3 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
4 for currently supported client and server kernel versions.
5 * Server primary kernels built and tested during release cycle:
6 3.10.0-1062.18.1.el7 (RHEL7.7)
7 3.10.0-1127.19.1.el7 (RHEL7.8)
8 3.10.0-1160.6.1.el7 (RHEL7.9)
9 * Other server kernels known to build and work at some point (others may also work):
10 3.10.0-862.14.4.el7 (RHEL7.5)
11 3.10.0-957.27.2.el7 (RHEL7.6)
12 4.18.0-80.11.2.el8 (RHEL8.0)
13 4.18.0-147.8.1.el8 (RHEL8.1)
14 4.18.0-193.28.1.el8 (RHEL8.2)
15 4.18.0-240.1.1.el8 (RHEL8.3)
16 4.4.120-92.70 (SLES12 SP2)
17 4.4.180-94.100 (SLES12 SP3)
18 4.4.0-131 (Ubuntu 16.04)
19 4.15.0-32 (Ubuntu 18.04)
20 5.4.0-48 (Ubuntu 20.04)
21 vanilla linux 5.4.0 (ZFS + ldiskfs)
22 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
23 * Client primary kernels built and tested during release cycle:
24 3.10.0-1062.18.1.el7 (RHEL7.7)
25 3.10.0-1127.19.1.el7 (RHEL7.8)
26 3.10.0-1160.6.1.el7 (RHEL7.9)
27 4.12.14-95.48 (SLES12 SP4)
28 4.12.14-122.20 (SLES12 SP5)
29 4.12.14-197.75 (SLES15 SP1)
30 5.3.18-24.24 (SLES15 SP2)
31 4.15.0-48 (Ubuntu 18.04)
32 * Other clients known to build on these kernels at some point (others may also work):
33 3.10.0-862.14.4.el7 (RHEL7.5)
34 3.10.0-957.27.2.el7 (RHEL7.6)
35 4.14.0-49.13.1.el7a (RHEL7.5)
36 4.14.0-115.2.2.el7a (RHEL7.6)
37 4.18.0-80.11.2.el8 (RHEL8.0)
38 4.18.0-147.8.1.el8 (RHEL8.1)
39 4.18.0-193.28.1.el8 (RHEL8.2)
40 4.18.0-240.1.1.el8 (RHEL8.3)
41 4.4.120-92.70 (SLES12 SP2)
42 4.4.180-94.100 (SLES12 SP3)
43 4.4.0-131 (Ubuntu 16.04)
44 5.4.0-37 (Ubuntu 20.04)
45 5.4.0 (vanilla kernel.org)
46 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
47 * Recommended ZFS version: 2.0.0
48 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
49 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
50 stack overflow. For more information, please refer to bugzilla 17630.
51 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
52 the re-exporting nodes to be mounted with "32bitapi" mount option
54 --------------------------------------------------------------------------------
58 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
59 for currently supported client and server kernel versions.
60 * Server primary kernels built and tested during release cycle:
61 3.10.0-1062.18.1.el7 (RHEL7.7)
62 3.10.0-1127.19.1.el7 (RHEL7.8)
63 3.10.0-1160.6.1.el7 (RHEL7.9)
64 * Other server kernels known to build and work at some point (others may also work):
65 3.10.0-862.14.4.el7 (RHEL7.5)
66 3.10.0-957.27.2.el7 (RHEL7.6)
67 4.18.0-80.11.2.el8 (RHEL8.0)
68 4.18.0-147.8.1.el8 (RHEL8.1)
69 4.18.0-193.28.1.el8 (RHEL8.2)
70 4.18.0-240.1.1.el8 (RHEL8.3)
71 4.4.120-92.70 (SLES12 SP2)
72 4.4.180-94.100 (SLES12 SP3)
73 4.4.0-131 (Ubuntu 16.04)
74 4.15.0-32 (Ubuntu 18.04)
75 5.4.0-48 (Ubuntu 20.04)
76 vanilla linux 5.4.0 (ZFS + ldiskfs)
77 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
78 * Client primary kernels built and tested during release cycle:
79 3.10.0-1062.18.1.el7 (RHEL7.7)
80 3.10.0-1127.19.1.el7 (RHEL7.8)
81 3.10.0-1160.6.1.el7 (RHEL7.9)
82 4.12.14-95.48 (SLES12 SP4)
83 4.12.14-122.20 (SLES12 SP5)
84 4.12.14-197.75 (SLES15 SP1)
85 5.3.18-24.24 (SLES15 SP2)
86 4.15.0-48 (Ubuntu 18.04)
87 * Other clients known to build on these kernels at some point (others may also work):
88 3.10.0-862.14.4.el7 (RHEL7.5)
89 3.10.0-957.27.2.el7 (RHEL7.6)
90 4.14.0-49.13.1.el7a (RHEL7.5)
91 4.14.0-115.2.2.el7a (RHEL7.6)
92 4.18.0-80.11.2.el8 (RHEL8.0)
93 4.18.0-147.8.1.el8 (RHEL8.1)
94 4.18.0-193.28.1.el8 (RHEL8.2)
95 4.18.0-240.1.1.el8 (RHEL8.3)
96 4.4.120-92.70 (SLES12 SP2)
97 4.4.180-94.100 (SLES12 SP3)
98 4.4.0-131 (Ubuntu 16.04)
99 5.4.0-37 (Ubuntu 20.04)
100 5.4.0 (vanilla kernel.org)
101 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
102 * Recommended ZFS version: 2.0.0
103 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
104 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
105 stack overflow. For more information, please refer to bugzilla 17630.
106 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
107 the re-exporting nodes to be mounted with "32bitapi" mount option
109 --------------------------------------------------------------------------------
113 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
114 for currently supported client and server kernel versions.
115 * Server primary kernels built and tested during release cycle:
116 3.10.0-1062.1.1.el7 (RHEL7.7)
117 * Other server kernels known to build and work at some point (others may also work):
118 3.10.0-862.14.4.el7 (RHEL7.5)
119 3.10.0-957.27.2.el7 (RHEL7.6)
120 4.18.0-80.7.1.el8 (RHEL8.0)
121 4.4.120-92.70 (SLES12 SP2)
122 4.4.180-94.100 (SLES12 SP3)
123 4.4.0-131 (Ubuntu 16.04)
124 4.15.0-32 (Ubuntu 18.04)
125 vanilla linux 4.6.7 (ZFS only)
126 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
127 * Client primary unpatched kernels built and tested during release cycle:
128 3.10.0-1062.1.1.el7 (RHEL7.7)
129 4.12.14-95.29 (SLES12 SP4)
130 4.15.0-48 (Ubuntu 18.04)
131 * Other clients known to build on unpatched kernels at some point (others may also work):
132 3.10.0-862.14.4.el7 (RHEL7.5)
133 3.10.0-957.27.2.el7 (RHEL7.6)
134 4.14.0-49.13.1.el7a (RHEL7.5)
135 4.14.0-115.2.2.el7a (RHEL7.6)
136 4.18.0-80.7.1.el8 (RHEL8.0)
137 4.4.120-92.70 (SLES12 SP2)
138 4.4.180-94.100 (SLES12 SP3)
139 4.4.0-131 (Ubuntu 16.04)
140 vanilla linux 4.15.0 (ZFS only)
141 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
142 * Recommended ZFS / SPL version: 0.7.13
143 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
144 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
145 stack overflow. For more information, please refer to bugzilla 17630.
146 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
147 the re-exporting nodes to be mounted with "32bitapi" mount option
148 * Note for Ubuntu 14.04.5:
149 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
150 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
152 --------------------------------------------------------------------------------
156 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
157 for currently supported client and server kernel versions.
158 * Server known to build on the following kernels (others may also work):
159 2.6.32-696.18.7.el6 (RHEL6.9)
160 3.10.0-862.14.4.el7 (RHEL7.5)
161 3.10.0-957.el7 (RHEL7.6)
162 4.4.120-92.70 (SLES12 SP2)
163 4.4.162-94.69 (SLES12 SP3)
164 4.4.0-131 (Ubuntu 16.04)
165 4.15.0-32 (Ubuntu 18.04)
166 vanilla linux 4.6.7 (ZFS only)
167 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
168 * Client known to build on unpatched kernels (others may also work):
169 2.6.32-696.18.7.el6 (RHEL6.9)
170 2.6.32-754.6.3.el6 (RHEL6.10)
171 3.10.0-862.14.4.el7 (RHEL7.5)
172 3.10.0-957.el7 (RHEL7.6)
173 4.14.0-49.13.1.el7a (RHEL7.5)
174 4.14.0-115.2.2.el7a (RHEL7.6)
175 4.4.120-92.70 (SLES12 SP2)
176 4.4.162-94.69 (SLES12 SP3)
177 4.4.0-131 (Ubuntu 16.04)
178 4.15.0-32 (Ubuntu 18.04)
179 vanilla linux 4.15.0 (ZFS only)
180 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
181 * Recommended ZFS / SPL version: 0.7.9
182 * Tested with ZFS / SPL version: 0.7.9
183 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
184 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
185 stack overflow. For more information, please refer to bugzilla 17630.
186 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
187 the re-exporting nodes to be mounted with "32bitapi" mount option
188 * Note for Ubuntu 14.04.5:
189 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
190 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
192 --------------------------------------------------------------------------------
194 2018-04-03 Intel Corporation
196 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
197 for currently supported client and server kernel versions.
198 * Server known to build on patched kernels:
199 2.6.32-431.29.2.el6 (RHEL6.5)
200 2.6.32-504.30.3.el6 (RHEL6.6)
201 2.6.32-573.26.1.el6 (RHEL6.7)
202 2.6.32-642.15.1.el6 (RHEL6.8)
203 2.6.32-696.18.7.el6 (RHEL6.9)
204 3.10.0-693.21.1.el7 (RHEL7.4)
205 3.0.101-0.47.71 (SLES11 SP3)
206 3.0.101-107 (SLES11 SP4)
207 3.12.74-60.64.40 (SLES12 SP1)
208 4.4.114-92.64 (SLES12 SP2)
209 4.4.114-94.11 (SLES12 SP3)
210 3.13.0-101 (Ubuntu 14.04, ZFS only)
211 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
212 4.4.0-116 (Ubuntu 16.04)
213 vanilla linux 4.6.7 (ZFS only)
214 * Client known to build on unpatched kernels:
215 2.6.32-431.29.2.el6 (RHEL6.5)
216 2.6.32-504.30.3.el6 (RHEL6.6)
217 2.6.32-573.26.1.el6 (RHEL6.7)
218 2.6.32-642.15.1.el6 (RHEL6.8)
219 2.6.32-696.18.7.el6 (RHEL6.9)
220 3.10.0-693.21.1.el7 (RHEL7.4)
221 3.0.101-0.47.71 (SLES11 SP3)
222 3.0.101-107 (SLES11 SP4)
223 3.12.74-60.64.40 (SLES12 SP1)
224 4.4.114-92.64 (SLES12 SP2)
225 4.4.114-94.11 (SLES12 SP3)
226 3.13.0-101 (Ubuntu 14.04)
227 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
228 4.4.0-116 (Ubuntu 16.04)
230 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
231 * Recommended ZFS / SPL version: 0.7.6
232 * Tested with ZFS / SPL version: 0.7.6
233 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
234 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
235 stack overflow. For more information, please refer to bugzilla 17630.
236 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
237 the re-exporting nodes to be mounted with "32bitapi" mount option
238 * Note for Ubuntu 14.04.5:
239 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
240 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
242 --------------------------------------------------------------------------------
244 2017-07-05 Intel Corporation
246 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
247 for currently supported client and server kernel versions.
248 * Server known to build on patched kernels:
249 2.6.32-431.29.2.el6 (RHEL6.5)
250 2.6.32-504.30.3.el6 (RHEL6.6)
251 2.6.32-573.26.1.el6 (RHEL6.7)
252 2.6.32-642.15.1.el6 (RHEL6.8)
253 2.6.32-696.3.1.el6 (RHEL6.9)
254 3.10.0-514.21.1.el7 (RHEL7.3)
255 3.0.101-0.47.71 (SLES11 SP3)
256 3.0.101-100 (SLES11 SP4)
257 3.12.74-60.64.40 (SLES12 SP1)
258 4.4.59-92.17 (SLES12 SP2)
259 vanilla linux 4.6.7 (ZFS only)
260 * Client known to build on unpatched kernels:
261 2.6.32-431.29.2.el6 (RHEL6.5)
262 2.6.32-504.30.3.el6 (RHEL6.6)
263 2.6.32-573.26.1.el6 (RHEL6.7)
264 2.6.32-642.15.1.el6 (RHEL6.8)
265 2.6.32-696.3.1.el6 (RHEL6.9)
266 3.10.0-514.21.1.el7 (RHEL7.3)
267 3.0.101-0.47.71 (SLES11 SP3)
268 3.0.101-100 (SLES11 SP4)
269 3.12.74-60.64.40 (SLES12 SP1)
270 4.4.59-92.17 (SLES12 SP2)
272 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
273 * Recommended ZFS / SPL version: 0.6.5.9
274 * Tested with ZFS / SPL version: 0.6.5.9
275 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
276 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
277 stack overflow. For more information, please refer to bugzilla 17630.
278 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
279 the re-exporting nodes to be mounted with "32bitapi" mount option
281 --------------------------------------------------------------------------------
283 12-07-2016 Intel Corporation
285 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
286 for currently supported client and server kernel versions.
287 * Server known to build on patched kernels:
288 2.6.32-431.29.2.el6 (RHEL6.5)
289 2.6.32-504.30.3.el6 (RHEL6.6)
290 2.6.32-573.26.1.el6 (RHEL6.7)
291 2.6.32-642.11.1.el6 (RHEL6.8)
292 3.10.0-514.el7 (RHEL7.3)
293 3.0.101-0.47.71 (SLES11 SP3)
294 3.0.101-80 (SLES11 SP4)
295 3.12.62-60.64.8 (SLES12 SP1)
296 vanilla linux 4.5.7 (ZFS only)
297 * Client known to build on unpatched kernels:
298 2.6.32-431.29.2.el6 (RHEL6.5)
299 2.6.32-504.30.3.el6 (RHEL6.6)
300 2.6.32-573.26.1.el6 (RHEL6.7)
301 2.6.32-642.11.1.el6 (RHEL6.8)
302 3.10.0-514.el7 (RHEL7.3)
303 3.0.101-0.47.71 (SLES11 SP3)
304 3.0.101-80 (SLES11 SP4)
305 3.12.62-60.64.8 (SLES12 SP1)
307 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
308 * Recommended ZFS / SPL version: 0.6.5.8
309 * Tested with ZFS / SPL version: 0.6.5.8
310 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
311 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
312 stack overflow. For more information, please refer to bugzilla 17630.
313 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
314 the re-exporting nodes to be mounted with "32bitapi" mount option
316 --------------------------------------------------------------------------------
318 02-29-2016 Intel Corporation
320 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
321 for currently supported client and server kernel versions.
322 * Server known to build on patched kernels:
323 2.6.32-431.29.2.el6 (RHEL6.5)
324 2.6.32-504.30.3.el6 (RHEL6.6)
325 2.6.32-573.12.1.el6 (RHEL6.7)
326 3.10.0-327.3.1.el7 (RHEL7.2)
327 3.0.101-0.47.71 (SLES11 SP3)
328 3.0.101-68 (SLES11 SP4)
329 vanilla linux 4.2.1 (ZFS only)
330 * Client known to build on unpatched kernels:
331 2.6.32-431.29.2.el6 (RHEL6.5)
332 2.6.32-504.30.3.el6 (RHEL6.6)
333 2.6.32-573.12.1.el6 (RHEL6.7)
334 3.10.0-327.3.1.el7 (RHEL7.2)
335 3.0.101-0.47.71 (SLES11 SP3)
336 3.0.101-68 (SLES11 SP4)
339 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
340 * Recommended ZFS / SPL version: 0.6.4.2
341 * Tested with ZFS / SPL version: 0.6.4.2
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 03-10-2015 Intel Corporation
352 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
353 for currently supported client and server kernel versions.
354 * Server known to build on patched kernels:
355 2.6.32-431.29.2.el6 (RHEL6.5)
356 2.6.32-504.8.1.el6 (RHEL6.6)
357 3.0.101-0.46 (SLES11 SP3)
358 * Client known to build on unpatched kernels:
359 2.6.32-431.29.2.el6 (RHEL6.5)
360 2.6.32-504.8.1.el6 (RHEL6.6)
361 3.10.0-123.20.1.el7 (RHEL7)
362 3.0.101-0.46 (SLES11 SP3)
363 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
364 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
365 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
366 stack overflow. For more information, please refer to bugzilla 17630.
367 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
368 the re-exporting nodes to be mounted with "32bitapi" mount option
370 Severity : enhancement
372 Description: control OST-index in IDIF via ROCOMPAT flag.
373 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
374 last_rcvd file. For new formatted OST device, it will be auto set;
375 for the case of upgrading from old OST device, you can enable it
376 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
377 enabled, for new created OST-object, its IDIF-in-LMA will contain
378 the OST-index; for the existing OST-object, the OSD will convert
379 old format IDIF as new format IDIF with OST-index stored in the
380 LMA EA when accessing such OST-object or via OI scrub. Once such
381 flag is enabled, it cannot be reverted back, so the system cannot
382 be downgraded to the orignal incompatible version.
384 --------------------------------------------------------------------------------
386 07-30-2014 Intel Corporation
388 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
389 for currently supported client and server kernel versions.
390 * Server known to build on patched kernels:
391 2.6.32-431.20.3.el6 (RHEL6.5)
392 3.0.101-0.31 (SLES11 SP3)
393 * Client known to build on unpatched kernels:
394 2.6.32-431.20.3.el6 (RHEL6.5)
395 3.0.101-0.31 (SLES11 SP3)
398 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
399 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
400 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
401 stack overflow. For more information, please refer to bugzilla 17630.
402 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
403 the re-exporting nodes to be mounted with "32bitapi" mount option
405 --------------------------------------------------------------------------------
407 10-31-2013 Intel Corporation
409 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
410 * Server builds for kernels:
411 2.6.32-358.18.1.el6 (RHEL6.4)
412 * Client builds for unpatched kernels:
413 2.6.32-358.18.1.el6 (RHEL6.4)
414 3.0.92-0.8 (SLES11 SP3)
416 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
417 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
418 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
419 stack overflow. For more information, please refer to bugzilla 17630.
420 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
421 the re-exporting nodes to be mounted with "32bitapi" mount option
423 --------------------------------------------------------------------------------
425 05-30-2013 Intel Corporation
427 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
428 * Server builds for kernels:
429 2.6.32-358.6.2.el6 (RHEL6.4)
430 * Client builds for unpatched kernels:
431 2.6.32-358.6.2.el6 (RHEL6.4)
432 3.0.74-0.6.6 (SLES11 SP2)
434 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
435 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
436 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
437 stack overflow. For more information, please refer to bugzilla 17630.
438 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
439 the re-exporting nodes to be mounted with "32bitapi" mount option
441 --------------------------------------------------------------------------------
443 9-30-2012 Whamcloud, Inc.
445 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
446 * Server support for kernels:
447 2.6.32-279.5.1.el6 (RHEL6)
448 * Client support for unpatched kernels:
449 2.6.32-279.5.1.el6 (RHEL6)
451 2.6.18-238.19.1.el5 (RHEL5)
452 2.6.16.60-0.69.1 (SLES10)
453 * Recommended e2fsprogs version: 1.41.90.wc4
454 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
455 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
456 stack overflow. For more information, please refer to bugzilla 17630.
457 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
458 the re-exporting nodes to be mounted with "32bitapi" mount option
460 Severity : enhancement
462 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
464 Severity : enhancement
466 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
468 Severity : enhancement
470 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
472 Severity : enhancement
474 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
476 Severity : enhancement
478 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
480 --------------------------------------------------------------------------------
482 04-30-2012 Whamcloud, Inc.
484 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
485 * Server support for kernels:
486 2.6.32-220.4.2.el6 (RHEL6)
487 * Client support for unpatched kernels:
488 2.6.32-220.4.2.el6 (RHEL6)
490 2.6.18-238.19.1.el5 (RHEL5)
491 2.6.16.60-0.69.1 (SLES10)
492 * Recommended e2fsprogs version: 1.41.90.wc4
493 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
494 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
495 stack overflow. For more information, please refer to bugzilla 17630.
496 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
497 the re-exporting nodes to be mounted with "32bitapi" mount option
499 Severity : enhancement
501 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
503 Severity : enhancement
505 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
507 Severity : enhancement
509 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
511 Severity : enhancement
513 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
515 --------------------------------------------------------------------------------
517 09-30-2011 Whamcloud, Inc.
519 * Server support for kernels:
520 2.6.18-238.19.1.el5 (RHEL5)
521 2.6.32-131.6.1.el6 (RHEL6)
522 * Client support for unpatched kernels:
523 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
524 2.6.18-238.19.1.el5 (RHEL5)
525 2.6.32-131.6.1.el6 (RHEL6)
526 2.6.29.4-167.fc11 (FC11)
527 2.6.16.60-0.69.1 (SLES10)
529 * Recommended e2fsprogs version: 1.41.90.wc3
530 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
531 Lustre filesystem with 4K stack may cause a stack overflow. For more
532 information, please refer to bugzilla 17630.
533 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
534 mounted with "32bitapi" mount option
536 Severity : enhancement
537 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
539 Severity : enhancement
540 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
542 Severity : enhancement
543 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
546 Frequency : only with SLES10
548 Description: Use OFED "KMP" provided by Novell
549 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
550 Lustre is now built against this package. That means you need to
551 install the ofed-kmp package from Novell for the patchless client.
552 Note that the ofed-kmp that Novell ships may not exactly match the
553 kernel version but should still be compatible.
555 Severity : enhancement
557 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
558 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
560 --------------------------------------------------------------------------------
562 2010-07-15 Oracle, Inc.
564 * Server support for kernels:
565 2.6.18-164.11.1.el5 (RHEL 5)
566 2.6.18-164.11.1.0.1.el5 (OEL 5)
567 * Client support for unpatched kernels:
568 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
569 2.6.18-164.11.1.el5 (RHEL 5),
570 2.6.18-164.11.1.0.1.el5 (OEL 5)
571 2.6.16.60-0.42.8 (SLES 10),
573 2.6.29.4-167.fc11 (FC11)
574 * Recommended e2fsprogs version: 1.41.10-sun2
576 Severity : enhancement
578 Description: Distribute lfs_migrate script with lustre
579 Details : The lfs_migrate mentioned in the manual is now included in the
580 lustre tree in lustre/scripts and will distributed in lustre RPMs.
582 Severity : enhancement
584 Description: Failover nid list added to import proc file.
586 Severity : enhancement
588 Description: "som_preview" mount option required to use SOM
589 Details : The clients will need to use som_preview mount option
593 Description: Lustre does not do 1MB IOs to HW RAID
594 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
596 Severity : enhancement
598 Description: lfs quota output is not very convenient for awk/sed-parsing
599 Details : Some positions in lfs quota output table could be empty or
600 non-empty which made it hard to parse it with scripts, now
601 a dash is put instead of space where there is not supposed
606 Description: Update to OFED 1.5.1
607 Details : OFED 1.5.1 will be used in environments that don't use
608 vendor supplied OFED.
612 Description: Remove lnet. prefix from lctl params display
613 Details : Also renamed lnet.memused to lnet_memused to
618 Description: Use conf_param -d to remove permanant settings.
622 Description: Add list_param -R
623 Details : lctl list_param now supports recursive mode.
626 Bugzilla : 21829, 22312
627 Description: Fix llobdstat.
628 Details : 1. Now updating the write_bytes counter
629 2. Port script fixes from 1.8.
633 Description: decrease the usage of memory on clients.
634 Details : 1. On clients, recycle dentries and inodes unused.
635 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
640 Description: llobdstat fix and enhancement
641 Details : add a counter to set a limit to how many samples will be returned
642 fix a wildcard in the path to limit to obdfilter stats only
646 Description: lfs check only allowed for root
647 Details : Code cleanups and added a regresion test of non-root use of
652 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
653 Details : Fix hr->hr_index overload.
657 Description: Fix FIEMAP functionality.
661 Description: FC11 patchless client support added
665 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
666 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
670 Description: New acc-sm script and various test-framework enhancments.
674 Description: Previously landed patch was reverted.
678 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
682 Description: Add ldiskfs maxdirsize mount option.
686 Description: Fix mkfs.lustre support for 16TB LUNs.
690 Description: Print mmp_check_interval
691 Details : print mmp_check_interval and make it possible to abort
692 mount operation in case it takes too long.
