718 * crucial basic fixes to 0.5.3
6719 * IOR, Iozone work over Elan
6722 2002-08-07 Phil Schwan <phil@clusterfs.com>
6723 * version 0_5_3, our first alpha
6724 * we use the new Portals iovs
6725 * documentation updates
6726 * bug fixes and cleanups
6727 * small changes in the DLM wire protocol
6729 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6730 * version 0_5_1 with some initial stability,
6731 * locking on MD and file I/O.
6732 * documentation updates
6733 * several bug fixes since 0.5.0
6734 * small changes in wire protocol
6736 2002-07-18 Phil Schwan <phil@clusterfs.com>
6738 * delivered as Lustre Light Alpha
6739 * fixed a crash after handling invalid MDS requests
6740 * fixed directory pages for architectures with non-4k pages sizes
6742 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6743 * release version v0_4_4
6744 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6746 * Fixes a number of interruption problems with OST operations.
6747 * Update documentation for portals header changes
6748 * Move all wire protocol structs/defines to lustre_idl.h
6749 * Fixes symlink length bug.
6750 * Add tcpdump to repository.
6752 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6753 * release version v0_4_3
6754 * Fixes statfs for inodes on extN.
6755 * Fixes bug in runtests which would delete /etc/hosts.
6756 * Use 64-bit object IDs wherever possible (not into VFS though)
6757 Remove ost_get_info, which is unused by lustre, and out of date.
6759 2002-07-03 Peter Braam <braam@clusterfs.com>
6760 * release version v0_4_2 Fixes a lookup error (type not passed)
6761 * move forward to head of Portals
6762 * move forward to latest Lustre kernel
6764 2002-06-25 Peter Braam <braam@clusterfs.com>
6765 * release version v0_4_1. Hopefully stable on single node use.