696 Description: Add cascading_rw to lustre mpi tests.
700 Description: Store SOM data in EA.
701 Details : Also several SOM related cleanups.
705 Description: Add refcount to md_open_data
709 Description: Use refcount on lov_request_set to prevent memory corruption.
713 Description: Statahead: use parent dir's i_mutex for synchronization.
717 Description: loadgen improvements.
721 Description: Enable peer health detection in ksocklnd and ko2iblnd.
725 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
729 Description: Fix race on obd_devcie:md_stats
733 Description: Fix SLES 11 client builds.
734 Details : Resolved uninitalized pointer gcc warnings.
738 Description: Allow non-root access for "lfs check".
739 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
743 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
745 Severity : enhancement
747 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
748 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
752 Description: Sanity test_27p fails.
753 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
754 Sanity tests 27 were cleaned up and improved.
758 Description: Unify req format on client/servers
759 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
760 buffers and swabbers approach..
765 Description: drop join file stuff.
767 Severity : enhancement
769 Description: Add 2.6.27 and SLES11 (patchless client) support
773 Description: Fix in ptlrpc_expire_one_request() to print the signed time
776 Severity : enhancement
778 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
781 Severity : enhancement
783 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
785 Severity : enhancement
787 Description: correctly handle big reply message.
788 Details : send LNet event if reply is bigger than buffer and adjust this buffer
793 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
795 Severity : enhancement
797 Description: Add support for OEL5.
801 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
803 Severity : enhancement
805 Description: "weak-modules" support
806 Details : Implement "weak-modules" support which enables kernel modules
807 to be used with any kernel that implements the same kABI. In
808 order to achieve this modules are now installed in
809 /lib/modules/$(uname -r)/updates/kernel on all distributions.
813 Description: client could not reconnect to OST because of an active request.
814 Details : The client now retries to reconnect to the same server, if a
815 connect request failed with EBUSY or -EAGAIN. Test case 35b
816 conf-sanity.sh has been created.
820 Description: Don't allow make backward step on assignin osc next id.
821 Details : race between allocation next id and ll_sync thread can be cause
822 of set wrong osc next id and can be kill valid ost objects.
826 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
830 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
834 Description: Error handling in osc_statfs_interpret() has been improved.
835 Details : Check in osc_statfs_interpret() for EBADR.
839 Description: Do not send statfs() requests to OSTs disabled by administrator.
840 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
842 Severity : enhancement
844 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
846 Severity : enhancement
848 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
850 Severity : enhancement
852 Description: Enhancement for lfs(1) command to use numeric uid/gid.
856 Description: Improved cleanup procedure for conf-sanity test_46a.
860 Description: Do not start bulk transfer, if client tried to reconnect.
861 Details : Do not start bulk transfer, if client tried to reconnect,
862 but failed due to active rpcs. Check for exp_abort_active_req
863 has been implemented.
867 Description: Prevent inconsistences between linux and lustre mount structures.
868 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
869 Make the sleep interruptible.
871 Severity : enhancement
873 Description: provide server to client comms path
874 Details : server to client communications path via new LDLM_SET_INFO rpc
878 Description: truncate starts GFP_FS allocation under transaction causing deadlock
879 Details : ldiskfs_truncate calls grab_cache_page which may start page
880 allocation under an open transaction. This may lead to
881 calling prune_icache with consequent lustre reentrance.
885 Description: send correctly lsm on open replay
886 Details : MDS is trust to LSM size on replay open, but client can set wrong size
891 Description: enable client interface failover
892 Details : When a child reconnects from another NID, properly update export
893 nid hash position and ldlm reverse import.
895 Severity : enhancement
897 Description: Communicate OST degraded/readonly state via statfs to MDS
898 Details : Flags in the statfs returned from OSTs indicate whether the
899 OST is in a degraded RAID state, or if the filesystem has
900 turned read-only after a filesystem error is detected.
905 Description: don't panic if EPROTO was hit when reading symlink
906 Details : correctly handling request reference in error cases.
908 Severity : enhancement
910 Description: Add LustreNetLink, a kernel-userspace communcation path.
912 Severity : enhancement
914 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
919 Description: access to llog context before init.
920 Details : move handling CATALOGS file at osc layer and forbid access to llog
924 Frequency : rare, only when client is writing to OST on local node
926 Description: Avoid deadlock for local client writes
927 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
928 memory freeing context. This allows OST threads to set the
929 PF_MEMALLOC flag on task structures in order to allocate memory
930 from reserved pools and complete IO.
931 Use GFP_HIGHUSER for OST allocations for non-local client writes,
932 so that the OST threads generate memory pressure and allow
933 inactive pages to be reclaimed.
935 Severity : enhancement
937 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
942 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
943 Details : move ldlm namespace creation in setup phase to avoid grab
944 _lprocfs_lock with cli_sem held.
948 Description: Temporarily disable grant shrink.
949 Details : Disable the feature for debugging.
952 Frequency : only when formatting test filesystems
954 Description: Unable to run several mkfs.lustre on loop devices at the same time
955 Details : mkfs.lustre returns error 256 on the concurrent loop devices
956 formatting. The solution is to proper handle the error.
958 Severity : enhancement
960 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
962 Severity : enhancement
964 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
969 Description: resolve race between obd_disconnect and class_disconnect_exports
970 Details : if obd_disconnect will be called to already disconnected export he
971 forget release one reference and osc module can't unloaded.
973 Severity : enhancement
975 Description: Allow tuning service thread via /proc
976 Details : For each service a new
977 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
978 created that can be used to set min/max thread counts, and get the
979 current number of running threads.
984 Descriptoin: don't return error if have partially created objects for file.
985 Details : lov_update_create_set uses set->set_success as index for created
986 objects, so if some requests will be failed, they will have hole
987 at end of array and qos_shrink_lsm for allocate correct lsm.
989 Severity : enhancement
991 Description: Update OFED support to 1.4.1
996 Description: Reduce small size read RPC
997 Details : Set read-ahead limit for every file and only do read-ahead when
998 available read-ahead pages are bigger than 1M to avoid small size
1001 Severity : enhancement
1003 Description: move AT tunable parameters for more consistent usage
1004 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1006 Severity : enhancement
1008 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1009 Details : allow skip disconnected ost for send statfs request and hide error
1015 Description: fix racy locking of mballoc block bitmaps causing BUG
1016 Details : The locking of the mballoc buddy bitmap and the in-memory
1017 block bitmap was using two different spin locks in some
1018 cases. This made it possible to incorrectly access the
1019 mballoc bitmap while another process was modifying it,
1020 causing a sanity assertion to fail. While no on-disk corruption
1021 was reported, there was some risk of this happening.
1024 Frequency : rare, on llog test 6
1026 Descriptoin: don't allow connect to already connected import
1027 Details : allowing connect to already connected import is hide connecting problem.
1032 Descriptoin: allow kill process which wait statahead result
1033 Details : for some reasons 'ls' can stick in waiting result from statahead,
1034 in this case need way for kill this process.
1036 Severity : enhancement
1038 Description: Add state history info file, enhance import info file
1039 Details : Track import connection state changes in a new osc/mdc proc file;
1040 add overview-type data to the osc/mdc import proc file.
1042 Severity : enhancement
1044 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1046 Severity : enhancement
1048 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1053 Descriptoin: OST grant too much space to client even there are not enough space.
1054 Details : Client will shrink its grant cache to OST if there are no write
1055 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1056 this grant cache if there are already not enough avaible space
1057 (left_space < total_clients * 32M).
1060 Frequency : start MDS on uncleanly shutdowned MDS device
1062 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1063 Details : stay in waiting mds<>ost recovery finished produce random bugs
1064 due race between two ll_sync thread for one lov target. send
1065 ACTIVATE event only if connect realy finished and import have
1069 Frequency : rare, connect and disconnect target at same time
1071 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1072 Details : don't call obd_disconnect under lov_lock. this long time
1073 operation and can block ptlrpcd which answer to connect request.
1078 Descriptoin: don't lose wakeup for imp_recovery_waitq
1079 Details : recover_import_no_retry or invalidate_import and import_close can
1080 both sleep on imp_recovery_waitq, but we was send only one wakeup
1084 Frequency : always with long access acl
1086 Descriptoin: mds can't pack reply with long acl.
1087 Details : mds don't control size of acl but they limited by reint/getattr
1090 Severity : enhancement
1092 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1094 Severity : enhancement
1096 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1099 Frequency : start MDS on uncleanly shutdowned MDS device
1101 Descriptoin: aborting recovery hang on MDS
1102 Details : don't throttle destroy RPCs for the MDT.
1105 Frequency : on remount
1107 Description: external journal device not working after the remount
1108 Details : clear dev_rdonly flag for external journal devices in
1114 Description: shutdown vs evict race
1115 Details : client_disconnect_export vs connect request race.
1116 if client will evicted at this time - we start invalidate
1117 thread without referece to import and import can be freed
1121 Frequency : rare, need acl's on inode.
1123 Description: client can't handle ost additional correctly
1124 Details : if ost was added after client connected to mds client can have
1125 hit lnet_try_match_md ... to big messages to wide striped files.
1126 in this case need teach client to handle config events about add
1127 lov target and update client max ea size at that event.
1129 Severity : enhancement
1131 Description: Changelogs
1132 Details : Changelogs are a lightweight mechanism to track filesystem
1133 metadata and namespace changes. The changelog is recorded
1134 permanently on the MDTs, and is periodically "consumed" / purged
1135 when records are no longer needed.
1137 Severity : enhancement
1139 Description: compact fld format with extents
1140 Details : Store range of seq rather than every seq in FLD. Seq
1141 controller update FLD rather than clients. In Case of CMD, mdt0
1142 has FLD, all other metadata server act as non persistent proxy
1143 for FLD queries and cache fld entries in fld cache.
1148 Description: don't skip ost target if they assigned to file
1149 Details : Drop slow OSCs if we can, but not for requested start idx.
1150 This means "if OSC is slow and it is not the requested
1151 start OST, then it can be skipped, otherwise skip it only
1152 if it is inactive/recovering/out-of-space.
1156 Description: more cleanup in mds_lov
1157 Details : not send LOV EA under replay, we can't know about they size at this
1158 time. Don't allow client connect to mds before any ost connected,
1159 for avoid problems with LOV EA size and returning EIO to client.
1161 Severity : enhancement
1163 Description: Interoperability at server side (Disk interoperability)
1165 Severity : enhancement
1167 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1169 Severity : enhancement
1171 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1173 Severity : enhancement
1175 Description: New client IO stack (CLIO).
1177 Severity : enhancement
1179 Description: Commit on sharing. Eliminate inter-client dependencies between
1180 uncommitted transactions by doing transaction commits.
1181 Thereby clients may recovery independently.
1184 Frequency : Create a symlink file with a very long name
1186 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1187 Details : If there is no extra space in the request for early cancels,
1188 ldlm_req_handles_avail() returns 0 instead of a negative value.
1190 Severity : enhancement
1192 Description: Add /proc entry for import status
1193 Details : The mdc, osc, and mgc import directories now have
1194 an import directory that contains useful import data for debugging
1195 connection problems.
1197 Severity : enhancement
1199 Description: Re-disable certain /proc logging
1200 Details : Enable and disable client's offset_stats, extents_stats and
1201 extents_stats_per_process stats logging on the fly.
1204 Frequency : Only on FC kernels 2.6.22+
1206 Description: oops in statahead
1207 Details : Do not drop reference count for the dentry from VFS when lookup,
1208 VFS will do that by itself.
1210 Severity : enhancement
1212 Description: Generic /proc file permissions
1213 Details : Set /Proc file permissions in a more generic way to enable non-
1214 root users operate on some /proc files.
1218 Description: Hitting mdc_commit_close() ASSERTION
1219 Details : Properly handle request reference release in
1220 ll_release_openhandle().
1224 Description: quota recovery deadlock during mds failover
1225 Details : This patch includes att18982, att18236, att18237 in bz14840.
1227 1. fix osts hang when mds does failover with quotaon
1228 2. prevent watchdog storm when osts threads wait for the
1233 Frequency : only patchless client
1234 Description: add workaround for race between add/remove dentry from hash
1236 Severity : enhancement
1238 Description: Allow OST glimpses to return PW locks
1242 Description: LBUG when llog conf file is full
1243 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1247 Description: Prevent import from entering FULL state when server in recovery
1251 Description: service mount cannot take device name with ":"
1252 Details : Only when device name contains ":/" will mount treat it as
1258 Description: replace ptlrpcd with the statahead thread to interpret the async
1259 statahead RPC callback
1263 Frequency : on recovery
1264 Description: I/O failures after umount during fail back
1265 Details : if client reconnected to restarted server we need join to recovery
1266 instead of find server handler is changed and process self
1267 eviction with cancel all locks.
1269 Severity : enhancement
1271 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1273 Severity : enhancement
1275 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1277 Severity : enhancement
1279 Description: Upcall on Lustre log has been dumped
1280 Details : Allow for a user mode script to be called once a Lustre log has
1281 been dumped. It passes the filename of the dumped log to the
1282 script, the location of the script can be specified via
1283 /proc/sys/lnet/debug_log_upcall.
1288 Description: avoid idr_remove called for id which is not allocated.
1289 Details : Move assigment s_dev for clustered nfs to end of initialization,
1290 to avoid problem with error handling.
1295 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1296 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1297 into CONN_USED_HASH and this prodice warning when put connection
1298 again in unused hash.
1301 Severity : enhancement
1303 Description: File striping can now be set to use an arbitrary pool of OSTs.
1305 Severity : enhancement
1307 Description: Export bytes_read/bytes_write count on OSC/OST.
1311 Description: Early reply size mismatch, MGC loses connection
1312 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1313 the connect flags are properly negotiated.
1318 Description: quotas are not honored with O_DIRECT
1319 Details : all writes with the flag O_DIRECT will use grants which leads to
1320 this problem. Now using OBD_BRW_SYNC to guard this.
1324 Description: add quota statistics
1325 Details : 1. sort out quota proc entries and proc code.
1326 2. add quota statistics
1328 Severity : enhancement
1330 Description: enable quota support for HEAD.
1334 Description: Properly propagate oinfo flags from lov to osc for statfs
1335 Details : restore missing copy oi_flags to lov requests.
1337 Severity : enhancement
1339 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1340 : l_getgroups(8), lst(8), routerstat(8)
1341 Details : included man pages for llobdstat(8), llstat(8),
1342 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1344 Severity : enhancement
1346 Description: configure's --enable-quota should check the
1347 : kernel .config for CONFIG_QUOTA
1348 Details : configure is terminated if --enable-quota is passed but
1349 : no quota support is in kernel
1353 Description: Remove portals compatibility
1354 Details : Remove portals compatibility, not interoperable with releases
1359 Description: Resolve device initialization race
1360 Details : Prevent proc handler from accessing devices added to the
1361 obd_devs array but yet be intialized.
1363 Severity : enhancement
1365 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1367 Severity : enhancement
1369 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1374 Description: Using wrong pointer in osc_brw_prep_request
1375 Details : Access to array[-1] can produce panic if kernel compiled with
1376 CONFIG_PAGE_ALLOC enabled
1380 Description: Client runs out of low memory
1381 Details : Consider only lowmem when counting initial number of llap pages
1385 Description: *optional* service tags registration
1386 Details : if the "service tags" package is installed on a Lustre node
1387 When the filesystem is mounted, a local-node service tag will
1388 be created. See http://inventory.sun.com/ for more information
1389 about the Service Tags asset management system.
1393 Description: Kernel BUG tries to release flock
1394 Details : Lustre does not destroy flock lock before last reference goes
1395 away. So always drop flock locks when client is evicted and
1396 perform unlock regardless of successfulness of speaking to MDS.
1400 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1404 Description: sanity test 65a fails if stripecount of -1 is set
1405 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1410 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1411 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1414 Severity : enhancement
1416 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1420 Description: do not process already freed flock
1421 Details : flock can possibly be freed by another thread before it reaches
1422 to ldlm_flock_completion_ast.
1426 Description: LBUG during stress test
1427 Details : Need properly lock accesses the flock deadlock detection list.
1431 Description: oops in page fault handler
1432 Details : kernel page fault handler can return two special 'pages' in error
1433 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1437 Description: timeout with invalidate import.
1438 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1439 should be handled by ptlrpcd. This produce long age waiting and
1440 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1442 Severity : enhancement
1444 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1448 Description: enable MGS and MDT services start separately
1449 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1450 a co-located MGS without starting the MGS, which is a complement
1451 to 'nosvc' mount option.
1455 Frequency : after recovery
1456 Description: precreate to many object's after del orphan.
1457 Details : del orphan st in oscc last_id == next_id and this triger growing
1458 count of precreated objects. Set flag LOW to skip increase count
1459 of precreated objects.
1463 Frequency : rare, on clear nid stats
1464 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1465 Details : when clean nid stats sometimes try destroy live entry,
1466 and this produce panic in free.
1470 Description: Stack overflow during MDS log replay
1471 ease stack pressure by using a thread dealing llog_process.
1475 Description: wait until IO finished before start new when do lock cancel.
1476 Details : VM protocol want old IO finished before start new, in this case
1477 need wait until PG_writeback is cleared until check dirty flag and
1478 call writepages in lock cancel callback.
1480 Severity : enhancement
1482 Description: using special macro for print time and cleanup in includes.
1486 Description: mds_mfd_close() ASSERTION(rc == 0)
1487 Details : In mds_mfd_close(), we need protect inode's writecount change
1488 within its orphan write semaphore to prevent possible races.
1492 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1496 Frequency : rare, on shutdown ost
1497 Description: don't hit live lock with umount ost.
1498 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1499 use shrink_dcache_sb instead.
1503 Description: don't panic with use echo client
1504 Details : echo client pass NULL as client nid pointer and this produce null
1505 pointer dereference.
1509 Description: fix build on ppc32
1510 Details : compile code with -m64 flag produce wrong object file for ppc32.
1514 Description: add message levels for liblustreapi
1518 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1519 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1520 will result in an -EAGAIN and in the exit call path we call
1521 alloc_rr() which will with increasing aggressiveness attempt to
1522 aquire precreated objects on the minimum number of required OSCs.
1526 Description: Use old size assignment to avoid deadlock
1527 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1528 the scheduling while holding a spinlock. We do not need locking
1529 for size in ll_update_inode() because size is only updated from
1530 the MDS for directories or files without objects, so there is no
1531 other place to do the update, and concurrent access to such inodes
1532 are protected by the inode lock.
1536 Description: resolve "_IOWR redefined" build error on SLES10
1540 Description: dump the memory debugging after all modules are unloaded to
1541 suppress false negative in conf_sanity test 39
1543 Severity : enhancement
1545 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1550 Description: fix SLES kernel versioning
1551 Details : the kernel version for our SLES 10 kernel did not include a "-"
1552 before the "smp" at the end. while this was not a problem in
1553 general, it did mean that software trying to use the kernel
1554 version to try to detect a vendor specific kernel would fail.
1555 this was most evident by the OFED build scripts.
1559 Description: Don't update lov_desc members until making sure they are valid
1560 Details : When updating lov_desc members via proc fs, need fix their
1561 validities before doing the real update.
1565 Description: don't put request into delay list while invalidate in flight.
1566 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1567 import in flight. this produce timeout for invalidate and sometimes
1568 can cause stale data.
1572 Frequency : on ppc only
1573 Description: not convert ost objects for directory because it's not exist.
1574 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1578 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1580 Severity : enhancement
1582 Description: Update to support 2.6.22.14 vanilla kernel.
1586 Frequency : rare, on recovery
1587 Description: read procfs can produce deadlock in some situation
1588 Details : Holding lprocfs lock which send rpc can produce block for destroy
1589 obd objects and this also block reconnect with -EALREADY. This
1590 isn't fix all lprocfs bugs - but make it rare.
1592 Severity : enhancement
1594 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1597 Frequency : frequent on X2 node
1599 Description: mdc_set_open_replay_data LBUG
1600 Details : Set replay data for requests that are eligible for replay.
1604 Description: lustre_mgs: operation 101 on unconnected MGS
1605 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1606 MGC, and late on MGC cannot successfully connect to MGS and a lot
1607 of the error messages complaining that MGS is not connected.
1611 Frequency : on network error
1612 Description: panic with double free request if network error
1613 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1614 but it's true only for synchronus enqueue, for async enqueue
1615 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1618 Severity : enhancement
1620 Description: client-side metadata stat-ahead during readdir(directory readahead)
1621 Details : perform client-side metadata stat-ahead when the client detects
1622 readdir and sequential stat of dir entries therein
1625 Frequency : on start mds
1627 Description: Implement get_info(last_id) in obdfilter.
1630 Frequency : occasional
1632 Description: Correctly check stale fid, not start epoch if ost not support SOM
1633 Details : open with flag O_CREATE need set old fid in op_fid3 because
1634 op_fid2 was overwritten with new generated fid, but mds can answer
1635 with one of these two fids and both is not stale. Setattr
1636 incorrectly started an epoch and assume will be called
1637 done_writing, but without SOM done_writing ever being called.
1640 Frequency : rare, depends on device drivers and load
1642 Description: MDS or OSS nodes crash due to stack overflow
1643 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1644 In some cases, in conjunction with device drivers that use a lot
1645 of stack the MDS (or possibly OSS) service threads could overflow
1646 the stack. One change which was identified to consume additional
1647 stack has been reworked to avoid the extra stack usage.
1650 Frequency : occasional
1652 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1653 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1654 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1655 oscc_has_objects and see OSC has no objests and call
1656 oscc_internal_create to resend create request.
1658 Severity : enhancement
1660 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1662 Severity : enhancement
1664 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1667 Frequency : very rare
1669 Description: Fix replay if there is an un-replied request and open
1670 Details : In some cases, older replay request will revert the
1671 mcd->mcd_last_xid on MDS which is used to record the client's
1672 latest sent request.
1674 Severity : enhancement
1676 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1678 Severity : enhancement
1680 Description: Add rhel5 support to HEAD.
1682 Serverity : enhancement
1684 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1689 Description: Don't allow skipping OSTs if index has been specified.
1690 Details : Don't allow skipping OSTs if index has been specified, make
1691 locking in internal create lots better.
1695 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1696 Details : access to bitfield in structure is always rounded to long
1697 and this produce problem with not atomic change any bit.
1701 Description: Lustre make rpms failed.
1702 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1703 builds Lustre rpms from tarball.
1706 Frequency : rare on shutdown ost
1708 Description: If llog cancel was not send before clean_exports phase, this can
1709 produce deadlock in llog code.
1710 Details : If llog thread has last reference to obd and call class_import_put
1711 this produce deadlock because llog_cleanup_commit_master wait when
1712 last llog_commit_thread exited, but this never success because was
1713 called from llog_commit_thread.
1717 Description: allow userland application know is lost one of stripes.
1718 Details : fill lvb_blocks with error code on ost and return it to
1719 application if error flag found.
1723 Description: NULL lov_tgts causing MDS oops
1724 Details : more safe checks for NULL lov_tgts for avoid oops.
1726 Severity : enhancement
1728 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1732 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1733 Details : If an OST is down the MDS will hang indefinitely in
1734 obd_statfs_rqset() waiting for the statfs data. While for
1735 MDS QOS usage of statfs, it should not stuck in waiting.
1737 Severity : enhancement
1739 Description: remote_acl support
1740 Details : Support ACL-based permission check for remote user.
1741 Support setfacl/getfacl for remote user with the utils
1742 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1743 the system "{s,g}etfacl" utils.
1745 Severity : enhancement
1747 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1749 Severity : enhancement
1751 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1755 Description: make mgs_setparam() handle fsname containing dash
1756 Details : fsname containing a dash does not work with lctl conf_param
1758 Severity : enhancement
1760 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1762 Severity : enhancement
1764 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1768 Frequency : rare, at shutdown
1769 Description: access already free / zero obd_namespace.
1770 Details : if client_disconnect_export was called without force flag set,
1771 and exist connect request in flight, this can produce access to
1772 NULL pointer (or already free pointer) when connect_interpret
1773 store ocd flags in obd_namespace.
1777 Frequency : only at startup
1778 Description: not alloc memory with spinlock held.
1779 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1780 if any spinlock held.
1782 Severity : enhancement
1784 Description: make lustre randomly fail allocating memory
1785 Details : Make lustre randomly failed allocating memory for testing purpose.
1787 Severity : enhancement
1789 Description: lost problems with lov objid file
1790 Details : Fixes some scability and access to not inited memory problems
1791 in work with lov objdid file.
1796 Description: lfs find does not continue on file error
1797 Details : Continue other files processing when a file/dir is absent.
1801 Description: Inconsistent usage of lustre_pack_reply()
1802 Details : Standardize the usage of lustre_pack_reply() such that it
1803 always generate a CERROR on failure.
1806 Frequency : occasional
1808 Description: MDS hang or stay in waiting lock
1809 Details : If client receive lock with CBPENDING flag ldlm need send lock
1810 cancel as separate rpc, to avoid situation when cancel request
1811 can't processed due all i/o threads stay in wait lock.
1815 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1819 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1823 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1827 Description: Update to RHEL4 latest kernel.
1829 Severity : enhancement
1831 Description: Build SLES10 patchless client fails
1832 Details : The configure was broken by run ./configure with
1833 --with-linux-obj=.... argument for patchless client. When the
1834 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1835 can't search header adequately. Use absolute path such as
1836 -I($LINUX)/include instead.
1840 Description: interrupt oig_wait produce painc on resend.
1841 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1842 private set, and this produce situation when rq_ptlrpcd_data not
1843 copyed to new allocated request and triggered LBUG on assert
1844 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1845 ptlrpcd set if request is changed and can be safety changed to use
1850 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1851 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1852 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1853 these macros and replace them with their difinitions instead.
1857 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1858 Details : Update which_patch & target file for SLES10 latest kernel.
1861 Description: Remove client patches from SLES 10 kernel.
1862 Details : This causes SLES 10 clients to behave as patchless clients
1863 even on a Lustre-patched (server) kernel.
1865 Severity : enhancement
1867 Description: self-adjustable client's lru lists
1868 Details : use adaptive algorithm for managing client cached locks lru
1869 lists according to current server load, other client's work
1870 pattern, memory activities, etc. Both, server and client
1871 side namespaces provide number of proc tunables for controlling
1874 Severity : enhancement
1876 Description: light-weight GSS support
1877 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1878 and reduce performance overhead.
1880 Severity : enhancement
1882 Description: Linux keyring support
1883 Details : Support using service of Linux keyring for Lustre GSS internal
1884 context refresh/cache mechanism.
1888 Description: Fix errors in lfs documentation
1889 Details : Fixes man pages
1893 Description: don't use GFP_* in generic Lustre code.
1894 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1898 Description: obdclass is limited by single OBD_ALLOC(idarray)
1899 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1903 Description: warnings with build patchless client with vanila 2.6.19 and up
1904 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1905 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1909 Frequency : Only for SLES
1910 Description: sanity_quota fail test_1
1911 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1912 file, which makes TSTID[2] inunique.
1916 Frequency : Only for SLES9
1917 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1918 Details : Update target/ChangeLog/which_patch .
1922 Frequency : Only for test_10 in sanity-quota.sh
1923 Description: a bug in quota test code
1924 Details : checking a test flag in a wrong place
1928 Frequency : Only for test_13 in sanity-quota.sh
1929 Description: a bug in quota test script
1930 Details : shouldn't check overfull allocated quota
1934 Description: jbd statistics
1935 Details : Port older jbd statistics patch for sles10
1937 Severity : enhancement
1939 Description: Add printing inode info into message about error in writepage.
1944 Description: reply_lock_interpret crash due to race with it and lock cancel.
1945 Details : Do not replay locks that are being cancelled. Do not reference
1946 locks by their address during replay, just by their handle.
1950 Frequency : When flocks are used.
1951 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1952 Details : Flock locks might destroy just granted lock if it could be merged
1953 with another existing flock, this is done in completion handler,
1954 so teach ldlm_cli_enquque_fini that this is a valid case for
1960 Description: Oops in read and write path when failing to allocate lock.
1961 Details : Check if lock allocation failed and return error back.
1965 Description: lstripe command fails for valid OST index
1966 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1967 instead of lov->desc.ld_active_tgt_count.
1971 Description: sanity.sh failed test 103
1972 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1977 Description: df doesn't work properly if diskfs blocksize != 4K
1978 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1982 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1983 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1985 Severity : enhancement
1987 Description: Adaptive timeouts
1988 Details : RPC timeouts adapt to changing server load and network
1989 conditions to reduce resend attempts and improve recovery time.
1993 Description: llapi_file_create() does not allow some changes
1994 Details : add llapi_file_open() that allows specifying the mode and
1995 open flags, and also returns an open file handle.
1999 Description: merge and cleanup kernel patches.
2000 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2004 Description: Add journal checksum support.(Kernel part)
2005 Details : The journal checksum feature adds two new flags i.e
2006 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2007 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2008 indicates that the commit block contains the checksum for
2009 the blocks described by the descriptor blocks. Now commit
2010 record can be sent to disk without waiting for descriptor
2011 blocks to be written to disk. This behavior is controlled
2012 using JBD2_FEATURE_ASYNC_COMMIT flag.
2016 Description: OSS needs mutliple precreate threads
2017 Details : Add ability to start more than one create thread per OSS.
2021 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2022 Details : execute lfs setstripe on client
2026 Description: mds_obd_create error creating tmp object
2027 Details : When the user sets quota on root, llog will be affected and can't
2028 create files and write files.
2031 Frequency : Always on ia64 patchless client, and possibly others.
2033 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2034 Details : This allows the patchless client to be loaded on architectures
2035 without this export.
2039 Description: RedHat Update kernel for RHEL5
2040 Details : Add the kernel config file for RHEL5.
2044 Description: RedHat Update kernel for RHEL5
2045 Details : Modify the kernel config file more closer RHEL5.
2049 Description: Build failure against Centos5 (RHEL5)
2050 Details : Define PAGE_SIZE when it isn't present.
2054 Description: test_7 on sanity-quota.sh can't be used on separate machines
2055 Details : add facet to handle it
2059 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2060 Details : with NFS, the anon dentry's parent was set to itself in
2061 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2062 corresponding dentry other than use rec->ur_name.
2064 Severity : enhancement
2066 Description: lfs setstripe enhancement
2067 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2071 Description: enable data checksumming by default
2072 Details : enable checksum by default, allow --disable-checksum
2073 configure option and "-o nochecksum" mount option. Checksums
2074 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2078 Description: lustre support for RHEL5
2079 Details : Add support for RHEL5.
2083 Description: Client eviction due to failover config
2084 Details : after a connection loss, the lustre client should attempt to
2085 reconnect to the last active server first before trying the
2086 other potential connections.
2090 Description: when mds and osts use different quota unit(32bit and 64bit),
2091 quota will be released repeatly.
2092 Details : void sending multiple quota reqs to mds, which will keep the
2093 status between the reqs.
2097 Description: osts not allocated evenly to files
2098 Details : change the condition to increase offset_idx
2101 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2103 Description: Data corruption for OSTs that are formatted larger than 2TB
2105 Details : When generating the bio request for lustre file writes the
2106 sector number would overflow a temporary variable before being
2107 used for the IO. The data reads correctly from Lustre (which
2108 will overflow in a similar manner) but other file data or
2109 filesystem metadata may be corrupted in some cases.
2113 Description: Tune the kernel for good SCSI performance.
2114 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2115 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2120 Description: rewrite ext2-derived code in obdclass/uuid.c
2121 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2122 from scratch preserving functionality.
2126 Description: "lfs find -obd UUID" prints directories
2127 Details : "lfs find -obd UUID" will return all directory names instead
2128 of just file names. It is incorrect because the directories
2129 do not reside on the OSTs.
2133 Description: use i_size_read and i_size_write in 2.6 port
2134 Details : replace inode->i_size access with i_size_read/write()
2137 Frequency : when removing large files
2139 Description: scheduling issue during removal of large Lustre files
2140 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2141 It causes scheduling issues when removing large files (17TB in the
2145 Frequency : only with liblustre clients on XT3
2147 Description: evictions taking too long
2148 Details : allow llrd to evict clients directly on OSTs
2151 Frequency : only on ppc
2153 Description: /proc/fs/lustre/devices broken on ppc
2154 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2155 We should make sure the type of 'index' is loff_t and then cast
2156 explicitly as needed below. Do not assign an explicitly cast
2160 Frequency : only for rhel5
2162 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2163 Details : Modify the target file & which_kernel.
2165 Severity : enhancement
2167 Description: omit set fsid for export NFS
2168 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2169 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2173 Frequency : Always, for kernels after 2.6.16
2174 Description: Fix warning idr_remove called for id=.. which is not allocated.
2175 Details : Last kernels save old s_dev before kill super and not allow
2176 to restore from callback - restore it before call kill_anon_super.
2180 Description: buffer overruns could theoretically occur
2181 Details : llapi_semantic_traverse() modifies the "path" argument by
2182 appending values to the end of the origin string, and a
2183 overrun may occur. Adding buffer overrun check in liblustreapi.
2187 Description: Fix error on 'ls .' at the top of the Lustre mount.
2188 Details : Don't revalidate dentry if it is a root dentry.
2192 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2193 Details : Modify vm-tunables-rhel4.patch.
2197 Description: Kernel config for 2.6.18-vanilla.
2198 Details : Modify targets/2.6-vanilla.target.in.
2199 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2200 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2201 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2202 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2206 Description: improve handling recoverable errors
2207 Details : if request processig with error which can be recoverable on server
2208 request should be resend, otherwise page released from cache and
2213 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2214 Details : Modify target file & which_patch.
2215 A flaw was found in the IA32 system call emulation provided
2216 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2217 value could be stored in the %RAX register, which could trigger an
2218 out-of-bounds system call table access. An untrusted local user
2219 could exploit this flaw to run code in the kernel
2220 (ie a root privilege escalation). (CVE-2007-4573).
2224 Description: O_DIRECT bypasses client statistics.
2225 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2226 read_ahead_stats were updated. copied the stats section from
2227 osc_send_oap_rpc() into async_internal().
2231 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2235 Description: change order of libsysio includes
2236 Details : '#include sysio.h' should always come before '#include xtio.h'
2240 Description: handle "serious error: objid * already exists" more gracefully
2241 Details : If LAST_ID value on disk is smaller than the objects existing in
2242 the O/0/d* directories, it indicates disk corruption and causes an
2243 LBUG(). If the object is 0-length, then we should use the existing
2244 object. This will help to avoid a full fsck in most cases.
2246 Severity : enhancement
2248 Description: adapt the lustre_config script to support the upgrade case
2249 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2250 server targets to 1.6 in parallel.
2254 Description: To avoid grant space > avaible space when the disk is almost
2255 full. Without this patch you might see the error "grant XXXX >
2256 available" or some LBUG about grant, when the disk is almost
2258 Details : In filter_check_grant, for non_grant cache write, we should
2259 check the left space by if (*left > ungranted + bytes), instead
2260 of (*left > ungranted), because only we are sure the left space
2261 is enough for another "bytes", then the ungrant space should be
2262 increase. In client, we should update cl_avail_grant only there
2263 is OBD_MD_FLGRANT in the reply.
2267 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2268 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2269 described in CVE-2007-4573. This problem could allow untrusted
2270 local users to gain root access.
2273 Frequency : when using O_DIRECT and quotas
2275 Description: Incorrect file ownership on O_DIRECT output files
2276 Details : block usage reported by 'lfs quota' does not take into account
2277 files that have been written with O_DIRECT.
2282 Description: touch file failed when fs is not full
2283 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2284 otherwise we can get ENOSP while fs is not full.
2288 Description: parallel lock callbacks
2289 Details : Instead of sending blocking and completion callbacks as separated
2290 requests, adding them to a set and sending in parallel.
2293 Frequency : only for Cray XT3
2294 Bugzilla : 12829/13455
2295 Description: Changing primary group doesn't change the group lustre assigns to
2297 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2298 overridden with the primary group provided by the group upcall,
2299 whereas the supplied fsgid can be trusted if it is in the list of
2300 supplementary groups returned by the group upcall.
2302 Severity : enhancement
2304 Description: Allow masking D_WARNING, D_ERROR messages from console
2305 Details : Console messages can now be disabled via lnet.printk.
2309 Description: User code with malformed file open parameter crashes client node
2310 Details : Before packing join_file req, all the related reference should be
2311 checked carefully in case some malformed flags cause fake
2312 join_file req on client.
2316 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2317 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2318 mode is equal to the granted mode. However, it should also check
2319 that LDLM_FL_CP_REQD is not set.
2323 Description: Heavy nfs access might result in deadlocks
2324 Details : After ELC code landed, it is now improper to enqueue any mds
2325 locks under och_sem, because enqueue might want to decide to
2326 cancel open locks for same inode we are holding och_sem for.
2330 Description: Client eviction while running blogbench
2331 Details : A lot of unlink operations with concurrent I/O can lead to a
2332 deadlock causing evictions. To address the problem, the number of
2333 oustanding OST_DESTROY requests is now throttled to
2334 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2335 callbacks are processed in priority.
2339 Description: enable ACLs on MDS by default
2340 Details : ACLs must be enabled on MDS by default.
2343 Frequency : PPC/PPC64 only
2345 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2346 Details : fix duplicated definitions between asm-ppc64/types.h and
2347 lustre_types.h on PPC.
2350 Frequency : PPC/PPC64 only
2352 Description: asm-ppc/segment.h does not exist
2353 Details : fix compile issue on PPC.
2357 Description: better handle error messages in extents code
2360 Frequency : RHEL4 only
2362 Description: mkfs is very slow on IA64/RHEL4
2363 Details : A performance regression has been discovered in the MPT Fusion
2364 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2365 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2366 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2368 Severity : enhancement
2370 Description: SNMP support enhancement
2371 Details : Adding total number of sampled request for an MDS node in snmp
2374 Severity : enhancement
2376 Description: Optimize ldlm waiting list processing for PR extent locks
2377 Details : When processing waiting list for read extent lock and meeting read
2378 lock that is same or wider to it that is not contended, skip
2379 processing rest of the list and immediatelly return current
2380 status of conflictness, since we are guaranteed there are no
2381 conflicting locks in the rest of the list.
2385 Description: Time out and refuse to reconnect
2386 Details : When the failover node is the primary node, it is possible
2387 to have two identical connections in imp_conn_list. We must
2388 compare not conn's pointers but NIDs, otherwise we can defeat
2389 connection throttling.
2393 Description: port llog fixes from b1_6 into HEAD
2394 Details : Port llog reference couting and some llog cleanups from b1_6
2395 (bug 10800) into HEAD, for protect from panic and access to already
2396 free llog structures.
2400 Description: Detect stride IO mode in read-ahead
2401 Details : When a client does stride read, read-ahead should detect that and
2402 read-ahead pages according to the detected stride pattern.
2406 Description: data checksumming impacts single node performance
2407 Details : add support for several checksum algorithm. Currently, only CRC32
2408 and Adler-32 are supported. The checksum type can be changed on
2409 the fly via /proc/fs/lustre/osc/*/checksum_type.
2413 Description: use adler32 for page checksums
2414 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2419 Description: build for x2 fails
2420 Details : fix compile issue on Cray systems.
2424 Description: Properly match for duplicate locks
2425 Details : Due to different lock order from skiplists code, we need to
2426 traverse entire list for now
2429 Frequency : only on PPC/SLES10
2431 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2432 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2433 be defined. Add a hack in mkfs_lustre.c to work around this compile
2438 Description: LASSERT on MDS when client holding flock lock dies
2439 Details : ldlm pool logic depends on number of granted locks equal to
2440 number of released locks which is not true for flock locks, so
2441 just exclude such locks from consideration.
2445 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2446 Details : Use fsfilt_sync() for both the whole filesystem sync and
2447 individual file sync to eliminate dangerous inode locking
2448 with I_LOCK that can lead to a deadlock.
2452 Description: performance in 1.6.3
2453 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2454 sure that 1MB requests are not fragmented by the block layer.
2456 Severity : enhancement
2458 Description: organize the server-side client stats on per-nid basis
2459 Details : Change the structure of stats under obdfilter and mds to
2467 The "uuid"s file would list the uuids of _active_ exports.
2468 And the clear entry is to clear all stats and stale nids.
2470 Severity : enhancement
2472 Description: eliminate client locks in face of contention
2473 Details : file contention detection and lockless i/o implementation
2474 for contended files.
2478 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2481 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2483 Description: MDS may lose file striping (and hence file data) in some cases
2484 Details : If there are additional extended attributes stored on the MDS,
2485 in particular ACLs, SELinux, or user attributes (if user_xattr
2486 is specified for the client mount options) then there is a risk
2487 of attribute loss. Additionally, the Lustre file striping
2488 needs to be larger than default (e.g. striped over all OSTs),
2489 and an additional attribute must be stored initially in the
2490 inode and then increase in size enough to be moved to the
2491 external attribute block (e.g. ACL growing in size) for file
2496 Description: skiplist implementation simplification
2497 Details : skiplists are used to group compatible locks on granted list
2498 that was implemented as tracking first and last lock of each
2499 lock group the patch changes that to using doubly linked lists
2503 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2504 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2505 because LASSERT on both the data supplied by a client, and the
2506 data on disk is dangerous and incorrect.
2508 Severity : enhancement
2510 Description: Slow truncate/writes to huge files at high offsets.
2511 Details : Directly associate cached pages to lock that protect those pages,
2512 this allows us to quickly find what pages to write and remove
2513 once lock callback is received.
2517 Description: more ldlm soft lockups
2518 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2519 starve other threads from the resource lock for a long time in
2520 case of long waiting queue, so change the debug level from
2521 D_OTHER to the less frequently used D_INFO.
2523 Severity : enhancement
2525 Description: add -gid, -group, -uid, -user options to lfs find
2529 Description: Hung threads in invalidate_inode_pages2_range
2530 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2531 once one is completed. As a result, some RPCs are stuck in the
2532 queue and are never sent.
2536 Description: filter threads hungs on waiting journal commit
2537 Details : Cleanup filter group llog code, then only filter group llog will
2538 be only created in the MDS/OST syncing process.
2542 Description: Procfs and llog threads access destoryed import sometimes.
2543 Details : Sync the import destoryed process with procfs and llog threads by
2544 the import refcount and semaphore.
2546 Severity : enhancement
2548 Description: openlock cache of b1_6 port to HEAD
2553 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2554 Details : If insertion of an extent fails, then discard the inode
2555 preallocation and free data blocks else it can lead to duplicate
2560 Description: don't always update ctime in ext3_xattr_set_handle()
2561 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2562 In some cases the ctime should not be updated, for example for
2563 2.0->1.8 compatibility it is necessary to delete an xattr and it
2564 should not update the ctime.
2568 Bugzilla : 15713/16362
2569 Description: Assertion in iopen_connect_dentry in 1.6.3
2570 Details : looking up an inode via iopen with the wrong generation number can
2571 populate the dcache with a disconneced dentry while the inode
2572 number is in the process of being reallocated. This causes an
2573 assertion failure in iopen since the inode's dentry list contains
2574 both a connected and disconnected dentry.
2578 Description: assertion failure in ldlm_handle2lock()
2579 Details : fix a race between class_handle_unhash() and class_handle2object()
2580 introduced in lustre 1.6.5 by bug 13622.
2585 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2586 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2591 Description: oops in sd_iostats_seq_show()
2592 Details : unloading/reloading the scsi low level driver triggers a kernel
2593 bug when trying to access the sd iostat file.
2598 Description: Kernel panics during QLogic driver reload
2599 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2600 patch, causing memory corruption.
2605 Description: journal_dev option does not work in b1_6
2606 Details : pass mount option during pre-mount.
2608 Severity : enhancement
2610 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2611 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2612 extent information of a file. It can be used to map logical blocks
2613 in a file to physical blocks in the block device.
2617 Description: LDLM soft lockups - improvement
2618 Details : It is be possible to send the lock handle along with each read
2619 or write request because the client is already doing a lock match
2620 itself so there isn't any reason the OST should have to re-do that
2624 Frequency : only on Cray X2
2626 Description: X2 build failures
2627 Details : fix build failures on Cray X2.
2631 Description: xid & resent requests
2632 Details : Initialize RPC XID from clock at startup (randomly if clock is
2635 Severity : enhancement
2637 Description: Add lustre_start utility to start or stop multiple Lustre servers
2642 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2643 Details : In case of memory pressure, list_del() can be called twice on
2644 req->rq_history_list, causing a kernel oops.
2648 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2649 Details : fix stack overflow in the distributed lock manager by defering
2650 export eviction after a failed AST to the elt thread instead of
2651 handling it in the dlm interpret routine.
2655 Description: Convert some comments to new format.
2656 Details : Update documenting comments to match doxygen conventions.
2660 Description: Grammar fixes.
2661 Details : A couple of trivial spelling fixes.
2665 Description: OSD_COUNTERS-mandatory
2666 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2667 to simplify the code.
2671 Description: simplify cmm_device freeing logic.
2672 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2676 Description: Add lockdep support to dt_object_operations locking interface.
2677 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2678 parameter indicating lock ordering. Update mdd code to use new
2683 Description: Introduce failloc constants for lockless IO tests.
2684 Details : Add two new failloc constants to test lockless IO. Only one of
2685 them in implemented---another is checked in yet to be landed
2690 Description: Add lockdep support for inode mutex.
2691 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2692 in the situations where Lustre has to lock more than one inode
2697 Description: Add optional invariants checking support.
2698 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2699 configure switch. This macro is to be used for consistency and
2700 sanity checks that are too expensive to be left in `production'
2705 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2706 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2711 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2712 Details : Use spin_lock_nested() in (the only) situation where more than
2713 one ldlm_lock is locked simultaneously. Also, fix possible
2714 dead-lock in ldlm_lock_change_resource() by enforcing particular
2719 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2720 Details : Instead of specifying each ldlm_lock call-back through separate
2721 parameter, wrap them into struct ldlm_callback_suite.
2725 Description: Kill join_lru obd method and its callers.
2726 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2727 in memory---a requirement implemented through ->o_join_lru() obd
2728 method in HEAD. Remove this method and its users.
2732 Description: Add asynchronous ldlm ENQUEUE completion handler.
2733 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2734 case that stock ldlm_completion_ast() cannot handle as it waits
2735 until lock is granted. Introduce new ldlm_completion_ast_async()
2736 for this. Also comment ldlm_completion_ast().
2740 Description: ldlm_error <-> errno conversion.
2741 Details : Add functions to map (rather arbitrary) between LDLM error codes
2742 and standard errno values. CLIO needs this to prevent LDLM specific
2743 constants from escaping ldlm and osc.
2747 Description: Kill unused ldlm_handle2lock_ns() function.
2748 Details : Kill unused ldlm_handle2lock_ns() function.
2752 Description: Add lu_ref support to ldlm_lock
2753 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2754 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2755 and ldlm_resource. LDLM interface has to be changed, because code that
2756 releases a reference on a lock, has to "know" what reference this is.
2757 In the most frequent case
2759 lock = ldlm_handle2lock(handle);
2761 LDLM_LOCK_PUT(lock);
2763 no changes are required. When any other reference (received _not_ from
2764 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2765 instead of LDLM_LOCK_PUT().
2767 Arguably, changes are pervasive, and interface requires some discipline
2768 for proper use. On the other hand, it was very instrumental in finding
2769 a few leaked lock references.
2773 Description: Add ldlm_lock_addref_try().
2774 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2775 attempts to addref a lock that might be being canceled
2780 Description: Add ldlm_weigh_callback().
2781 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2782 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2783 hard-coded `number of pages' logic.
2787 Description: Add lockdep annotations to llog code.
2788 Details : Use appropriately tagged _nested() locking calls in the places
2789 where llog takes more than one ->lgh_lock lock.
2793 Description: Add loi_kms_set().
2794 Details : Wrap kms updates into a helper function.
2798 Description: Constify instances of struct lsm_operations.
2799 Details : Constify instances of struct lsm_operations.
2803 Description: lu_conf support.
2804 Details : On a server, a file system object is uniquely identified
2805 by a fid, which is sufficient to locate and load all object
2806 state (inode). On a client, on the other hand, more data are
2807 necessary instantiate an object. Change lu_object_find() and
2808 friends to take additional `lu_conf' argument describing object.
2809 Typically this includes layout information.
2813 Description: lu_context fixes.
2814 Details : Introduce new lu_context functions that are needed on the client
2815 side, where some system threads (ptlrpcd) are shared by multiple
2816 modules, and so cannot be stopped during module shutdown.
2820 Description: Add start and stop methods to lu_device_type_operations.
2821 Details : Introduce two new methods in lu_device_type_operations, that are
2822 invoked when first instance of a given type is created and last one
2823 is destroyed respectively. This is need by CLIO.
2827 Description: Add lu_ref support to struct lu_device.
2828 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2829 track leaked references.
2833 Description: Introduce lu_kmem_descr.
2834 Details : lu_kmem_descr and its companion interface allow to create
2835 and destroy a number of kmem caches at once.
2839 Description: Fix lu_object finalization race.
2840 Details : Fix a race between lu_object_find() finding an object and its
2841 concurrent finalization. This race is (most likely) not possible
2842 on the server, but might happen on the client.
2846 Description: Introduce lu_ref interface.
2847 Details : lu_ref is a debugging module allowing to track references to
2848 a given object. It is quite cpu expensive, and has to be
2849 explicitly enabled with --enable-lu_ref. See usage description
2854 Description: Factor lu_site procfs stats into a separate function.
2855 Details : Separate lu_site stats printing code into a separate function
2856 that can be reused on a client.
2860 Description: Constify instances of struct {lu,dt,md}_device_operations.
2861 Details : Constify instances of struct {lu,dt,md}_device_operations.
2865 Description: Introduce struct md_site and move meta-data specific parts of
2866 struct lu_site here.
2867 Details : Move md-specific fields out of struct lu_site into special struct
2868 md_site, so that lu_site can be used on a client.
2872 Description: Kill mdd_lov_destroy().
2873 Details : Remove unused mdd code.
2877 Description: Add st_block checking to multistat.c.
2878 Details : Add st_block checking to multistat.c.
2882 Description: Add lu_ref support to struct obd_device.
2883 Details : Add lu_ref tracking to obd_device.
2887 Description: Kill obd_set_fail_loc().
2888 Details : Remove unused code.
2892 Description: Add special type for ptlrpc_request interpret functions.
2893 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2894 there. Actual usage will be in CLIO.
2898 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2899 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2900 doesn't work with lockdep.
2904 Description: Add rwv.c test program.
2905 Details : New testing program exercising readv(2) and writev(2) (Qian).
2909 Description: Add sendfile.c test program.
2910 Details : New testing program exercising sendfile(2) (Jay).
2914 Description: Ratelimit a message that can be very frequent.
2915 Details : Ratelimit a memory allocation failure message that can
2920 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2921 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2925 Description: Liblustre build fixes.
2926 Details : Liblustre build fixes.
2930 Description: libcfs: add cfs_{need,cond}_resched() interface.
2931 Details : libcfs: add cfs_{need,cond}_resched() definition and
2932 implementations for Linux, NT, and liblustre.
2934 Severity : enhancement
2936 Description: More exported tunables for mballoc
2937 Details : Add support for tunable preallocation window and new tunables for
2938 large/small requests
2942 Description: Detect corruption of block bitmap and checking for preallocations
2943 Details : Checks validity of on-disk block bitmap. Also it does better
2944 checking of number of applied preallocations. When corruption is
2945 found, it turns filesystem readonly to prevent further corruptions.
2949 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2950 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2951 window to 0. If it does hit cache miss, and read-pattern is still
2952 stride-io mode, does not reset the stride window, but also does not
2953 increase the stride window length in this case.
2957 Frequency : only for big-endian servers
2958 Description: Check if system is big-endian while mounting fs with extents feature
2959 Details : Mounting a filesystem with extents feature will fail on big-endian
2960 systems since ext3-based ldiskfs is not supported on big-endian
2961 systems. This can be over-riden with "bigendian_extents" mount option.
2963 Severity : enhancement
2965 Description: The root squash functionality
2966 Details : A security feature, which is to prevent users from being able
2967 to mount lustre on their desktop, run as root, and delete
2968 all of the files in the filesystem. The goal is accomplished by
2969 remapping user id (UID) and group id (GID) of the root user to
2970 a UID and GID specified by the system administartor via Lustre
2971 configuration management server (MGS). The functionality also
2972 allows to specify sets of clients for which the remapping does
2977 Description: Excessive recovery window
2978 Details : With AT enabled, the recovery window can be excessively long (6000+
2979 seconds). To address this problem, we no longer use
2980 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2981 timeout no longer depends on the service time, it is set to
2982 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2983 time via pb_service_time.
2987 Description: Watchdog triggered on MDS failover
2988 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2989 the OSTs know that the MDS "UUID" can be reused for the same export
2990 from a different NID, so we do not need to wait for the export to be
2994 Frequency : rare, only if using MMP with Linux RAID
2996 Description: MMP doesn't work with Linux RAID
2997 Details : While using HA for Lustre servers with Linux RAID, it is possible
2998 that MMP will not detect multiple mounts. To make this work we
2999 need to unplug the device queue in RAID when the MMP block is being
3000 written. Also while reading the MMP block, we should read it from
3001 disk and not the cached one.
3003 Severity : enhancement
3005 Description: open file using fid
3006 Details : A file can be opened using just its fid, like
3007 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3010 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3013 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3014 Details : add proper locking for creation of jbd_4k slab cache
3018 Description: MMP check in ext3_remount() fails without displaying any error
3019 Details : When multiple mount protection fails during remount, proper error
3022 Severity : enhancement
3024 Description: Allow stripe size to be up to 4G-64k
3025 Details : Fix math logic to allow large stripe sizes.
3029 Description: add check for >8TB ldiskfs filesystems
3030 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3031 Don't allow >8TB ldiskfs filesystems to be mounted without
3032 force_over_8tb mount option
3037 Descriptoin: OSS complains "too many blocks" and crashes
3038 Details : In filter_commitrw_write():
3039 1. lquota_chkquota();
3040 2. obdo_from_inode();
3041 3. lquota_pending_commit();
3042 "2." has changed uid and gid in this case, then caused
3043 lquota_chkquota() and lquota_pending_commit() to use
3044 different uid and gid. That is the root of the bug.
3048 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3049 Details : Detect and print out the 'stripe count = -1' case properly.
3053 Description: Out or order replies might be lost on replay
3054 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3055 request with tid smaller than one currently being added, add it
3056 to the start, not end of the list.
3058 --------------------------------------------------------------------------------
3060 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3062 * Support for kernels:
3063 2.6.5-7.283 (SLES 9),
3064 2.6.9-55.EL (RHEL 4),
3065 2.6.16.46-0.14 (SLES 10),
3066 2.6.18.8 vanilla (kernel.org)
3067 * Client support for unpatched kernels:
3068 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3069 2.6.16 - 2.6.22 vanilla (kernel.org)
3070 * Due to recently discovered recovery problems, we do not recommend
3071 using patchless RHEL 4 clients with this or any earlier release.
3072 * Recommended e2fsprogs version: 1.39.cfs8
3073 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3074 * Starting with this release, the ldiskfs backing filesystem required
3075 by Lustre is now in its own package, lustre-ldiskfs. This package
3076 should be installed. It is versioned separately from Lustre and
3077 may be released separately in future.
3081 Description: block reactivating mgc import until all deactivates complete
3082 Details : Fix race when failing back MDT/MGS to itself (testing)
3085 Frequency : at statup only
3087 Description: mds_lov_synchronize race leads to various problems
3088 Details : simultaneous MDT->OST connections at startup can cause the
3089 sync to abort, leaving the OSC in a bad state.
3091 Severity : enhancement
3093 Description: add optional extra BUILD_VERSION info
3094 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3095 to override the lustre version.
3098 Frequency : 2.6.18 servers only
3100 Description: ll_kern_mount() doesn't release the module reference
3101 Details : The ldiskfs module reference count never drops down to 0
3102 because ll_kern_mount() doesn't release the module reference.
3107 Description: server LBUG when using old ost_num_threads parameter
3108 Details : Accept the old ost_num_threads parameter but warn that it
3109 is deprecated, and fix an off-by-one error that caused an LBUG.
3114 Description: Transient SCSI error results in persistent IO issue
3115 Details : iobuf->dr_error is not reinitialized to 0 between two
3119 Frequency : sometimes when underlying device returns I/O errors
3121 Description: OSTs not going read-only during write failures
3122 Details : OSTs are not remounted read-only when the journal commit threads
3123 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3124 instead of the ext3 wrappers.
3128 Description: poor connect scaling with increasing client count
3129 Details : Don't run filter_grant_sanity_check for more than 100 exports
3130 to improve scaling for large numbers of clients.
3133 Frequency : SLES10 only
3135 Description: sanity-quota.sh quotacheck failed: rc = -22
3136 Details : Quotas cannot be enabled on SLES10.
3139 Frequency : liblustre clients only
3141 Description: getdirentries does not give error when run on compute nodes
3142 Details : getdirentries does not fail when the size specified as an argument
3143 is too small to contain at least one entry
3145 Severity : enhancement
3147 Description: Add LNET router traceability for debug purposes
3148 Details : If a checksum failure occurs with a router as part of the
3149 IO path, the NID of the last router that forwarded the bulk data
3150 is printed so it can be identified.
3155 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3156 Details : Due to a race condition, liblustre clients were occasionally
3157 evicted incorrectly.
3159 Severity : enhancement
3161 Description: lfs setstripe use optional parameters instead of postional
3164 Severity : enhancement
3166 Description: Nanosecond timestamp support for ldiskfs
3167 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3168 resolution timestamps. There is not yet support for this at
3169 the Lustre filesystem level.
3172 Frequency : during server recovery
3174 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3175 Details : request with rq_no_resend flag not awake l_wait_event if they get
3179 Frequency : nfs export on patchless client
3181 Description: connectathon hang when test nfs export over patchless client
3182 Details : Disconnected dentry cannot be found with lookup, so we do not need
3183 to unhash it or make it invalid
3186 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3187 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3188 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3193 Description: open req refcounting wrong on reconnect
3194 Details : If reconnect happened between getting open reply from server and
3195 call to mdc_set_replay_data in ll_file_open, we will schedule
3196 replay for unreferenced request that we are about to free.
3197 Subsequent close will crash in variety of ways.
3198 Check that request is still eligible for replay in
3199 mdc_set_replay_data().
3204 Description: disable writes to filesystem when reading health_check file
3205 Details : the default for reading the health_check proc file has changed
3206 to NOT do a journal transaction and write to disk, because this
3207 can cause reads of the /proc file to hang and block HA state
3208 checking on a healthy but otherwise heavily loaded system. It
3209 is possible to return to the previous behaviour during configure
3210 with --enable-health-write.
3212 Severity : enhancement
3214 Description: 64-bit inode version
3215 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3216 to the inode. This will be required for version-based recovery.
3221 Description: MDS fails to start if a duplicate client export is detected
3222 Details : in some rare cases it was possible for a client to connect to
3223 an MDS multiple times. Upon recovery the MDS would detect this
3224 and fail during startup. Handle this more gracefully.
3226 Severity : enhancement
3228 Description: Add -o localflock option to simulate old noflock
3230 Details : This will achieve local-only flock/fcntl locks
3236 Description: log_commit_thread vs filter_destroy race leads to crash
3237 Details : Take import reference before releasing llog record semaphore
3242 Description: Wrong request locking in request set processing
3243 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3244 imp_delayed_list, in this place should be used imp_lock.
3247 Frequency : when reconnection
3249 Description: Grant Leak when osc reconnect to OST
3250 Details : When osc reconnect ost, OST(filter) should check whether it
3251 should grant more space to client by comparing fed_grant and
3252 cl_avail_grant, and return the granted space to client instead
3253 of "new granted" space, because client will call osc_init_grant
3254 to update the client grant space info.
3257 Frequency : when client reconnect to OST
3259 Description: Grant Leak when osc do resend and replay bulk write
3260 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3261 bulk write request, because the grant info while be sync between
3262 OSC and OST when reconnect, and we should ignore the grant info
3263 these of resend/replay write req.
3268 Description: Grant space more than available left space sometimes.
3269 Details : When then OST is about to be full, if two bulk writing from
3270 different clients came to OST. Accord the avaliable space of the
3271 OST, the first req should be permitted, and the second one
3272 should be denied by ENOSPC. But if the seconde arrived before
3273 the first one is committed. The OST might wrongly permit second
3274 writing, which will cause grant space > available space.
3277 Frequency : when client is evicted
3279 Description: Grant might be wrongly erased when osc is evicted by OST
3280 Details : when the import is evicted by server, it will fork another
3281 thread ptlrpc_invalidate_import_thread to invalidate the
3282 import, where the grant will be set to 0. While the original
3283 thread will update the grant it got when connecting. So if
3284 the former happened latter, the grant will be wrongly errased
3285 because of this race.
3290 Description: Checking Stale with correct fid
3291 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3292 to check whether it is stale, because sometimes, we want the
3293 enqueue happened anyway, and op_data.fid2 will not be initialized.
3295 Severity : enhancement
3297 Description: update patchless client
3298 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3301 Frequency : only with 2.4 kernel
3303 Description: random memory corruption
3304 Details : size of struct ll_inode_info is to big for union inode.u and this
3305 can be cause of random memory corruption.
3310 Description: Memory leak in recovery
3311 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3312 It should also check obd_fail before fsfilt_start, otherwise if
3313 fsfilt_start return -EROFS,(failover mds during mds recovery).
3314 then the req will return with repmsg->transno = 0 and rc = EROFS.
3315 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3316 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3317 be freed no matter whether fsfilt_commit success or not.
3320 Frequency : only with huge count clients
3322 Description: Prevents from taking the superblock lock in llap_from_page for
3324 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3325 for llap_from_page call in ll_removepage prevents from taking the
3326 superblock lock for a soon died page.
3331 Description: Not check open intent error before release open handle
3332 Details : in some rare cases, the open intent error is not checked before
3333 release open handle, which may cause
3334 ASSERTION(open_req->rq_transno != 0), because it tries to release
3335 the failed open handle.
3340 Description: Set cat log bitmap only after create log success.
3341 Details : in some rare cases, the cat log bitmap is set too early. and it
3342 should be set only after create log success.
3346 Description: Accessing a block bevice can re-enable I/O when Lustre is
3347 tearing down a device.
3348 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3353 Description: service threads may hog cpus when there are a lot of requests
3354 Details : Insert cond_resched to give other threads a chance to use some CPU
3359 Description: the cat log was not initialized in recovery
3360 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3361 unlink log on mds will not be initialized until mds post
3362 recovery. And also in mds post recovery, the unlink log will
3363 initialization will be done asynchronausly, so there will be race
3364 between add unlink log and unlink log initialization.
3368 Description: brw_stats were being printed incorrectly
3369 Details : brw_stats were being printed as log2 but all of them were not
3370 recorded as log2. Also remove some code duplication arising from
3371 filter_tally_{read,write}.
3375 Frequency : rare, only in recovery.
3376 Description: ASSERTION(req->rq_type != LI_POISON) failed
3377 Details : imp_lock should be held while iterating over imp_sending_list for
3378 prevent destroy request after get timeout in ptlrpc_queue_wait.
3382 Description: replay-single.sh test 52 fails
3383 Details : A lock's skiplist need to be cleanup when it being unlinked
3384 from its resource list.
3388 Description: Short directio read returns full requested size rather than
3390 Details : Direct I/O operations should return actual amount of bytes
3391 transferred rather than requested size.
3393 Severity : enhancement
3395 Description: metadata RPC reduction (e.g. for rm performance)
3396 Details : decrease the amount of synchronous RPC between clients and servers
3397 by canceling conflicing lock before the operation on the client
3398 and packing thier handles into the main operation RPC to server.
3400 Severity : enhancement
3402 Description: Async OSC create to avoid the blocking unnecessarily.
3403 Details : If an OST has no remaining object, system will block on the
3404 creation when it needs to create a new object on this OST. Now,
3405 ways use pre-created objects when available, instead of blocking on
3406 an empty osc while others are not empty. If we must block, we
3407 block for the shortest possible period of time.
3411 Description: improve handling recoverable errors
3412 Details : if request processig with error which can be recoverable on server
3413 request should be resend, otherwise page released from cache and
3416 Severity : enhancement
3418 Description: refine locking for avoid write wrong info into lov_objid file
3419 Details : fix possible races with add new target and write/update data in
3422 --------------------------------------------------------------------------------
3424 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3429 Frequency : on some architectures
3431 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3432 Details : Uninitialized flags sometimes cause configuration commands to
3436 Frequency : patchless clients only
3438 Description: missing __iget() symbol export
3439 Details : The __iget() symbol export is missing. To avoid the need for
3440 this on patchless clients the deathrow inode reaper is turned
3441 off, and we depend on the VM to clean up old inodes. This
3442 dependency was during via the fix for bug 12181.
3444 --------------------------------------------------------------------------------
3446 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3448 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3449 INTEROPERATE with older versions automatically. In many cases a
3450 special upgrade step is needed. Please read the
3451 user documentation before upgrading any part of a live system.
3452 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3453 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3454 * WARNING: Lustre configuration and startup changes are required with
3455 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3459 Severity : enhancement
3461 Description: MountConf
3462 Details : Lustre configuration is now managed via mkfs and mount
3463 commands instead of lmc and lconf. New obd types (MGS, MGC)
3464 are added for dynamic configuration management. See
3465 https://mail.clusterfs.com/wikis/lustre/MountConf for
3468 Severity : enhancement
3470 Description: dynamic OST addition
3471 Details : OSTs can now be added to a live filesystem
3473 Severity : enhancement
3475 Description: startup order invariance
3476 Details : MDTs and OSTs can be started in any order. Clients only
3477 require the MDT to complete startup.
3479 Severity : enhancement
3481 Description: parallel, asynchronous orphan cleanup
3482 Details : orphan cleanup is now performed in separate threads for each
3483 OST, allowing parallel non-blocking operation.
3485 Severity : enhancement
3487 Description: optimized stripe assignment
3488 Details : stripe assignments are now made based on ost space available,
3489 ost previous usage, and OSS previous usage, in order to try
3490 to optimize storage space and networking resources.
3492 Severity : enhancement
3494 Description: Permanently set tunables
3495 Details : All writable /proc/fs/lustre tunables can now be permanently
3496 set on a per-server basis, at mkfs time or on a live system.
3498 Severity : enhancement
3500 Description: Lustre message v2
3501 Details : Add lustre message format v2.
3503 Severity : enhancement
3505 Description: client OST exclusion list
3506 Details : Clients can be started with a list of OSTs that should be
3507 declared "inactive" for known non-responsive OSTs.
3511 Description: SPEC SFS validation failure on NFS v2 over lustre.
3512 Details : Changes the blocksize for regular files to be 2x RPC size,
3513 and not depend on stripe size.
3515 Severity : enhancement
3517 Description: Multiple MD RPCs in flight.
3518 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3519 To avoid overly-overloading MDS, introduce a limit on number of
3520 MDS RPCs in flight for a single client and add /proc controls
3521 to adjust this limit.
3523 Severity : enhancement
3525 Description: client read/write statistics
3526 Details : Add client read/write call usage stats for performance
3527 analysis of user processes.
3528 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3529 file access. extents_stats shows chunk size distribution.
3530 extents_stats_per_process show chunk size distribution per
3533 Severity : enhancement
3535 Description: mds statistics
3536 Details : Add detailed mds operations statistics in
3537 /proc/fs/lustre/mds/*/stats.
3541 Description: Failure of copying files with lustre special EAs.
3542 Details : Client side always return success for setxattr call for lustre
3543 special xattr (currently only "trusted.lov").
3548 Description: Refcount LNET uuids
3549 Details : The global LNET uuid list grew linearly with every startup;
3550 refcount repeated list entries instead of always adding to
3553 Severity : enhancement
3555 Description: Dynamic service threads
3556 Details : Within a small range, start extra service threads
3557 automatically when the request queue builds up.
3560 Frequency : mixed-endian client/server environments
3562 Description: mixed-endian crashes
3563 Details : The new msg_v2 system had some failures in mixed-endian
3566 Severity : enhancement
3568 Description: Easy OST removal
3569 Details : OSTs can be permanently deactivated with e.g. 'lctl
3570 conf_param lustre-OST0001.osc.active=0'
3572 Severity : enhancement
3574 Description: MGS proc entries
3575 Details : Added basic proc entries for the MGS showing what filesystems
3578 Severity : enhancement
3580 Description: provide MGS failover
3581 Details : Added config lock reacquisition after MGS server failover.
3583 Severity : enhancement
3585 Description: add Linux 2.4 support
3586 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3590 Description: a large application tries to do I/O to the same resource and dies
3591 in the middle of it.
3592 Details : Check the req->rq_arrival time after the call to
3593 ost_brw_lock_get(), but before we do anything about
3594 processing it & sending the BULK transfer request. This
3595 should help move old stale pending locks off the queue as
3596 quickly as obd_timeout.
3599 Frequency : when an incorrect nid is specified during startup
3601 Description: ptlrpc connect to non-existant node causes kernel crash
3602 Details : LNET can't be re-entered from an event callback, which
3603 happened when we expire a message after the export has been
3604 cleaned up. Instead, hand the zombie cleanup off to another
3607 Severity : enhancement
3609 Description: plain/inodebits lock performance improvement
3610 Details : Grouping plain/inodebits in granted list by their request modes
3611 and bits policy, thus improving the performance of search through
3615 Frequency : only if OST filesystem is corrupted
3617 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3618 Details : for a short time RPCs with bulk IO are in the replay list,
3619 but replay of bulk IOs is unimplemented. If the OST filesystem
3620 is corrupted due to disk cache incoherency and then replay is
3621 started it is possible to trip an assertion. Avoid putting
3622 committed RPCs into the replay list at all to avoid this issue.
3625 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3627 Description: System hang on startup
3628 Details : This bug allowed the liblustre (e.g. catamount) client to
3629 return to the app before handling all startup RPCs. This
3630 could leave the node unresponsive to lustre network traffic
3631 and manifested as a server ptllnd timeout.
3633 Severity : enhancement
3635 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3636 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3637 boolean to control whether to print peer debug info when a
3638 client's RPC times out.
3641 Frequency : only for kernels with patches from Lustre below 1.4.3
3643 Description: Remove old rdonly API
3644 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3647 Frequency : only for devices with external journals
3649 Description: Set external device read-only also
3650 Details : During a commanded failover stop, we set the disk device
3651 read-only while the server shuts down. We now also set any
3652 external journal device read-only at the same time.
3655 Frequency : when upgrading from 1.4 while trying to change parameters
3657 Description: The wrong (new) MDC name was used when setting parameters for
3658 upgraded MDT's. Also allows changing of OSC (and MDC)
3659 parameters if --writeconf is specified at tunefs upgrade time.
3662 Frequency : when setting specific ost indicies
3664 Description: QOS code breaks on skipped indicies
3665 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3666 created with --index need not be sequential.
3668 Severity : enhancement
3670 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3671 Details : The uninit_groups feature works in conjunction with the kernel
3672 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3673 up the pass1 processing of e2fsck. This is a read-only feature
3674 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3675 mount filesystems that have had this feature enabled.
3677 Severity : enhancement
3679 Description: Improve multi-block allocation algorithm to avoid fragmentation
3680 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3681 allocation locality and avoid filesystem fragmentation.
3683 ------------------------------------------------------------------------------
3685 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3687 * Support for kernels:
3688 2.4.21-47.0.1.EL (RHEL 3)
3689 2.6.5-7.283 (SLES 9)
3690 2.6.9-42.0.10.EL (RHEL 4)
3691 2.6.12.6 vanilla (kernel.org)
3692 2.6.16.27-0.9 (SLES 10)
3693 * Recommended e2fsprogs version: 1.39.cfs5
3695 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3699 Frequency : occasional, depends on client load and configuration
3700 Bugzilla : 12181, 12203
3701 Description: data loss for recently-modified files
3703 Details : In some cases it is possible that recently written or created
3704 files may not be written to disk in a timely manner (this should
3705 normally be within 30s unless client IO load is very high).
3706 The problem appears as zero-length files or files that are a
3707 multiple of 1MB in size after a client crash or client eviction
3708 that are missing data at the end of the file.
3710 This problem is more likely to be hit on clients where files are
3711 repeatedly created and unlinked in the same directory, clients
3712 have a large amount of RAM, have many CPUs, the filesystem has
3713 many OSTs, the clients are rebooted frequently, and/or the files
3714 are not accessed by other nodes after being written.
3716 The presence of the problem can be detected by looking at
3717 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3718 smaller than the second (nr_unused) then dirty files will not
3719 be flushed automatically to disk. "sync; sleep 10" should be
3720 run several times on the node before unmounting it to update
3721 Lustre (this is also safe to run on nodes without this problem).
3723 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3724 that can cause this same problem, so customers using that kernel
3725 also need to update the kernel in addition to Lustre. In order
3726 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3728 It is normal that files written just before a client crash (less
3729 than 30s) may not yet have been flushed to disk, even for local
3733 Frequency : frequent on thin XT3 nodes
3735 Description: UUID collision on thin XT3 Linux nodes
3736 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3737 randomly, since we relied on an insufficiently-seeded PRNG.
3742 Description: OSS hangs after "All ost request buffers busy"
3743 Details : A deadlock between quota and journal operations caused OSS
3744 hangs after printing "All ost request buffers busy."
3747 Frequency : always on liblustre builds
3749 Description: Cleanup compiler warnings on liblustre
3752 Frequency : always on liblustre builds on XT3
3754 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3758 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3759 > 32000 subdirectories
3760 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3761 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3762 subdirectory count crosses 32000. This will aid e2fsck to
3763 correctly handle more than 32000 subdirectories.
3766 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3768 Description: System hang on startup
3769 Details : This bug allowed the liblustre (e.g. catamount) client to
3770 return to the app before handling all startup RPCs. This
3771 could leave the node unresponsive to lustre network traffic
3772 and manifested as a server ptllnd timeout.
3774 Severity : enhancement
3776 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3777 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3778 boolean to control whether to print peer debug info when a
3779 client's RPC times out.
3784 Description: make O_SYNC working on 2.6 kernels
3785 Details : 2.6 kernels use different method for mark pages for write,
3786 so need add a code to lustre for O_SYNC work.
3791 Description: Failure to close file and release space on NFS
3792 Details : Put inode details into lock acquired in ll_intent_file_open.
3793 Use mdc_intent_lock in ll_intent_open to properly
3794 detect all kind of errors unhandled by mdc_enqueue.
3799 Description: proc file read during shutdown sometimes raced obd removal,
3801 Details : Add lock to prevent obd access after proc file removal.
3804 Frequency : Only for files larger than 4GB on 32-bit clients.
3806 Description: improperly doing page alignment of locks
3807 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3808 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3813 Description: RPCs being resent when they shouldn't be
3814 Details : Some RPCs that should not be resent are being resent. This
3815 can cause inconsistencies in the RPC state machine. Do not
3816 resend such requests.
3819 Frequency : rare, only with NFS export
3821 Description: Crash on NFS re-export node
3822 Details : under very unusual load conditions an assertion is hit in
3823 ll_intent_file_open()
3826 Frequency : only if OST filesystem is corrupted
3828 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3829 Details : for a short time RPCs with bulk IO are in the replay list,
3830 but replay of bulk IOs is unimplemented. If the OST filesystem
3831 is corrupted due to disk cache incoherency and then replay is
3832 started it is possible to trip an assertion. Avoid putting
3833 committed RPCs into the replay list at all to avoid this issue.
3838 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3839 Details : Large single O_DIRECT read and write calls can fail to allocate
3840 a sufficiently large buffer to process the request. In case of
3841 allocation failure the allocation is retried with a smaller
3842 buffer and broken into smaller requests.
3844 Severity : enhancement
3846 Description: Add -o localflock option to simulate old noflock behaviour.
3847 Details : This will achieve local-only flock/fcntl locks coherentness.
3852 Description: versioning check is incomplete
3853 Details : Checking the version difference of client vs. server, report
3854 error if the gap is too big.
3859 Description: add support PG_writeback bit
3860 Details : add support for PG_writeback bit for Lustre, for more carefull
3861 work with page cache in 2.6 kernel. This also fix some deadlocks
3862 and remove hack for work O_SYNC with 2.6 kernel.
3864 Severity : enhancement
3866 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3867 Details : The uninit_groups feature works in conjunction with the kernel
3868 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3869 up the pass1 processing of e2fsck. This is a read-only feature
3870 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3871 mount filesystems that have had this feature enabled.
3873 Severity : enhancement
3875 Description: Improve multi-block allocation algorithm to avoid fragmentation
3876 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3877 allocation locality and avoid filesystem fragmentation.
3879 ------------------------------------------------------------------------------
3881 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3883 * Support for kernels:
3884 2.6.9-42.0.3.EL (RHEL 4)
3885 2.6.5-7.276 (SLES 9)
3886 2.4.21-47.0.1.EL (RHEL 3)
3887 2.6.12.6 vanilla (kernel.org)
3888 2.6.16.21-0.8 (SLES10)
3889 * Recommended e2fsprogs version: 1.39.cfs2-0
3891 * The backwards-compatible /proc/sys/portals symlink has been removed
3892 in this release. Before upgrading, please ensure that you change
3893 any configuration scripts or /etc/sysctl.conf files that access
3894 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3895 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3896 in advance of the upgrade on any system running Lustre 1.4.6 or
3897 newer, since /proc/sys/lnet was added in that version.
3898 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3902 Frequency : only when quota is used
3904 Description: avoid scanning export list for quota master
3905 Details : Change the algorithms to avoid scanning export list in order
3906 to improve the efficiency.
3909 Frequency : MDS failover only, very rarely
3911 Description: "went back in time" messages on mds failover
3912 Details : The greatest transno may be lost when the current operation
3913 finishes with an error (transno==0) and the client's last_rcvd
3914 record is over-written. Save the greatest transno in the
3915 mds_last_transno for this case.
3918 Frequency : always for specific kernels and striping counts
3920 Description: client may get "Matching packet too big" without ACL support
3921 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3922 when trying to access files in certain configurations. The
3923 clients should in fact be denied when mounting because they do
3924 not understand ACLs.
3927 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3929 Description: many clients connecting with IO in progress causes connect timeouts
3930 Details : Avoid synchronous journal commits to avoid delays caused by many
3931 clients connecting/disconnecting when bulk IO is in progress.
3932 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3933 OST_IO_PORTAL to avoid delays behind potentially many pending
3937 Frequency : occasionally with multiple writers to a single file
3939 Description: shared writes to file may result in wrong size reported by stat()
3940 Details : Allow growing of kms when extent lock is cancelled
3943 Frequency : always with random mmap IO to multi-striped file
3945 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3946 Details : Only if the hole is at the end of OST object so that kms is too
3947 small. Fix is to increase kms accordingly in ll_nopage.
3950 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3952 Description: writes to a missing object would leak memory on the OST
3953 Details : If there is an inconsistency between the MDS and OST filesystems,
3954 such that the MDS references an object that doesn't exist, writes
3955 to that object will leak memory due to incorrect cleanup in the
3956 error handling path, eventually running out of memory on the OST.
3961 Description: Creating too long symlink causes lustre errors
3962 Details : Check symlink and name lengths before sending requests to MDS.
3965 Frequency : only if flock is enabled (not on by default)
3967 Description: posix locks not released on fd closure on 2.6.9+
3968 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3969 kernels, this caused such locks not to be released on fd close and
3970 then assertions on fs unmount about still used locks.
3973 Frequency : MDS failover only, very rarely
3975 Description: clients may get ASSERTION(granted_lock != NULL)
3976 Details : When request was taking a long time, and a client was resending
3977 a getattr by name lock request. The were multiple lock requests
3978 with the same client lock handle and
3979 mds_getattr_name->fixup_handle_for_resent_request found one of the
3980 lock handles but later failed with ASSERTION(granted_lock != NULL).
3985 Description: handle->h_buffer_credits > 0, assertion failure
3986 Details : h_buffer_credits is zero after truncate, causing assertion
3987 failure. This patch extends the transaction or creates a new
3991 Frequency : NFS re-export or patchless client
3992 Bugzilla : 11179, 10796
3993 Description: Crash on NFS re-export node (__d_move)
3994 Details : We do not want to hash the dentry if we don't have a lock.
3995 But if this dentry is later used in d_move, we'd hit uninitialised
3996 list head d_hash, so we just do this to init d_hash field but
3997 leave dentry unhashed.
4000 Frequency : NFS re-export or patchless client
4002 Description: NFS exports has problem with symbolic link
4003 Details : lustre client didn't properly install dentry when re-exported
4004 to NFS or running patchless client.
4007 Frequency : NFS re-export or patchless client
4009 Description: Various nfs/patchless fixes.
4010 Details : fixes reuse disconected alias for lookup process - this fixes
4011 warning "find_exported_dentry: npd != pd",
4012 fix permission error with open files at nfs.
4013 fix apply umask when do revalidate.
4016 Frequency : occasional
4018 Description: Crash on NFS re-export node
4019 Details : calling clear_page() on the wrong pointer triggered oops in
4020 generic_mapping_read().
4023 Frequency : rarely, using O_DIRECT IO
4025 Description: unaligned directio crashes client with LASSERT
4026 Details : check for unaligned buffers before trying any requests.
4029 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4031 Description: stale data returned from RAID cache
4032 Details : If only a small amount of IO is done to the RAID device before
4033 reading it again it is possible to get stale data from the RAID
4034 cache instead of reading it from disk.
4037 Frequency : always for sles10 kernel
4039 Description: sles10 support
4040 Details : ll_follow_link: compile fixes and using of nd_set_link
4041 under newer kernels.
4044 Frequency : depends on arch, kernel and compiler version, always on sles10
4047 Description: recursive or deep enough symlinks cause stack overflow
4048 Details : getting rid of large stack-allocated variable in
4052 Frequency : depends on hardware
4054 Description: lustre write performance loss in the SLES10 kernel
4055 Details : the performance loss is caused by using of write barriers in the
4056 ext3 code. The SLES10 kernel turns barrier support on by
4057 default. The fix is to undo that change for ldiskfs.
4059 ------------------------------------------------------------------------------
4061 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4063 * Support for kernels:
4064 2.6.9-42.0.3EL (RHEL 4)
4065 2.6.5-7.276 (SLES 9)
4066 2.4.21-47.0.1.EL (RHEL 3)
4067 2.6.12.6 vanilla (kernel.org)
4071 Frequency : quota enabled and large files being deleted
4073 Description: releasing more than 4GB of quota at once hangs OST
4074 Details : If a user deletes more than 4GB of files on a single OST it
4075 will cause the OST to spin in an infinite loop. Release
4076 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4081 Description: statfs data retrieved from /proc may be stale or zero
4082 Details : When reading per-device statfs data from /proc, in the
4083 {kbytes,files}_{total,free,avail} files, it may appear
4084 as zero or be out of date.
4087 Frequency : systems with MD RAID1 external journal devices
4089 Description: lconf's call to blkid is confused by RAID1 journal devices
4090 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4091 of returning all block devices that match the journal UUID.
4094 Frequency : always, for aggregate stripe size over 4GB
4096 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4097 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4098 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4101 Frequency : always if "lfs find" used on a local file/directory
4103 Description: "lfs find" segfaults if used on a local file/directory
4104 Details : The case where a directory component was not specified wasn't
4105 handled correctly. Handle this properly.
4108 Frequency : always on ppc64
4110 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4111 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4112 perform endianness conversion.
4115 Frequency : rarely (truncate to non-zero file size after write under load)
4116 Bugzilla : 10730, 10687
4117 Description: Files padded with zeros to next 4K multiple
4118 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4119 it is possible that files that are truncated to a non-zero size
4120 immediately after being written are filled with zero bytes beyond
4121 the truncated size. No file data is lost.
4123 Severity : enhancement
4125 Description: Allow recovery/failover for liblustre clients.
4126 Details : liblustre clients were unaware of failover configurations until
4129 Severity : enhancement
4131 Description: user file locks should fail when not mounting with flock option
4132 Details : Set up an error-returning stub in ll_file_operations.lock field
4133 to prevent incorrect behaviour when client is mounted without
4134 flock option. Also, set up properly f_op->flock field for
4138 Frequency : always on ia64
4140 Description: "lfs df" loops on printing out MDS statfs information
4141 Details : The obd_ioctl_data was not initialized and in some systems
4142 this caused a failure during the ioctl that did not return
4143 an error. Initialize the struct and return an error on failure.
4146 Frequency : SLES 9 only
4148 Description: Error of copying files with lustre special EAs as root
4149 Details : Client side always return success for setxattr call for lustre
4150 special xattr (currently only "trusted.lov").
4153 Frequency : rarely on clusters with both ia64+i386 clients
4155 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4156 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4157 clients are doing concurrent writes to the same file it is
4158 possible that the ia64 clients may LASSERT because the OST
4159 extent locks are not PAGE_SIZE aligned. Ensure that grown
4160 locks are always aligned on the request boundary.
4163 Frequency : specific use, occasional
4165 Description: Overwriting in use executable truncates on-disk binary image
4166 Details : If one node attempts to overwrite an executable in use by
4167 another node, we now correctly return ETXTBSY instead of
4168 truncating the file.
4170 Severity : enhancement
4172 Description: Async OSC create to avoid the blocking unnecessarily.
4173 Details : If an OST has no remaining object, system will block on the creating
4174 when need to create a new object on this OST. Now, ways use
4175 pre-created objects when available, instead of blocking on an
4176 empty osc while others are not empty. If we must block, we block
4177 for the shortest possible period of time.
4182 Description: chmod on Lustre root is propagated to other clients
4183 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4184 invalid by the follow_mount time.
4187 Frequency : liblustre clients only
4189 Description: Race in 'instant cancel' lock handling could lead to such locks
4190 never to be granted in case of SMP MDS
4191 Details : Do not destroy not yet granted but cbpending locks in
4195 Frequency : replay/resend of open
4197 Description: non null lock assetion failure in mds_intent_policy
4198 Details : Trying to replay/resend lockless open requests resulted in
4199 mds_open() returning 0 with no lock. Now it sets a flag if
4200 a lock is going to be returned.
4202 Severity : enhancement
4204 Description: Checksum enhancements
4205 Details : New checksum enhancements allow for resending RPCs that failed
4208 Severity : enhancement
4210 Description: Tunables on number of dirty pages in cacche
4211 Details : Allow to set limit on number of dirty pages cached.
4216 Description: client crash on unmount - lock still has references
4217 Details : In some error handling cases it was possible to leak a lock
4218 reference on a client while accessing a file. This was not
4219 harmful to the client during operation, but would cause the
4220 client to crash when the filesystem is unmounted.
4223 Frequency : specific case, rare
4225 Description: ETXTBSY on mds though file not in use by client
4226 Details : ETXTBSY is no longer incorrectly returned when attempting to
4227 chmod or chown a directory that the user previously tried to
4228 execute or a currently-executing binary.
4231 Frequency : extremely rare except on liblustre-based clients
4233 Description: Lustre space not freed when files are deleted
4234 Details : Clean up open-unlinked files after client eviction. Previously
4235 the unlink was skipped and the files remained as orphans.
4240 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4241 Details : In some cases it was possible to send a blocking callback to
4242 a client doing a glimpse, even though that client didn't get
4243 a lock granted. When the glimpse lock is cancelled on the OST
4244 the freed lock is left on the waiting list and corrupted the list.
4247 Frequency : all core dumps
4249 Description: Broke core dumps to lustre
4250 Details : Negative dentry may be unhashed if parent does not have UPDATE
4251 lock, but some callers, e.g. do_coredump, expect dentry to be
4252 hashed after successful create, hash it in ll_create_it.
4254 ------------------------------------------------------------------------------
4256 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4258 * Support for kernels:
4259 2.6.9-42.0.2.EL (RHEL 4)
4260 2.6.5-7.276 (SLES 9)
4261 2.4.21-40.EL (RHEL 3)
4262 2.6.12.6 vanilla (kernel.org)
4266 Frequency : always on RHEL 3
4268 Description: Number of open files grows over time
4269 Details : The number of open files grows over time, whether or not
4270 Lustre is started. This was due to a filp leak introduced
4271 by one of our kernel patches.
4273 ------------------------------------------------------------------------------
4275 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4277 * Support for kernels:
4278 2.6.9-42.EL (RHEL 4)
4279 2.6.5-7.276 (SLES 9)
4280 2.4.21-40.EL (RHEL 3)
4281 2.6.12.6 vanilla (kernel.org)
4286 Bugzilla : 5719, 9635, 9792, 9684
4287 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4288 Details : If a server is under heavy load and cannot reply to new
4289 connection requests before the client resends the (re)connect,
4290 the connection handling code can behave badly if two service
4291 threads are concurrently handing separate (re)connections from
4292 the same client. Add better locking to the connection handling
4293 code, and ensure that only a single connection will be processed
4294 for a given client UUID, even if the lock is dropped.
4296 Severity : enhancement
4298 Description: add TCP zero-copy support to kernel
4299 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4300 sends if the hardware supports scatter-gather and checksumming.
4301 This allows socklnd to do client-write and server-read more
4302 efficiently and reduce CPU utilization from skbuf copying.
4305 Frequency : only if NFS exporting from client
4307 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4308 Details : If setting attributes on a file created under NFS that had
4309 never been opened it would be possible to oops the client
4310 if the file had no objects.
4313 Frequency : always for liblustre
4315 Description: liblustre client does MDS+OSTs setattr RPC for each write
4316 Details : When doing a write from a liblustre client, the client
4317 incorrectly issued an RPC to the MDS and each OST the file was
4318 striped over in order to update the timestamps. When writing
4319 with small chunks and many clients this could overwhelm the MDS
4320 with RPCs. In all cases it would slow down the write because
4321 these RPCs are unnecessary.
4323 Severity : enhancement
4325 Description: allow number of MDS service threads to be changed at module load
4326 Details : It is now possible to change the number of MDS service threads
4327 running. Adding "options mds mds_num_threads={N}" to the MDS's
4328 /etc/modprobe.conf will set the number of threads for the next
4329 time Lustre is restarted (assuming the "mds" module is also
4330 reloaded at that time). The default number of threads will
4331 stay the same, 32 for most systems.
4336 Description: OST crash if filesystem is unformatted or corrupt
4337 Details : If an OST is started on a device that has never been formatted
4338 or if the filesystem is corrupt and cannot even mount then the
4339 error handling cleanup routines would dereference a NULL pointer.
4344 Description: NULL pointer deref in llap_from_page.
4345 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4346 incorrect) mapping if the page was truncated/reclaimed while it was
4347 searched for. Check for this condition and skip such pages when
4348 doing readahead. Introduce extra check to llap_from_page() to
4349 verify page->mapping->host is non-NULL (so page is not anonymous).
4352 Frequency : Sometimes when using sys_sendfile
4354 Description: "page not covered by a lock" warnings from ll_readpage
4355 Details : sendfile called ll_readpage without right page locks present.
4356 Now we introduced ll_file_sendfile that does necessary locking
4357 around call to generic_file_sendfile() much like we do in
4361 Frequency : with certain MDS communication failures at client mount time
4363 Description: NULL pointer deref after failed client mount
4364 Details : a client connection request may delayed by the network layer
4365 and not be sent until after the PTLRPC layer has timed out the
4366 request. If the client fails the mount immediately it will try
4367 to clean up before the network times out the request. Add a
4368 reference from the request import to the obd device and delay
4369 the cleanup until the network drops the request.
4372 Frequency : occasionally during client (re)connect
4374 Description: assertion failure during client (re)connect
4375 Details : processing a client connection request may be delayed by the
4376 client or server longer than the client connect timeout. This
4377 causes the client to resend the connection request. If the
4378 original connection request is replied in this interval, the
4379 client may trip an assertion failure in ptlrpc_connect_interpret()
4380 which thought it would be the only running connect process.
4383 Frequency : only with obd_echo servers and clients that are rebooted
4385 Description: kernel BUG accessing uninitialized data structure
4386 Details : When running an obd_echo server it did not start the ping_evictor
4387 thread, and when a client was evicted an uninitialized data
4388 structure was accessed. Start the ping_evictor in the RPC
4389 service startup instead of the OBD startup.
4391 Severity : enhancement
4392 Bugzilla : 10193 (patchless)
4393 Description: Remove dependency on various unexported kernel interfaces.
4394 Details : No longer need reparent_to_init, exit_mm, exit_files,
4395 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4398 Frequency : rare (only users of deprecated and unsupported LDAP config)
4400 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4401 Details : LDAP apparently contains 'lustreName' attributes instead of
4402 'name'. A simple remapping of the name is sufficient.
4405 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4407 Description: waiting_locks_callback trips kernel BUG if client is evicted
4408 Details : Running with the dump_on_timeout debug flag turned on makes
4409 it possible that the waiting_locks_callback() can try to dump
4410 the Lustre kernel debug logs from an interrupt handler. Defer
4411 this log dumping to the expired_lock_main() thread.
4413 Severity : enhancement
4415 Description: Support NFS exporting on 2.6 kernels.
4416 Details : Implement non-rawops metadata methods for NFS server to use without
4417 changing NFS server code.
4420 Frequency : very rare (synthetic metadata workload only)
4422 Description: two racing renames might cause an MDS thread to deadlock
4423 Details : Running the "racer" program may cause one MDS thread to rename
4424 a file from being the source of a rename to being the target of
4425 a rename at exactly the same time that another thread is doing
4426 so, and the second thread has already enqueued these locks after
4427 doing a lookup of the target and is trying to relock them in
4428 order. Ensure that we don't try to re-lock the same resource.
4431 Frequency : only very large systems with liblustre clients
4433 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4434 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4435 all OSTs in parallel. This allows the eviction of stale liblustre
4436 clients to proceed much faster than if they were done in series,
4437 and also offers similar improvements for other set_info RPCs.
4442 Description: excessive CPU usage during initial read phase on client
4443 Details : During the initial read phase on a client, it would agressively
4444 retry readahead on the file, consuming too much CPU and impacting
4445 performance (since 1.4.5.8). Improve the readahead algorithm
4446 to avoid this, and also improve some other common cases (read
4447 of small files in particular, where "small" is files smaller than
4448 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4453 Description: MDS crash when receiving packet with unknown intent.
4454 Details : Do not LBUG in unknown intent case, just return -EFAULT
4456 Severity : enhancement
4457 Bugzilla : 9293, 9385
4458 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4459 Details : Do not serialize getattr (non-intent version) and statfs.
4462 Frequency : occasional, when OST network is overloaded/intermittent
4464 Description: client evicted by OST after bulk IO timeout
4465 Details : If a client sends a bulk IO request (read or write) the OST
4466 may evict the client if it is unresposive to its data GET/PUT
4467 request. This is incorrect if the network is overloaded (takes
4468 too long to transfer the RPC data) or dropped the OST GET/PUT
4469 request. There is no need to evict the client at all, since
4470 the pinger and/or lock callbacks will handle this, and the
4471 client can restart the bulk request.
4474 Frequency : Always when mmapping file with no objects
4476 Description: client crashes when mmapping file with no objects
4477 Details : Check that we actually have objects in a file before doing any
4478 operations on objects in ll_vm_open, ll_vm_close and
4484 Description: Request leak when working with deleted CWD
4485 Details : Introduce advanced request refcount tracking for requests
4486 referenced from lustre intent.
4488 Severity : Enhancement
4490 Description: Cache open file handles on client.
4491 Details : MDS now will return special lock along with openhandle, if
4492 requested and client is allowed to hold openhandle, even if unused,
4493 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4494 closing files for every read/write openration.
4496 Severity : Enhancement
4498 Description: Cache open negative dentries on client when possible.
4499 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4500 negative dentries on lock revocation.
4505 Description: Remounting a client read-only wasn't possible with a zconf mount
4506 Details : It wasn't possible to remount a client read-only with llmount.
4508 Severity : enhancement
4509 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4510 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4511 MPICH is included to add Lustre-specific ADIO interfaces.
4512 This is based closely on the UFS ADIO layer and only differs
4513 in file creation, in order to allow the OST striping to be set.
4514 This is user-contributed code and not supported by CFS.
4519 Description: extended inode attributes (immutable, append-only) work improperly
4520 when 2.4 and 2.6 kernels are used on client/server or vice versa
4521 Details : Introduce kernel-independent values for these flags.
4523 Severity : enhancement
4526 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4527 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4528 granularity. Now, Fractional MB granularity is be supported,
4529 this is very useful for low memory system.
4531 Severity : enhancement
4533 Description: Getattr by fid
4534 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4535 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4536 have locks and avoids getattr rpc requests for it on every stat.
4539 Frequency : Always, for filesystems larger than 2TB
4541 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4542 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4543 filesystems larger than 2TB to be mounted with mballoc enabled.
4546 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4548 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4549 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4550 filesystems larger than 2TB is mounted then the kernel will
4551 silently corrupt the start of the filesystem. CONFIG_LBD is
4552 enabled for all CFS-supported kernels, but the possibility of
4553 this happening with a modified kernel config exists.
4555 Severity : enhancement
4557 Description: add client O_DIRECT support for 2.6 kernels
4558 Details : It is now possible to do O_DIRECT reads and writes to files
4559 in the Lustre client mountpoint on 2.6 kernel clients.
4561 Severity : enhancement
4563 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4564 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4565 parallel, not waiting for response from every OST before sending
4566 a rpc to the next OST.
4571 Description: setattr vs write race when updating file timestamps
4572 Details : Client processes that update a file timestamp into the past
4573 right after writing to the file (e.g. tar) it is possible that
4574 the updated file modification time can be reset to the current
4575 time due to a race between processing the setattr and write RPC.
4577 Severity : enhancement
4579 Description: Bring 'lfs find' closer in line with regular Linux find.
4580 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4581 -print0 options and obtains all the needed info through the lustre
4584 Severity : enhancement
4586 Description: support up to 1024 configured devices on one node
4587 Details : change obd_dev array from statically allocated to dynamically
4588 allocated structs as they are first used to reduce memory usage
4593 Description: Flush dirty partially truncated pages during truncate
4594 Details : Immediatelly flush partially truncated pages in filter_setattr,
4595 this way we completely avoid having any pages in page cache on OST
4596 and can retire ugly workarounds during writes to flush such pages.
4601 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4602 Details : protect inode from truncation within vfs_unlink() context
4603 just take a reference before calling vfs_unlink() and release it
4604 when parent's i_sem is free.
4609 Description: last_id value checked outside lock on OST caused LASSERT failure
4610 Details : If there were multiple MDS->OST object precreate requests in
4611 flight, it was possible that the OST's last object id was checked
4612 outside a lock and incorrectly tripped an assertion. Move checks
4613 inside locks, and discard old precreate requests.
4616 Frequency : always, if extents are used on OSTs
4618 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4619 Details : OSTs using the extents format would not zero the high 16 bits of
4620 the index physical block number. This is not a problem for any
4621 OST filesystems smaller than 16TB, and no kernels support ext3
4622 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4623 new/modified files) and can be fixed for existing filesystems
4624 with e2fsprogs-1.39-cfs1.
4629 Description: import connection selection may be incorrect if timer wraps
4630 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4631 import connections to be ignored if the 32-bit jiffies counter
4632 wraps. Use a 64-bit jiffies counter.
4635 Frequency : very large clusters immediately after boot
4637 Description: LNET request buffers exhausted under heavy short-term load
4638 Details : If a large number of client requests are generated on a service
4639 that has previously never seen so many requests it is possible
4640 that the request buffer growth cannot keep up with the spike in
4641 demand. Instead of dropping incoming requests, they are held in
4642 the LND until the RPC service can accept more requests.
4645 Frequency : Sometimes during replay
4647 Description: Assertion failure in ll_local_open after replay.
4648 Details : If replay happened on an open request reply before we were able
4649 to set replay handler, reply will become not swabbed tripping the
4650 assertion in ll_local_open. Now we set the handler right after
4651 recognising of open request
4654 Frequency : very rare
4656 Description: kernel reports "badness in vsnprintf"
4657 Details : Reading from the "recovery_status" /proc file in small chunks
4658 may cause a negative length in lprocfs_obd_rd_recovery_status()
4659 call to vsnprintf() (which is otherwise harmless). Exit early
4660 if there is no more space in the output buffer.
4662 Severity : enhancement
4664 Description: clear OBD RPC statistics by writing to them
4665 Details : It is now possible to clear the OBD RPC statistics by writing
4666 to the "stats" file.
4671 Description: Client mtime is not the same on different clients after utimes
4672 Details : In some cases, the client was using the utimes() syscall on
4673 a file cached on another node. The clients now validate the
4674 ctime from the MDS + OSTs to determine which one is right.
4679 Description: Inability to activate failout mode
4680 Details : lconf script incorrectly assumed that in pythong string's numeric
4681 value is used in comparisons.
4684 Frequency : always with multiple stripes per file
4686 Description: Inefficient object allocation for mutli-stripe files
4687 Details : When selecting which OSTs to stripe files over, for files with
4688 a stripe count that divides evenly into the number of OSTs,
4689 the MDS is always picking the same starting OST for each file.
4690 Return the OST selection heuristic to the original design.
4695 Description: mount failures may take full timeout to return an error
4696 Details : Under some heavy load conditions it is possible that a
4697 failed mount can wait for the full obd_timeout interval,
4698 possibly several minutes, before reporting an error.
4699 Instead return an error as soon as the status is known.
4701 Frequency : quota enabled and large files being deleted
4703 Description: releasing more than 4GB of quota at once hangs OST
4704 Details : If a user deletes more than 4GB of files on a single OST it
4705 will cause the OST to spin in an infinite loop. Release
4706 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4711 Description: statfs data retrieved from /proc may be stale or zero
4712 Details : When reading per-device statfs data from /proc, in the
4713 {kbytes,files}_{total,free,avail} files, it may appear
4714 as zero or be out of date.
4717 Frequency : systems with MD RAID1 external journal devices
4719 Description: lconf's call to blkid is confused by RAID1 journal devices
4720 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4721 of returning all block devices that match the journal UUID.
4724 Frequency : always, for aggregate stripe size over 4GB
4726 Description: assertion fails when trying to use 4GB stripe size
4727 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4728 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4731 Frequency : always on ppc64
4733 Description: the first write on an ext3 filesystem with mballoc got stuck
4734 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4735 perform endianness conversion.
4737 ------------------------------------------------------------------------------
4739 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4741 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4742 INTEROPERATE with older versions automatically. Please read the
4743 user documentation before upgrading any part of a live system.
4744 * WARNING: Lustre networking configuration changes are required with
4745 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4748 * Support for kernels:
4749 2.6.9-22.0.2.EL (RHEL 4)
4750 2.6.5-7.244 (SLES 9)
4751 2.6.12.6 vanilla (kernel.org)
4754 Severity : enhancement
4755 Bugzilla : 7981/8208
4756 Description: Introduced Lustre Networking (LNET)
4757 Details : LNET is new networking infrastructure for Lustre, it includes
4758 a reorganized network configuration mode (see the user
4759 documentation for full details) as well as support for routing
4760 between different network fabrics. Lustre Networking Devices
4761 (LNDs) for the supported network fabrics have also been
4762 created for this new infrastructure.
4764 Severity : enhancement
4765 Description: Introduced Access control lists
4766 Details : clients can set ACLs on files and directories in order to have
4767 more fine-grained permissions than the standard Unix UGO+RWX.
4768 The MDS must be started with the "-o acl" mount option.
4770 Severity : enhancement
4771 Description: Introduced filesystem quotas
4772 Details : Administrators may now establish per-user quotas on the
4775 Severity : enhancement
4777 Description: Configuration change for the XT3
4778 The PTLLND is now used to run Lustre over Portals on the XT3
4779 The configure option(s) --with-cray-portals are no longer used.
4780 Rather --with-portals=<path-to-portals-includes> is used to
4781 enable building on the XT3. In addition to enable XT3 specific
4782 features the option --enable-cray-xt3 must be used.
4787 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4788 Details : A race between allocating a new llcd and re-getting the llcd_lock
4789 allowed another thread to grab newly-allocated llcd.
4791 Severity : enhancement
4793 Description: 2.6 OST async journal commit and locking fix to improve performance
4794 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4795 2.6 kernels are now async as they already were in 2.4 kernels so
4796 that they can commit concurrently with the network bulk transfer.
4797 For block-allocated files the filter allocation semaphore is held
4798 to avoid filesystem fragmentation during allocation. BKL lock
4799 removed for 2.6 xattr operations where it is no longer needed.
4804 Description: lconf incorrectly determined whether two IP networks could talk
4805 Details : In some more complicated routing and multiple-network
4806 configurations, lconf will avoid trying to make a network
4807 connection to a disjoint part of the IP space. It was doing the
4808 math incorrectly for one set of cases.
4813 Description: Fix for potential infinite loop processing records in an llog.
4814 Details : If an llog record is corrupted/zeroed, it is possible to loop
4815 forever in llog_process(). Validate the llog record length
4816 and skip the remainder of the block on error.
4819 Frequency : occasional (liblustre only)
4821 Description: liblustre could not open files whose last component is a symlink
4822 Details : sysio_path_walk() would incorrectly pass the open intent to
4823 intermediate path components.
4826 Frequency : rare (liblustre only with non-standard tuning)
4827 Bugzilla : 7201 (7350)
4828 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4829 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4830 after it has been released, only do this for referenced locks
4832 Severity : enhancement
4834 Description: specifying an (invalid) directory default stripe_size of -1
4835 would reset the directory default striping
4836 Details : stripe_size -1 was used internally to signal directory stripe
4837 removal, now use "all default" to signal dir stripe removal
4838 as a directory striping of "all default" is not useful
4841 Frequency : common for large clusters running liblustre clients
4843 Description: doing an ls when liblustre clients are running is slow
4844 Details : sending a glimpse AST to a liblustre client waits for every AST
4845 to time out, as liblustre clients will not respond. Since they
4846 cannot cache data we refresh the OST lock LVB from disk instead.
4848 Severity : enhancement
4850 Description: doing an ls at the same time as file IO can be slow
4851 Details : enqueue and other "small" requests can be blocked behind many
4852 large IO requests. Create a new OST IO portal for non-IO
4853 requests so they can be processed faster.
4856 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4858 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4859 created file with mode 000
4860 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4861 allow an ftruncate() as a non-root user to a file with mode 000.
4862 Lustre now respects this flag to disable mode checks when
4863 truncating a file owned by the user
4866 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4869 Description: Revoking locks from clients that went dead or catatonic might take
4871 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4872 cancellation of such locks instant on servers without waiting for
4873 any reply from clients. Clients drops these locks when cancel
4874 notification from server is received without replying.
4877 Frequency : liblustre-only, when liblustre client dies or becomes busy
4879 Description: Doing ls on Linux clients can take a long time with active
4881 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4882 granting such locks to it in the first place if possible. Locks
4883 are taken by proxy on the OST during the read or write and
4884 dropped immediately afterward. Add connect flags handling, do
4885 not grant locks to liblustre clients for glimpse ASTs.
4887 Severity : enhancement
4889 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4890 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4891 sequential reads and behaves badly with applications doing random
4892 reads. Improve it by reading ahead at least read region, and
4893 avoiding excessive large RPC for small reads.
4895 Severity : enhancement
4897 Description: Creating more than 1000 files for a single job may cause a load
4898 imbalance on the OSTs if there are also a large number of OSTs.
4899 Details : qos_prep_create() uses an OST index reseed value that is an
4900 even multiple of the number of available OSTs so that if the
4901 reseed happens in the middle of the object allocation it will
4902 still utilize the OSTs as uniformly as possible.
4907 Description: OST or MDS may oops in ping_evictor_main()
4908 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4909 but doesn't restart at beginning of obd_exports_timed list
4912 Severity : enhancement
4914 Description: improve by-nid export eviction on the MDS and OST
4915 Details : allow multiple exports with the same NID to be evicted at one
4916 time without re-searching the exports list.
4919 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4921 Description: MDS may oops in groups_free()
4922 Details : in rare race conditions a newly allocated group_info struct is
4923 freed again, and this can be NULL. The 2.4 compatibility code
4924 for groups_free() checked for a NULL pointer, but 2.6 did not.
4927 Frequency : common for liblustre clients doing little filesystem IO
4928 Bugzilla : 9352, 7313
4929 Description: server may evict liblustre clients accessing contended locks
4930 Details : if a client is granted a lock or receives a completion AST
4931 with a blocking AST already set it would not reply to the AST
4932 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4935 Frequency : lfs setstripe, only systems with more than 160 OSTs
4937 Description: unable to set striping with a starting offset beyond OST 160
4938 Details : llapi_create_file() incorrectly limited the starting stripe
4939 index to the maximum single-file stripe count.
4942 Frequency : LDAP users only
4944 Description: lconf did not handle in-kernel recovery with LDAP properly
4945 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4947 Severity : enhancement
4949 Description: bind OST threads to NUMA nodes to improve performance
4950 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4951 node and do their allocations there to localize memory access
4953 Severity : enhancement
4955 Description: llmount can determine client NID directly from Myrinet (GM)
4956 Details : the client NID code from gmnalnid was moved directly into
4957 llmount, removing the need to use this or specifying the
4958 client NID explicitly when mounting GM clients with zeroconf
4961 Frequency : if client is started with down MDS
4963 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4964 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4965 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4966 ptlrpc_import_delayed_req() because we didn't check if the
4967 request was interrupted, and we also didn't break out of the
4968 event loop if there was no timeout
4973 Description: data loss during non-page-aligned writes to a single file from
4974 both multiple nodes and multiple threads on one node at same time
4975 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4976 inconsistency led to false short-reads, that were cached and later
4977 used by ->prepare_write() to fill in partially written page,
4978 leading to data loss.
4981 Frequency : always, if lconf --abort_recovery used
4983 Description: lconf --abort_recovery fails with 'Operation not supported'
4984 Details : lconf was attempting to abort recovery on the MDT device and not
4987 Severity : enhancement
4989 Description: remove cleanup logs
4990 Details : replace lconf-generated cleanup logs with lustre internal
4991 cleanup routines. Eliminates the need for client-cleanup and
4994 Severity : enhancement
4996 Description: add support for EAs (user and system) on lustre filesystems
4997 Details : it is now possible to store extended attributes in the Lustre
4998 client filesystem, and with the user_xattr mount option it
4999 is possible to allow users to store EAs on their files also
5001 Severity : enhancement
5003 Description: Add possibility (config option) to show minimal available OST free
5005 Details : When compiled with --enable-mindf configure option, statfs(2)
5006 (and so, df) will return least minimal free space available from
5007 all OSTs as amount of free space on FS, instead of summary of
5008 free spaces of all OSTs.
5010 Severity : enhancement
5012 Description: do not expand extent locks acquired on OST-side
5013 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5014 by server: they are not cached anyway.
5017 Frequency : when mmap is used/binaries executed from Lustre
5019 Description: Unmmap pages before throwing them away from read cache.
5020 Details : llap_shrink cache now attempts to unmap pages before discarding
5021 them (if unmapping failed - do not discard). SLES9 kernel has
5022 extra checks that trigger if this unmapping is not done first.
5027 Description: lconf didn't resolve symlinks before checking to see whether a
5028 given mountpoint was already in use
5031 Frequency : when migrating failover services
5032 Bugzilla : 6395, 9514
5033 Description: When migrating a subset of services from a node (e.g. failback
5034 from a failover service node) the remaining services would
5035 time out and evict clients.
5036 Details : lconf --force (implied by --failover) sets the global obd_timeout
5037 to 5 seconds in order to quickly disconnect, but this caused
5038 other RPCs to time out too quickly. Do not change the global
5039 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5041 Severity : enhancement
5042 Frequency : if MDS is started with down OST
5043 Bugzilla : 9439,5706
5044 Description: Allow startup/shutdown of an MDS without depending on the
5045 availability of the OSTs.
5046 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5047 Add appropriate locking and lov-osc refcounts for safe
5048 cleaning. Add osc abort_inflight calls in case the
5049 synchronize never started.
5052 Frequency : occasional (Cray XT3 only)
5054 Description: root not authorized to access files in CRAY_PORTALS environment
5055 Details : The client process capabilities were not honoured on the MDS in
5056 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5057 been accessed by an authorized user then root was able to access
5058 the file on the local client also. The root user capabilities
5059 are now allowed on the MDS, as this environment has secure UID.
5062 Frequency : occasional
5064 Description: ldiskfs "too long searching" message happens too often
5065 Details : A debugging message (otherwise harmless) prints too often on
5066 the OST console. This has been reduced to only happen when
5067 there are fragmentation problems on the filesystem.
5072 Description: Division by zero in statfs when all OSCs are inactive
5073 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5074 lov_statfs divides by value returned by lov_get_stripecnt().
5078 Bugzilla : 9489, 3273
5079 Description: First write from each client to each OST was only 4kB in size,
5080 to initialize client writeback cache, which caused sub-optimal
5081 RPCs and poor layout on disk for the first written file.
5082 Details : Clients now request an initial cache grant at (re)connect time
5083 and so that they can start streaming writes to the cache right
5084 away and always do full-sized RPCs if there is enough data.
5085 If the OST is rebooted the client also re-establishes its grant
5086 so that client cached writes will be honoured under the grant.
5091 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5092 Details : Now I/O RPCs go to different portal number and (presumably) fast
5093 lock requests (and glimses) and other RPCs get their own service
5094 threads pool that should be able to service those RPCs
5097 Severity : enhancement
5099 Description: Ability to exchange lustre version between client and servers and
5100 issue warnings at client side if client is too old. Also for
5101 liblustre clients there is ability to refuse connection of too old
5103 Details : New 'version' field is added to connect data structure that is
5104 filled with version info. That info is later checked by server and
5108 Frequency : rare, liblustre only.
5109 Bugzilla : 9296, 9581
5110 Description: Two simultaneous writes from liblustre at offset within same page
5111 might proceed at the same time overwriting eachother with stale
5113 Details : I/O lock withing llu_file_prwv was released too early, before data
5114 actually was hitting the wire. Extended lock-holding time until
5115 server acknowledges receiving data.
5118 Frequency : extremely rare. Never observed in practice.
5120 Description: avoid generating lustre_handle cookie of 0.
5121 Details : class_handle_hash() generates handle cookies by incrementing
5122 global counter, and can hit 0 occasionaly (this is unlikely, but
5123 not impossible, because initial value of cookie counter is
5124 selected randonly). Value of 0 is used as a sentinel meaning
5125 "unassigned handle" --- avoid it. Also coalesce two critical
5126 sections in this function into one.
5128 Severity : enhancement
5130 Description: allow liblustre clients to delegate truncate locking to OST
5131 Details : To avoid overhead of locking, liblustre client instructs OST to
5132 take extent lock in ost_punch() on client's behalf. New connection
5133 flag is added to handle backward compatibility.
5135 Severity : enhancement
5136 Bugzilla : 4928, 7341, 9758
5137 Description: allow number of OST service threads to be specified
5138 Details : a module parameter allows the number of OST service threads
5139 to be specified via "options ost ost_num_threads={N}" in the
5140 OSS's /etc/modules.conf or /etc/modprobe.conf.
5144 Bugzilla : 6146, 9635, 9895
5145 Description: servers crash with bad pointer in target_handle_connect()
5146 Details : In rare cases when a client is reconnecting it was possible that
5147 the connection request was the last reference for that export.
5148 We would temporarily drop the export reference and get a new
5149 one, but this may have been the last reference and the export
5150 was just destroyed. Get new reference before dropping old one.
5152 Severity : enhancement
5153 Frequency : if client is started with failover MDS
5155 Description: Allow multiple MDS hostnames in the mount command
5156 Details : Try to read the configuration from all specified MDS
5157 hostnames during a client mount in case the "primary"
5160 Severity : enhancement
5162 Description: Stop sending data to evicted clients as soon as possible.
5163 Details : Check if the client we are about to send or are sending data to
5164 was evicted already. (Check is done every second of waiting,
5165 for which l_wait_event interface was extended to allow checking
5166 of exit condition at specified intervals).
5169 Frequency : rare, normally only when NFS exporting is done from client
5171 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5173 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5176 Frequency : always, if srand() is called before liblustre initialization
5178 Description: Liblustre uses system PRNG disturbing its usage by user application
5179 Details : Introduce internal to lustre fast and high-quality PRNG for
5180 lustre usage and make liblustre and some other places in generic
5181 lustre code to use it.
5183 Severity : enhancement
5184 Bugzilla : 9477, 9557, 9870
5185 Description: Verify that the MDS configuration logs are updated when xml is
5186 Details : Check if the .xml configuration logs are newer than the config
5187 logs stored on the MDS and report an error if this is the case.
5188 Request --write-conf, or allow starting with --old_conf.
5190 Severity : enhancement
5192 Description: Handle symlinks in the path when checking if Lustre is mounted.
5193 Details : Resolve intermediate symlinks when checking if a client has
5194 mounted a filesystem to avoid duplicate client mounts.
5199 Description: lconf can hit an error exception but still return success.
5200 Details : The lconf command catches the Command error exception at the top
5201 level script context and will exit with the associated exit
5202 status, but doesn't ensure that this exit status is non-zero.
5207 Description: failure of ptlrpc thread startup can cause oops
5208 Details : Starting a ptlrpc service thread can fail if there are a large
5209 number of threads or the server memory is very fragmented.
5210 Handle this without oopsing.
5213 Frequency : always, only if liblustre and non-default acceptor port was used
5215 Description: liblustre cannot connect to servers with non-default acceptor port
5216 Details : tcpnal_set_default_params() was not called and was therefore
5217 ignoring the environment varaible TCPNAL_PORT, as well as other
5218 TCPNAL_ environment variables
5223 Description: two objects could be created on the same OST for a single file
5224 Details : If an OST is down, in some cases it was possible to create two
5225 objects on a single OST for a single file. No problems other
5226 than potential performance impact and spurious error messages.
5230 Bugzilla : 5681, 9562
5231 Description: Client may oops in ll_unhash_aliases
5232 Details : Client dcache may become inconsistent in race condition.
5233 In some cases "getcwd" can fail if the current directory is
5239 Description: Inode refcounting problems in NFS export code
5240 Details : link_raw functions used to call d_instantiate without obtaining
5241 extra inode reference first.
5245 Bugzilla : 9942, 9903
5246 Description: Referencing freed requests leading to crash, memleaks with NFS.
5247 Details : We used to require that call to ll_revalidate_it was always
5248 followed by ll_lookup_it. Also with revalidate_special() it is
5249 possible to call ll_revalidate_it() twice for the same dentry
5250 even if first occurence returned success. This fix changes semantic
5251 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5252 reference on a request referred from the intent.
5253 ll_intent_release() then releases such a request.
5256 Frequency : rare, normally benchmark loads only
5258 Description: unlinked inodes were kept in memory on the client
5259 Details : If a client is repeatedly creating and unlinking files it
5260 can accumulate a lot of stale inodes in the inode slab cache.
5261 If there is no other client load running this can cause the
5262 client node to run out of memory. Instead flush old inodes
5263 from client cache that have the same inode number as a new inode.
5266 Frequency : SLES9 2.6.5 kernel and long filenames only
5267 Bugzilla : 9969, 10379
5268 Description: utime reports stale NFS file handle
5269 Details : SLES9 uses out-of-dentry names in some cases, which confused
5270 the lustre dentry revalidation. Change it to always use the
5274 Frequency : rare, unless heavy write-truncate concurrency is continuous
5275 Bugzilla : 4180, 6984, 7171, 9963, 9331
5276 Description: OST becomes very slow and/or deadlocked during object unlink
5277 Details : filter_destroy() was holding onto the parent directory lock
5278 while truncating+unlinking objects. For very large objects this
5279 may block other threads for a long time and slow overall OST
5280 responsiveness. It may also be possible to get a lock ordering
5281 deadlock in this case, or run out of journal credits because of
5282 the combined truncate+unlink. Solution is to do object truncate
5283 first in one transaction without parent lock, and then do the
5284 final unlink in a new transaction with the parent lock. This
5285 reduces the lock hold time dramatically.
5288 Frequency : rare, 2.4 kernels only
5290 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5291 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5292 before unmounting their filesystem in order to avoid deadlock.
5293 The kernel_locked() function in 2.4 kernels only checks whether
5294 the kernel lock is held, not whether it is this process that is
5295 holding it as 2.6 kernels do.
5300 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5301 Details : The client ptlrpc code may be trying to reconnect to a down
5302 server before a previous connection attempt has timed out.
5303 Increase the reconnect interval to be longer than the connection
5304 timeout interval to avoid sending duplicate connections to
5308 Frequency : echo_client brw_test command
5310 Description: fix echo_client to work with OST preallocated code
5311 Details : OST preallocation code (5137) didn't take echo_client IO path
5312 into account: echo_client calls filter methods outside of any
5313 OST thread and, hence, there is no per-thread preallocated
5314 pages and buffers to use. Solution: hijack pga pages for IO. As
5315 a byproduct, this avoids unnecessary data copying.
5319 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5320 Description: Client can oops in mdc_commit_close() after open replay
5321 Details : It was possible for the MDS to return an open request with no
5322 transaction number in mds_finish_transno() if the client was
5323 evicted, but without actually returning an error. Clients
5324 would later try to replay that open and may trip an assertion
5325 Simplify the client close codepath, and always return an error
5326 from the MDS in case the open is not successful.
5329 Frequency : rare, 2.6 OSTs only
5331 Description: OST may deadlock under high load on fragmented files
5332 Details : If there was a heavy load and highly-fragmented OST filesystems
5333 it was possible to have all the OST threads deadlock waiting on
5334 allocation of biovecs, because the biovecs were not released
5335 until the entire RPC IO was completed. Instead, release biovecs
5336 as soon as they are complete to ensure forward IO progress.
5338 Severity : enhancement
5340 Description: Support for specifying external journal device at mount
5341 Details : If an OST or MDS device is formatted with an external journal
5342 device, this device major/minor is stored in the ext3 superblock
5343 and may not be valid for failover. Allow detecting and
5344 specifying the external journal at mount time.
5349 Description: Mounting an MDS with pending unlinked files may cause oops
5350 Details : target_finish_recovery() calls mds_postrecov() which returned
5351 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5352 considers this an error and immediately begins cleaning up the
5353 lov, just after starting the mds_lov process
5355 Severity : enhancement
5357 Description: Implement 'lfs df' to report actual free space on per-OST basis
5358 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5359 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5360 usage of objects; '-h' to report in human readable format.
5362 ------------------------------------------------------------------------------
5364 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5371 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5372 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5373 at startup during orphan unlinking. Instead, skip buddy update
5374 before setup, as it will be regenerated after recovery is complete.
5379 Description: If an OST is inactive, its locks might reference stale inodes.
5380 Details : lov_change_cbdata() must iterate over all namespaces, even if
5381 they are inactive to clear inode references from the lock.
5383 Severity : enhancement
5384 Frequency : occasional, if non-standard max_dirty_mb used
5386 Description: Client will block write RPCs if not enough grant
5387 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5388 then the client will block writes while waiting for another RPC
5389 to complete instead of consuming its dirty limit. With change
5390 we get improved performance when max_dirty_mb is small.
5392 Severity : enhancement
5393 Bugzilla : 3389, 6253
5394 Description: Add support for supplementary groups on the MDS.
5395 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5396 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5397 lookups for user supplementary groups into a cache.
5401 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5402 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5403 for liblustre O_EXCL is enforced by MDS.
5407 Description: readdir never returns NULL in liblustre.
5408 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5409 dentry in struct dirent.
5413 Frequency : liblustre only, depends on application IO pattern
5414 Description: liblustre clients evicted if not contacting servers
5415 Details : Don't put liblustre clients into the ping_evictor list, so
5416 they will not be evicted by the pinger ever.
5418 Severity : enhancement
5420 Description: Add ability to evict clients by NID from MDS.
5421 Details : By echoing "nid:$NID" string into
5422 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5423 $NID would be instantly evicted from this MDS and from all active
5424 OSTs connected to it.
5428 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5429 Details : lookup_it_finish() used to query file size from OSTs that was not
5434 Description: service threads change working directory to that of init
5435 Details : Starting lustre service threads may pin the working directory
5436 of the parent thread, making that filesystem busy. Threads
5437 now change to the working directory of init to avoid this.
5441 Frequency : during shutdown only
5442 Description: shutdown with a failed MDS or OST can cause unmount to hang
5443 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5448 Frequency : 2.6 only
5449 Description: chmod/chown may include an extra supplementary group
5450 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5451 supplementary group and if none is specified this is used.
5454 Bugzilla : 5479 (6816)
5456 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5457 Details : If lookup is in progress on a file that is unlinked we might try
5458 to revalidate the inode and fail in revalidate after lookup is
5459 complete and ll_file_open() enqueues the open again but
5460 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5463 Frequency : always, if lconf --abort_recovery used
5465 Description: lconf --abort_recovery fails with 'Operation not supported'
5466 Details : lconf was attempting to abort recovery on the MDT device and not
5469 ------------------------------------------------------------------------------
5471 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5476 Frequency : rare (only unsupported configurations with a node running as an
5478 Bugzilla : 6514, 5137
5479 Description: Mounting a Lustre file system on a node running as an OST could
5481 Details : OSTs now preallocates memory needed to write out data at
5482 startup, instead of when needed, to avoid having to
5483 allocate memory in possibly low memory situations.
5484 Specifically, if the file system is mounted on on OST,
5485 memory pressure could force it to try to write out data,
5486 which it needed to allocate memory to do. Due to the low
5487 memory, it would be unable to do so and the node would
5488 become unresponsive.
5490 Severity : enhancement
5492 Description: Addition of lconf --service command line option
5493 Details : lconf now accepts a '--service <arg>' option, which is
5494 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5496 Severity : enhancement
5498 Description: Failover mode is now the default for OSTs.
5499 Details : By default, OSTs will now run in failover mode. To return to
5500 the old behaviour, add '--failout' to the lmc line for OSTs.
5502 Severity : enhancement
5504 Description: Health checks are now provided for MDS and OSTs
5505 Details : Additional detailed health check information on MSD and OSTs
5506 is now provided through the procfs health_check value.
5509 Frequency : occasional, depends on IO load
5511 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5512 after numerous create/delete cycles
5513 Details : The ext3 inode allocation policy would not allocate new inodes
5514 very well on the OSTs because there are no new directories
5515 being created. Instead we look for groups with free space if
5516 the parent directories are nearly full.
5521 Description: Network or server problems during mount may cause partially
5522 mounted clients instead of returning an error.
5523 Details : The config llog parsing code may overwrite the error return
5524 code during mount error handling, returning success instead
5530 Description: MDS can fail to allocate large reply buffers
5531 Details : After long uptimes the MDS can fail to allocate large reply
5532 buffers (e.g. zconf client mount config records) due to memory
5533 fragmentation or consumption by the buffer cache. Preallocate
5534 some large reply buffers so that these replies can be sent even
5535 under memory pressure.
5539 Frequency : rare (liblustre)
5540 Description: fsx running with liblustre complained that using truncate() to
5541 extend the file doesn't work. This patch corrects that issue.
5542 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5543 ATTR_SIZE and lsm use in llu_setattr_raw.
5547 Frequency : rare, only 2.6 kernels
5548 Description: Unusual file access patterns on the MDS may result in inode
5549 data being lost in very rare circumstances.
5550 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5551 code in the 2.6 kernel caused the inode and/or EA data not to
5552 be read from disk, causing single-file corruption.
5556 Frequency : rare, only 2.6 filesystems using extents
5557 Description: Heavy concurrent write and delete load may cause data corruption.
5558 Details : It was possible under high-load situations to have an extent
5559 metadata block in the block device cache from a just-unlinked
5560 file overwrite a newly-allocated data block. We now unmap any
5561 metadata buffers that alias just-allocated data blocks.
5565 Frequency : filesystems with default stripe_count larger than 77
5566 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5567 Details : lconf specifies an inode size of 4096 bytes when the default
5568 stripe_count is larger than 77. This conflicts with the default
5569 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5570 this case to avoid pinning too much memory for large EAs.
5572 ------------------------------------------------------------------------------
5574 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5579 Frequency : rare (extremely heavy IO load with hundreds of clients)
5581 Description: Client is evicted, gets IO error writing to file
5582 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5583 caused clients to be evicted for AST timeouts. The fixes in
5584 bug 5192 mean we no longer need to have such short AST timeouts
5585 so ldlm_timeout has been increased.
5588 Frequency : occasional during --force or --failover shutdown under load
5589 Bugzilla : 5949, 4834
5590 Description: Server oops/LBUG if stopped with --force or --failover under load
5591 Details : a collection of import/export refcount and cleanup ordering
5592 issues fixed for safer force cleanup
5595 Frequency : only filesystems larger than 120 OSTs
5596 Bugzilla : 5990, 6223
5597 Description: lfs getstripe would oops on a very large filesystem
5598 Details : lov_getconfig used kfree on vmalloc'd memory
5601 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5602 Bugzilla : 6242, 6243
5603 Description: reading from files that had been truncated to a non-zero size
5604 but never opened returned no data
5605 Details : ll_file_read() reads zeros from no-object files to EOF
5610 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5611 Details : There's little harm in aborting MDS/OSS recovery and letting it
5612 try again, so I removed the LASSERT and return an error instead.
5614 Severity : enhancement
5616 Description: New debugging infrastructure for tracking down data corruption
5617 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5618 (b) cover more of the client-side code path, and (c) try to narrow
5619 down where problems occurred
5623 Bugzilla : 3819, 4364, 4397, 6313
5624 Description: Racing close and eviction MDS could cause assertion in mds_close
5625 Details : It was possible to get multiple mfd references during close and
5626 client eviction, leading to one thread referencing a freed mfd.
5628 Severity: : enhancement
5629 Bugzilla : 3262, 6359
5630 Description: Attempts to reconnect to servers are now more aggressive.
5631 Details : This builds on the enhanced upcall-less recovery that was added
5632 in 1.4.2. When trying to reconnect to servers, clients will
5633 now try each server in the failover group every 10 seconds. By
5634 default, clients would previously try one server every 25 seconds.
5639 Description: After recovery, certain operations trigger a failed
5640 assertion on a client.
5641 Details : Failing over an mds, using lconf -d --failover, while a
5642 client was doing a readdir() call would cause the client to
5643 LBUG after recovery completed and the readdir() was resent.
5645 Severity : enhancement
5647 Description: Default groups are now added by lconf
5648 Details : You can now run lconf --group <servicename> without having to
5649 manually add groups with lmc.
5652 Frequency : occasional
5654 Description: Nodes with an elan id of 0 trigger a failed assertion
5657 Frequency : always when accessing e.g. tty/console device nodes
5659 Description: tty and some other devices nodes cannot be used on lustre
5660 Details : file's private_data field is used by device data and lustre
5661 values in there got lost. New field was added to struct file to
5662 store fs-specific private data.
5665 Frequency : when exporting Lustre via NFS
5667 Description: NFSD failed occasionally when looking up a path component
5668 Details : NFSD is looking up ".." which was broken in ext3 directories
5669 that had grown large enough to become hashed.
5672 Frequency : Clusters with multiple interfaces not on the same subnet
5674 Description: Nodes will repeatedly try to reconnect to an interface which it
5675 cannot reach and report an error to the log.
5676 Details : Extra peer list entries will be created by lconf with some peers
5677 unreachable. lconf now validates the peer before adding it.
5680 Frequency : Only if a default stripe is set on the filesystem root.
5682 Description: Setting a default stripe on the filesystem root prevented the
5683 filesystem from being remounted.
5684 Details : The client was sending extra request flags in the root getattr
5685 request and did not allocate a reply buffer for the dir EA.
5688 Frequency : occasional, higher if lots of files are accessed by one client
5689 Bugzilla : 6159, 6097
5690 Description: Client trips assertion regarding lsm mismatch/magic
5691 Details : While revalidating inodes the VFS looks up inodes with ifind()
5692 and in rare cases can find an inode that is being freed.
5693 The ll_test_inode() code will free the lsm during ifind()
5694 when it finds an existing inode and then the VFS later attaches
5695 this free lsm to a new inode.
5699 Bugzilla : 6422, 7030
5700 Description: MDS deadlock between mkdir and client eviction
5701 Details : Creating a new file via mkdir or mknod (starting a transaction
5702 and getting the ns lock) can deadlock with client eviction
5703 (gets ns lock and trying to finish a synchronous transaction).
5706 Frequency : occasional
5707 Description: While starting a server, the fsfilt_ext3 module could not be
5709 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5710 kernels. Previously, lconf would still use the ext3 name
5711 when trying to load modules. Now, it will correctly use
5712 ext3 on 2.4 and ldiskfs on 2.6.
5714 Severity : enhancement
5715 Description: The default stripe count has been changed to 1
5716 Details : The interpretation of the default stripe count (0, to lfs
5717 or lmc) has been changed to mean striping across a single
5718 OST, rather than all available. For general usage we have
5719 found a stripe count of 1 or 2 works best.
5721 Severity : enhancement
5722 Description: Add support for compiling against Cray portals.
5723 Details : Conditional compiling for some areas that are different
5727 Frequency : occasional
5728 Bugzilla : 6409, 6834
5729 Description: Creating files with an explicit stripe count may lead to
5730 a failed assertion on the MDS
5731 Details : If some OSTs are full or unavailable, creating files may
5732 trigger a failed assertion on the MDS. Now, Lustre will
5733 try to use other servers or return an error to the
5737 Frequency : occasional
5739 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5740 caused lock timeout during readahead (since 1.4.2).
5741 Details : Processes doing readahead might match a lock that hasn't been
5742 granted yet if there are overlapping and conflicting lock
5743 requests. The readahead process waits on ungranted lock
5744 (original lock is CBPENDING), while OST waits for that process
5745 to cancel CBPENDING read lock and eventually evicts client.
5747 Severity : enhancement
5749 Description: Initial enabling of flock support for clients
5750 Details : Implements fcntl advisory locking and file status functions.
5751 This feature is provided as an optional mount flag (default
5752 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5753 locking are implemented yet, and those that are are not guaranteed
5754 to be completely correct in production environments.
5755 mount -t lustre -o [flock|noflock] ...
5758 Frequency : occasional
5760 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5761 assertion in the ext3 JBD (journaling) layer.
5762 Details : The b_committed_data struct is protected by the big kernel lock
5763 in 2.4 kernels, serializing journal_commit_transaction() and
5764 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5765 access to this struct. In 2.6 kernels there is finer grained
5766 locking to improve SMP performance of the JBD layer.
5770 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5775 Description: merge and cleanup kernel patches.
5778 -----------------------------------------------------------------------------
5780 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5782 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5783 versions of Lustre. Please update all servers and clients to
5784 version 1.4.2 or later at the same time. You must also run
5785 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5786 to update the configuration logs.
5788 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5789 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5790 - swab llog records of type '0' so we get proper header size/idx (5861)
5791 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5792 - fix rename of one directory over another leaking an inode (5953)
5793 - avoid SetPageDirty on 2.6 (5981)
5794 - don't re-add just-being-destroyed locks to the waiting list (5653)
5795 - when creating new directories, inherit the parent's custom
5796 striping settings if present parent (3048)
5797 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5798 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5799 - don't hold client locks on temporary worklist from l_lru (5666)
5800 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5801 - automatically evict dead clients (5921)
5802 - Update file size properly in create+truncate+fstat case (6196)
5803 - Do not unhash mountpoint dentries, do not allow removal of
5805 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5806 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5807 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5808 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5809 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5810 - Correctly handle reads of files with no objects (6243)
5811 - lctl recover will also mark a device active if deactivate used (5933)
5813 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5814 - allow --write-conf on an MDS with different nettype than client (5619)
5815 - don't write config llogs to MDS for mounts not from that MDS (5617)
5816 - lconf should create multiple TCP connections from a client (5201)
5817 - init scripts are now turned off by default; run chkconfig --on
5818 lustre and chkconfig --on lustrefs to use them
5819 - upcalls are no longer needed for clients to recover to failover
5821 - add --abort-recovery option to lconf to abort recovery on device
5823 - add support for an arbitrary number of OSTs (3026)
5824 - Quota support protocol changes.
5825 - forward compatibility changes to wire structs (6007)
5826 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5827 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5828 - improved "lustre status" script
5829 - initialize blocksize for non-regular files (6062)
5830 - added --disable-server and --disable-client configure options (5782)
5831 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5832 - Vanilla 2.4.29 support
5833 - increase maximum number of obd devices to 520 (6242)
5834 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5835 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5836 SLES 9 2.6.5 kernels
5837 - limit stripes per file to 160 (the maximum EA size) (6093)
5839 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5842 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5843 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5844 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5845 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5846 - small ext3 extents cleanups and fixes (5733)
5847 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5848 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5849 - avoid needless client->OST connect, fix handle mismatch (5317)
5850 - fix DLM error path that led to out-of-sync client, long delays (5779)
5851 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5852 - fix several locking issues related to i_size (5492,5624,5654,5672)
5853 - don't move pending lock onto export if it is already evicted (5683)
5854 - fix kernel oops when creating .foo in unlinked directory (5548)
5855 - fix deadlock in obdfilter statistics vs. object create (5811)
5856 - use time_{before,after} to avoid timer jiffies wrap (5882)
5857 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5858 - Do not leak request if server was not able to process it (5154)
5859 - If mds_open unable to find parent dir, make that negative lookup(5154)
5860 - don't create new directories with extent-mapping (5909, 5936)
5862 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5863 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5864 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5865 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5866 - support for RHEL4 kernel on i686 (5773)
5867 - provide error messages when incompatible logs are encountered (5898)
5869 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5870 * version 1.4.0.10 (1.4.1 release candidate 1)
5872 - don't keep a lock reference when lock is not granted (4238)
5873 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5874 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5875 - drop import inflight refcount on signal_completed_replay error (5255)
5876 - unlock page after async write error during send (3677)
5877 - handle missing objects in filter_preprw_read properly (5265)
5878 - no transno return for symlink open, don't save no-trasno open (3440)
5879 - don't try to complete elan receive that already failed (4012)
5880 - free RPC server reply state on error (5406)
5881 - clean up thread from ptlrpc_start_thread() on error (5160)
5882 - readahead could read extra page into cache that wasn't ejected (5388)
5883 - prevent races in class_attach/setup/cleanup/detach (5260)
5884 - don't dereference de->d_inode after l_dput of de (5458)
5885 - use "int" for stripe value returned from lock_to_stripe (5544)
5886 - mballoc allocation and error-checking fixes in 2.6 (5504)
5887 - block device patches to fix I/O request sizes in 2.6 (5482)
5888 - look up hostnames for IB nals (5602)
5889 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5890 - don't start multiple acceptors for the same port (5277)
5891 - fix incorrect LASSERT in mds_getattr_name (5635)
5892 - export a proc file for general "ping" checking (5628)
5893 - fix "lfs check" to not block when the MDS is down (5628)
5895 - service request history (4965)
5896 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5897 - create an "evict_client" /proc entry on OSTs, like the MDS has
5898 - fix mount usage message, return errors per mount(8) (5168)
5899 - change grep [] to grep "[]" in tests so they work in more UMLs
5900 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5901 - remove ancient LOV_MAGIC_V0 EA support (5047)
5902 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5903 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5904 - allow readahead allocations to fail when low on memory (5383)
5905 - mmap locking landed again, after considerable improvement (2828)
5906 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5908 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5911 - send OST transaction number in read/write reply to free req (4966)
5912 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5913 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5914 - the watchdog thread now runs as interruptible (5246)
5915 - flock/lockf fixes (but it's still disabled, pending 5135)
5916 - don't use EXT3 constants in llite code (5094)
5917 - memory shortage at startup could cause assertion (5176)
5919 - reorganization of lov code
5920 - single portals codebase
5922 - add extents/mballoc support (5025)
5923 - direct I/O reads in the obdfilter (4048)
5924 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5926 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5929 - send OST transaction number in read/write reply to free req (4966)
5930 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5931 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5932 - the watchdog thread now runs as interruptible (5246)
5933 - handle missing objects in filter_preprw_read properly (5265)
5934 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5935 - drop import inflight refcount on signal_completed_replay error (5255)
5936 - unlock page after async write error during send (3677)
5937 - return original error code on reconstructed replies (3761)
5938 - no transno return for symlink open, don't save no-trasno open (3440)
5940 - add pid to ldlm debugging output (4922)
5941 - bump the watchdog timeouts -- we can't handle 30sec yet
5942 - extra debugging for orphan dentry/inode bug (5259)
5944 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5947 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5948 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5949 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5950 - fix NULL dereference in /proc/sys/portals/routes (4827)
5951 - allow failed mdc_close() operations to be interrupted (4561)
5952 - stop precreate on OST before MDS would time out on it (4778)
5953 - don't send partial-page writes before EOF from client (4410)
5954 - discard client grant for sub-page writes on large-page clients (4520)
5955 - don't free dentries not owned by NFS code, check generation (4806)
5956 - fix lsm leak if mds_create_objects() fails (4801)
5957 - limit debug_daemon file size, always print CERROR messages (4789)
5958 - use transno after validating reply (3892)
5959 - process timed out requests if import state changes (3754)
5960 - update mtime on OST during writes, return in glimpse (4829)
5961 - add mkfsoptions to LDAP (4679)
5962 - use ->max_readahead method instead of zapping global ra (5039)
5963 - don't interrupt __l_wait_event() during strace
5965 - add software watchdogs to catch hung threads quickly (4941)
5966 - make lustrefs init script start after nfs is mounted
5967 - fix CWARN/ERROR duplication (4930)
5968 - return async write errors to application if possible (2248)
5969 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5970 - print NAL number in %x format (4645)
5971 - update barely-supported suse-2.4.21-171 series (4842)
5972 - support for sles 9 %post scripts
5973 - support for building 2.6 kernel-source packages
5974 - support for sles km_* packages
5976 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5979 - ignore -ENOENT errors in osc_destroy (3639)
5980 - notify osc create thread that OSC is being cleaned up (4600)
5981 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5982 - reconstruct ost_handle() like mds_handle() (4657)
5983 - create a new thread to do import eviction to avoid deadlock (3969)
5984 - let lconf resolve symlinked-to devices (4629)
5985 - don't unlink "objects" from directory with default EA (4554)
5986 - hold socknal file ref over connect in case target is down (4394)
5987 - allow more than 32000 subdirectories in a single directory (3244)
5988 - fix blocks count for O_DIRECT writes (3751)
5989 - OST returns ENOSPC from object create when no space left (4539)
5990 - don't send truncate RPC if file size isn't changing (4410)
5991 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5992 - bind to privileged port in socknal and tcpnal (3689)
5994 - rate limit CERROR/CWARN console message to avoid overload (4519)
5995 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5996 - basic mmap support (3918)
5997 - kernel patch series update from b1_4 (4711)
5999 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6002 - avoid crash during MDS cleanup with OST shut down (2775)
6003 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6004 - don't use bad inodes on the MDS (3744)
6005 - dynamic object preallocation to improve recovery speed (4236)
6006 - don't hold spinlock over lock dumping or change debug flags (4401)
6007 - don't zero obd_dev when it is force cleaned (3651)
6008 - print grants to console if they go negative (4431)
6009 - "lctl deactivate" will stop automatic recovery attempts (3406)
6010 - look for existing locks in ldlm_handle_enqueue() (3764)
6011 - don't resolve lock handle twice in recovery avoiding race (4401)
6012 - revalidate should check working dir is a directory (4134)
6014 - don't always mark "slow" obdfilter messages as errors (4418)
6016 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6019 - don't close LustreDB during write_conf until it is done (3860)
6020 - fix typo in lconf for_each_profile (3821)
6021 - allow dumping logs from multiple threads at one time (3820)
6022 - don't allow multiple threads in OSC recovery (3812)
6023 - fix debug_size parameters (3864)
6024 - fix mds_postrecov to initialize import for llog ctxt (3121)
6025 - replace config semaphore with spinlock (3306)
6026 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6027 - don't allow enqueue to complete on a destroyed export (3822)
6028 - down write_lock before checking llog header bitmap (3825)
6029 - recover from lock replay timeout (3764)
6030 - up llog sem before sending rpc (3652)
6031 - reduce ns lock hold times when setting kms (3267)
6032 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6033 - fix NULL deref and obd_dev leak on setup error (3312)
6034 - replace some LBUG about llog ops with error handling (3841)
6035 - don't match INVALID dentries from d_lookup and spin (3784)
6036 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6037 - fix invalid assertion in ptlrpc_set_wait (3880)
6039 - add libwrap support for the TCP acceptor (3996)
6040 - add /proc/sys/portals/routes for non-root route listing (3994)
6041 - allow setting MDS UUID in .xml (2580)
6042 - print the stack of a process that LBUGs (4228)
6044 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6047 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6048 - make sure to unset replay flag from failed open requests (3440)
6049 - if default stripe count is 0, use OST count for inode size (3636)
6050 - update parent mtime/ctime on client for create/unlink (2611)
6051 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6052 - free recovery state on server during a forced cleanup (3571)
6053 - unregister_reply for resent reqs (3063)
6054 - loop back devices mounting and status check on 2.6 (3563)
6055 - fix resource-creation race that can provoke i_size == 0 (3513)
6056 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6057 - more debugging for page-accounting assertion (3746)
6058 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6059 - avoid deadlock after precreation failure (3758)
6060 - fix race and lock order deadlock in orphan handling (3450, 3750)
6061 - add validity checks when grabbing inodes from l_ast_data (3599)
6063 - add /proc/.../recovery_status to obdfilter (3428)
6064 - lightweight CDEBUG infrastructure, debug daemon (3668)
6065 - change default OSC RPC parameters to be better on small clusters
6066 - turn off OST read cache for files smaller than 32MB
6067 - install man pages and include them in rpms (3100)
6068 - add new init script for (un)mounting lustre filesystems (2593)
6069 - run chkconfig in %post for init scripts (3701)
6070 - drop scimac NAL (unmaintained)
6072 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6075 - clean kiobufs before and after use (3485)
6076 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6077 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6078 - change init script to fail more clearly if not run as root (1528)
6079 - allow clients to reconnect during replay (1742)
6080 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6081 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6082 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6083 - fix cond_resched (3554)
6084 - only evict unfinished clients after recovery (3515)
6085 - allow bulk resend, prevent data loss (3570)
6086 - dynamic ptlrpc request buffer allocation (2102)
6087 - don't allow unlinking open directory if it isn't empty (2904)
6088 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6089 - remove extraneous obd dereference causing LASSERT failure (3334)
6090 - don't use get_cycles() when creating temp. files on the mds (3156)
6091 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6092 - handle EEXIST for set-stripe, set proper directory name (3336)
6094 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6095 - fix ksocknal_fmb_callback() error messages (2918)
6097 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6100 - don't copy lvb into (possibly NULL) reply on error (2983)
6101 - don't deref dentry after dput, don't free lvb on error (2922)
6102 - use the kms to determine writeback rpc length (2947)
6103 - increment oti_logcookies when osc is inactive (2948)
6104 - update client's i_blocks count via lvb messages (2543)
6105 - handle intent open/close of special files properly (1557)
6106 - mount MDS with errors=remount-ro, like obdfilter (2009)
6107 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6108 - don't use cancelling-locks' kms values (2947)
6109 - use highest lock extent for kms, not last one (2925)
6110 - don't dereference ERR_PTR() dentry in error handling path (3107)
6111 - fix thread race in portals_debug_dumplog() (3122)
6112 - create lprocfs device entries at setup instead of at attach (1519)
6113 - common AST error handler, don't evict client on completion race (3145)
6114 - zero nameidata in detach_mnt in 2.6 (3118)
6115 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6116 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6117 - initialize RPC timeout timer earlier for 2.6 (3219)
6118 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6119 - print nal/nid for unknown nid (3258)
6120 - additional checks for oscc recovery before doing precreate (3284)
6121 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6122 - don't crash in mdc_close for bad permissions on open (3285)
6123 - zero i_rdev for non-device files (3147)
6124 - clear page->private before handing to FS, better assertion (3119)
6125 - tune the read pipeline (3236)
6126 - fix incorrect decref of invalidated dentry (2350)
6127 - provide read-ahead stats and refine rpc in flight stats (3328)
6128 - don't hold journal transaction open across create RPC (3313)
6129 - update atime on MDS at close time (3265)
6130 - close LDAP connection when recovering to avoid server load (3315)
6131 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6132 - don't leak open file on MDS after open resend (3325)
6133 - serialize filter_precreate and filter_destroy_precreated (3329)
6134 - loop device shouldn't call sync_dev() for nul device (3092)
6135 - clear page cache after eviction (2766)
6136 - resynchronize MDS->OST in background (2824)
6137 - refuse to mount the same filesystem twice on same mountpoint (3394)
6138 - allow llmount to create routes for mounting behind routers (3320)
6139 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6140 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6141 - fix rare problem with rename on htree directories (3417)
6143 - allow default OST striping configuration per directory (1414)
6144 - fix compilation for qswnal for 2.6 kernels (3125)
6145 - increase maximum number of MDS request buffers for large systems
6146 - change liblustreapi to be useful for external progs like lfsck (3098)
6147 - increase local configuration timeout for slow disks (3353)
6148 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6150 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6153 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6154 - don't overwrite extent policy data in reply if lock was blocked (2901)
6155 - drop filter export grants atomically with removal from device (2663)
6156 - del obd_self_export from work_list in class_disconnect_exports (2908)
6157 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6158 - swab reply message in mdc_close, other PPC fixes (2464)
6159 - fix destroying of named logs (2325)
6160 - overwrite old logs when running lconf --write_conf (2264)
6161 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6162 - fix race in target_handle_connect (2898)
6163 - mds_reint_create() should take same inode create lock (2926)
6164 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6165 - don't close files for self_export to avoid uninitialized obd (2936)
6166 - allow MDS with the same name as client node (2939)
6167 - hold dentry reference for closed log files for unlink (2325)
6168 - reserve space for all logs during transactions (2059)
6169 - don't evict page beyond end of stripe extent (2925)
6170 - don't oops on a deleted current working directory (2399)
6171 - handle hard links to targets without a parent properly (2517)
6172 - don't dereference NULL lock when racing during eviction (2867)
6173 - don't grow lock extents when lots of conflicting locks (2919)
6175 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6178 - account for cache space usage on clients to avoid data loss (974)
6179 - lfsck support in lustre kernel code (2349)
6180 - reduce journal credits needed for BRW writes (2370)
6181 - orphan handling to avoid losing space on client/server crashes
6182 - ptlrpcd can be blocked, stopping ALL progress (2477)
6183 - use lock value blocks to assist in proper KMS, faster stat (1021)
6184 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6185 - recovery for initial connections (2355)
6186 - fixes for mds_cleanup_orphans (1934)
6187 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6188 - block all file creations until orphan recovery completes (1901)
6189 - client remove rq_connection from request struct (2423)
6190 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6191 - recovery timer race (2670)
6192 - mdc_close recovey bug (2532)
6193 - ptlrpc cleanup bug (2710)
6194 - mds timeout on local locks (2588)
6195 - namespace lock held during RPCs (2431)
6196 - handle interrupted sync write properly (2503)
6197 - don't try to handle a message that hasn't been replied to (2699)
6198 - client assert failure during cleanup after abort recovery (2701)
6199 - leak mdc device after failed mount (2712)
6200 - ptlrpc_check_set allows timedout requests to complete (2714)
6201 - wait for inflight reqs when ptlrpcd finishes (2710)
6202 - make sure unregistered services are removed from the srv_list
6203 - reset bulk XID's when resending them (caught by 1138 test)
6204 - unregister_bulk after timeout
6205 - fix lconf error (2694)
6206 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6207 - readahead locks pages, leaves pending causing memory pressure (2673)
6208 - increase OST request buffers to 4096 on large machines (2729)
6209 - fix up permission of existing directories in simple_mkdir (2661)
6210 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6211 - don't assign transno to errored transactions (2742)
6212 - don't delete objects on OST if given a bogus objid from MDS (2751)
6213 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6214 - if rq_no_resend, then timeout request after recovery (2432)
6215 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6216 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6217 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6218 - MDS recovery shouldn't delete objects that it has given out (2730)
6219 - if enqueue arrives after completion, don't clobber LVB (2819)
6220 - don't unlock pages twice when trigger_group_io returns error (2814)
6221 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6222 - don't write pages to disk if there was an error (1450)
6223 - don't ping imports that have recovery disabled (2676)
6224 - take buffered bytes into account when balancing socknal conn (2817)
6225 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6226 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6227 - remove little-endian swabbing of llog records (1987)
6228 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6229 - retry reposting request buffers if they fail (1191)
6230 - grow extent at grant time to avoid granting a revoked lock (2809)
6231 - lock revoke doesn't evict page if covered by a second lock (2765)
6232 - disable VM readahead to avoid reading outside lock extents (2805)
6234 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6235 - updated kernel patches for hp-2.4.20 kernel (2681)
6237 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6240 - fix truncated write corruption (2366)
6241 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6243 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6244 - ensure len(uuid) < 37 in lmc (1171)
6245 - fix ia64 OOPS in llog_test (2255)
6246 - zero end of page at obdfilter for partial page writes (2648)
6247 - don't leave stale dentries around after renames (bug 2428)
6248 - fix timeouts when evicting a client with a single lock held (2642)
6249 - set deadline for the initial HELLO message to drain (2634)
6250 - print out dotted-quad IP addresses in the socknal (2302)
6252 - additional debugging for MDS client eviction problem (2443)
6253 - fix mkfsoptions support for osts (2603, 2604)
6255 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6258 - add series for the vanilla 2.6.0 kernel
6259 - add series for the vanilla 2.4.24 kernel
6260 - add series for a cray x86/64 UL kernel drop
6261 - fix xattr patches for the vanilla 2.4.19 series
6263 - generate true UUIDs in lmc (1171)
6264 - have portals stack dumping break in UML (2466)
6265 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6266 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6267 - wake callback waiting threads on client eviction (2460)
6268 - Add --ptldebug and --subsystem to lmc (1719)
6269 - update assertion to allow safe interrupt allocation
6270 - set rq_no_resend for cancel requests (2432)
6271 - recalculate ptlrpcd timeout after resend (2494)
6272 - call vfs_rmdir when removing pending directories (2368)
6273 - fix renaming a file to itself (2429)
6274 - lmc creates a default one-stripe lov (2454)
6275 - expand procfs space to handle large clusters (2326)
6276 - increase UML stack to avoid overflow
6277 - update lconf's list of debug and subsystem masks
6278 - fix lfs find --obd (2510)
6279 - /proc tunable for disabling filter read caching (2591)
6280 - stop rpm packages from altering slapd.conf (2301)
6281 - disable nagle in the socknal under 0conf (2578)
6282 - choose mds inode size based on stripe count (2572)
6283 - fix kernel-source rpm problems (2516)
6285 - add --disable-doc to avoid pdf generation (2421)
6286 - update documentation, tests, type-os, comments
6287 - avoid format warnings on ia64
6288 - remove the TOE NAL
6289 - tiny code cleanups by removing unused fields
6291 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6294 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6295 - remove the most blatant lies from BUILDING file (2371)
6296 - change default debug level to reasonable production setting
6297 - reduce client side cache size to reduce cache flush time
6298 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6299 - make TCP zerocopy and pinger support enabled by default (2476)
6300 - sync writes completed after process exits caused crashes (2319)
6301 - maintain correct mount count on the MDS (2356)
6302 - backout 1557, because 2316 wasn't really fixed
6303 - better file I/O statistics gathering in /proc
6304 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6305 - another kernel patch to fix zero-copy TCP function export
6306 - don't take duplicate lock when processing re-sent getattr (2420)
6307 - lctl uses obd_self_export instead of creating new conn (2353)
6308 - MDS/OST recovery case which requires object creation asserted (2425)
6309 - move lfs from /usr/sbin to /usr/bin in packages
6310 - fix race between mds_client_add and mds_client_free (2417)
6311 - use kmalloc instead of slabs in portals (2430)
6312 - don't create duplicate records when a failover MDS is present (2442)
6313 - remove unnecessary mount age check (2332)
6314 - don't remove directory inodes from locks prematurely (2451)
6315 - don't break if MDS service name is the same as hostname (2103)
6316 - fix races in client write RPC generation when cache full (2482)
6318 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6321 - remove now-unused request->rq_obd (278)
6322 - if an allocation fails, print out how much memory we've used (1933)
6323 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6324 - add the "configurable stack size" patch to most series files (1256)
6325 - ability to write large log records, for 100+ OST configs (2306)
6326 - fix NULL deref when filter_prep fails (2314)
6327 - fix operator precedence error in filter_sync
6328 - dynamic allocation of socknal TX descriptors (2315)
6329 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6330 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6331 - ensure that CWARN messages in Portals always get to the syslog
6332 - __init/__exit are not for prototype decls (ldlm_init/exit)
6333 - x86-64 compile warning fixes
6334 - fix gateway LMC keyword conflict (2318)
6335 - fix MDS lock inversions in getattr/reint paths (1844)
6336 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6337 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6338 - DLM race condition prevented some lock evictions (2328)
6339 - ENOMEM detection and retry on socknal sends (2230)
6340 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6341 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6342 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6343 - better error messages when a client is rejected during recovery (1505)
6344 - avoid cancelling locks which were never granted, after failure (2330)
6345 - fix i_sem/journal inversion in mds_client_add (2333)
6346 - fix truncate/getattr lock cycle deadlock (2334)
6347 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6348 - fix two rare exit paths which could leak an l_lock() ref (2321)
6349 - fencepost error in MDS/OST orphan recovery (2226)
6350 - make log record alignment 8 bytes (1988)
6351 - lstripe now fails when requested offset > ost_count (2237)
6352 - ensure that all kernel series have a complete list.h (1607)
6353 - fix crashes in special-file operations (2316)
6354 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6356 - allow configurable automake binary, for testing new versions
6357 - small update to the lfs documentation
6359 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6361 * fix negative export reference count in fsfilt_sync (2312)
6363 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6364 * release candidate 0.9.1
6366 - orphans are moved into the PENDING directory for possible recovery
6367 - replayed opens now open by fid for orphan/rename safety (1042)
6368 - last close of an orphan inode generates a transno (683)
6369 - chdir() and mount() now pin the directory entry (1020)
6370 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6371 - discard very old requests without processing them (1502)
6372 - remove some common, well-understood CERRORs (1505)
6373 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6374 - clear "grant" flags in OST replies until OST grant code lands (1644)
6375 - fix read performance by not clobbering i_blksize on client (1598)
6376 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6377 - make LRU size a /proc tunable, clears locks when reduced (707)
6378 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6379 - server locks take references on exports now (1558)
6380 - build fixes for 2.4.20-rh trees (1663)
6381 - return an error from lov_create if all OSCs are inactive (1751)
6382 - fix import levels when a reconnect happens without a timeout (1597)
6383 - exit early from mds_open if we get a lookup error (1749)
6384 - partial page read at EOF wouldn't wait for disk before sending (1642)
6385 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6386 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6387 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6388 - write/truncate lock inversion (1639)
6389 - Don't auto-load obdclass, portals modules during cleanup (1495)
6390 - fix timestamps from jumping to "now" (1763)
6391 - extra journal assertions (1648)
6392 - add an extra multiunlink test (1771)
6393 - fix read_record/write_record API (1776)
6394 - fix leak of offset_extent, possible incorrect i_size later (1772)
6395 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6396 - Debugging for the kqswnal_get_idle_tx problems (1820)
6397 - Allow recovery to be attempted multiple times (1536)
6398 - Write out MDS last_rcvd file after it is first created (1600)
6399 - Fix tx_descriptor leak in failed transmit situations (1827)
6400 - ext3 journaling fixes for assertion failure after IO error (1871)
6401 - class_export_put() on freed export after completion AST error (1896)
6402 - Fix revalidate looping in VFS (1322)
6403 - Don't access a freed export during MDS_REINT timeout (1521)
6404 - Add open-unlink recovery support on the MDS (1673,1764)
6405 - Return an error if no MDS data was read from last_rcvd (1946)
6406 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6407 - Don't setup a disk device that doesn't match exported UUID (317)
6408 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6409 - avoid committing NULL handle in force close
6410 - local.sh is now a one-stripe LOV configuration
6411 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6412 - fix ext3 htree duplicate directory entry corruption (1516)
6413 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6414 - update to new LOV EA format (2097)
6415 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6417 2003-06-15 Phil Schwan <phil@clusterfs.com>
6420 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6421 - per-import/export recovery handling (958, 931, 959)
6422 - multiple last-rcvd slots, for serving multiple FSes (949)
6423 - connections are again shared between multiple imp/exports (963, 964)
6424 - "umount -f" would hang if any requests needed to be sent (393, 978)
6425 - avoid pinning large req buffer by copying for queued messages (989)
6426 - add "uuid" to "lctl device" command to help upcalls (991)
6427 - "open" RPCs with transnos would confuse recovery counters (1037)
6428 - do proper endian conversion of all wire messages (288, 340, 891)
6429 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6430 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6431 - fix ptlrpc_connection leak in target_handle_connect (1174)
6432 - fix import refcounting bug in OST and MDS cleanup (1134)
6433 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6434 - fix very unlikely obd_types race condition (501)
6435 - remove osc_open hack for echo_client (1187)
6436 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6437 - a failure in read_inode2 leads to deadlock (1139)
6438 - cancel ack-locks as soon as transaction is committed (1072)
6439 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6440 - make sure to commitrw after any preprw to avoid deadlock (1162)
6441 - failing to execute a file in a lustre FS would lock inode (1203)
6442 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6443 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6444 - obd_uuid2tgt would walk off the end of the list (1255)
6445 - on IA64 the osc would give portals incorrect bulk size (1258)
6446 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6447 - fix lock inversion caused by new llite matching code (1282)
6448 - limit the number of dirty pages on a client to 10MB (1286)
6449 - timed out locks were not being corrected cancelled (1289)
6450 - fix O_DIRECT above 4GB on IA-32 (1292)
6451 * major user-visible changes
6452 - fail out/fail over policy now controlled by the upcall (993)
6454 - add OBD_PING to check server availability and failure (954)
6455 - lustre messages are now sent in sending host order (288, 340, 891)
6456 - add eadatalen to MDS getattr reply (340)
6457 - OST read replies may contain second buffer, with per-page status (593)
6459 2003-03-11 Phil Schwan <phil@clusterfs.com>
6462 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6463 - failed open()s could cause deadlock; fixed (867, 869)
6464 - stop cancelling OST locks when files are closed (481)
6465 - overlapping XID spaces caused network corruption (851, 853)
6466 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6467 - setattr_raw added, to do single-RPC, server-side setattrs
6468 - lmc/lconf syntax change for OST UUIDs
6469 - fix crashy race condition between ptlrpc_free_req and osc_close
6470 - don't use request in mdc_enqueue if we hit a timeout (889)
6471 - don't set the inode i_size for regular files from the MDS (896)
6472 - handle out of order completion AST (842)
6473 - don't LBUG if a lock request times out after receiving AST (913)
6474 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6475 - if a bad lock AST arrives, send an error instead of dropping entirely
6476 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6477 - fix leak in bulk IO when only partially completed (899, 900, 926)
6478 - fix O_DIRECT for ia64 (55)
6479 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6480 - C-z after timeout could hang a process forever; fixed (977)
6482 - client-side I/O cache (678, 924, 929, 941, 970)
6484 - READPAGE and SETATTRs which don't take server-side locks get
6487 2003-02-11 Phil Schwan <phil@clusterfs.com>
6490 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6491 - Fix fsfilt_extN_readpage() to read a full page of directory
6492 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6493 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6494 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6495 - Most intent operations were removed, in favour of a new RPC mode
6496 that does a single RPC to the server and bypasses most of the VFS
6497 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6498 - Aggressively cancel local locks on DLM servers
6499 - mds_reint_unlink sends EA to the client if it's the last nlink.
6500 client uses that EA to unlink OST objects.
6501 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6502 - recursive symlinks were fixed (439)
6503 - fixed NULL deref in DEBUG_REQ
6504 - filter_update_lastobjid no longer calls sync, which annoyed extN
6505 - fixed multi-client small-writes to a single file problem (445)
6506 - fixed mtime updates during file writes (607)
6507 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6508 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6509 - fixed Linux OST size reporting problem (444, 656)
6510 - OST now updates object mtime with writes or setattr (607, 619)
6511 - client verifies file size before zeroing page past EOF (445)
6512 - OST now writes last allocated objid to disk with allocation (108)
6513 - LOV on echo now works (409)
6515 - mds_reint_unlink sends a new buffer, with the EA included. this
6516 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6517 set if a regular file was being unlinked, and it was the last link
6518 - use PtlGet from the target for bulk writes (315)
6519 - OST now updates object mtime with writes or setattr (607, 619)
6520 - LDLM now has a grant-time callback to revalidate locked items, if
6522 - Many MDS operations were reorganized to combat race conditions
6524 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6525 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6527 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6530 - Fully reactivate OST imports after reconnection (512, others)
6531 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6532 - More graceful error handling for truncating on dead OST (515)
6533 - Don't error out unless we're actually accessing dead stripes (474)
6534 - Fix garbage sizes when stripes are missing (410)
6535 - LRU counters were broken, causing constant lock purge (433, 432)
6536 - garbage on read from stripes with failed OSTs (441)
6537 - mark OSCs as active before reconnecting during recovery (438)
6538 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6539 - lfind did not preserve OST order in output (443)
6540 - symlinks cause hung clients, incorrect data (439)
6541 - stop dereferencing request after dropping refcount (457)
6542 - don't LASSERT(spin_is_locked) on non-SMP (455)
6543 - fixes for many rename() bugs
6544 - fstat didn't correctly synchronize attributes (399)
6545 - server must handle lock cancellation during blocking AST prep (487)
6546 - bulk descriptors were free()d too soon (511)
6547 - fix paths in lconf, which would load incorrect modules (451, 507)
6548 - fix confusing lconf 'host not found' error message (386)
6549 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6550 - fix race condition in mdc_blocking_ast() for inode access (526)
6551 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6552 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6553 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6554 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6555 - fix lstripe in several ways (kernel side) (527)
6556 - fix request leak in ldlm_cli_enqueue (262)
6557 - incorrect OSC was marked inactive after OST failure
6558 - call mds_fs_cleanup before unmounting filesystem (524)
6559 - fix races between taking ns_lock and ldlm_lock_change_resource
6560 - fix races updating LOV export open file list
6561 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6562 - fix recovery NULL deref in ldlm_cli_cancel_unused
6563 - fix some DLM races by using new hash table for lock handles (419)
6564 - permit the client to specify desired inodes, at replay
6565 - duplicate requests when we queue them for replay reintegration
6566 - fix last_rcvd offset calculation
6567 - sync after each recovered transaction, so we always make progress
6568 - never, not always, ERESTART requests without transnos
6569 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6571 - skip replay if the MDS says that the client is already connected
6572 - don't check for a recovery-enabled export to match lctl's UUID
6573 - don't INC_USE_COUNT for phantom exports
6574 - don't crash when cleaning up phantom exports (567)
6575 - don't double-finish or set replay data for errored mdc_open requests
6576 - abort requests when they time out, so we don't get old replies
6577 - send/receive replies for AST messages again
6578 - if the client says that it doesn't have the lock, cancel it on the
6580 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6581 mark it as destroyed, it will all work out when decref is called
6582 - fix module use counts (22, 581)
6584 - ASTs now expect a reply (server cancels lock on error reply)
6586 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6589 - fix many simultaneous client startup (392)
6590 - fix dentry->d_it clobbering
6591 - credentials weren't being shipped for readdir/getattr operations
6592 - remove invalid assertions triggered during some concurrent MD
6594 - proper Lustre versions added (336, 389)
6595 - fix memory leak for create error case (398)
6596 - fix LOV locking bug that would get cli/srv out of sync
6597 - fix echo client over LOV (409)
6598 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6599 - fix double-O_EXCL intent crash (424)
6600 - avoid sending multiple lock CANCELs (352)
6602 - MDS can do multi-client recovery (modulo bugs in new code)
6604 - many updates, edits, cleanups
6606 2002-11-18 Phil Schwan <phil@clusterfs.com>
6609 - fix null d_it dereference (346)
6610 - fix full OST/dbench hang (333)
6611 - fix permission problem with file removal (286)
6612 - fix removal of OSCs from LOV when they fail
6613 - fix NULL deref during bulk timeout (214)
6614 - fix problems related to multiple filesystems on one MDS (241)
6615 - fixed serious subtle metadata locking bugs
6616 - free locks on clients when inodes are removed due to memory
6618 - fix inode pointer in lock data (285)
6619 - partial support for multiple MDS on a single host (241)
6620 - data locks weren't cancelled at clear_inode time (290, 311)
6621 - intent locks could lead to unbounded lock growth (205)
6622 - added a maximum lock count, an LRU list, and a flusher
6623 - fix multiple rename (365)
6624 - properly abstracted the echo client
6625 - OSC locked 1 byte too many; fixed
6626 - rewrote brw callback code:
6627 - fixed recovery bugs related to LOVs (306)
6628 - fixed too-many-pages-in-one-write crash (191)
6629 - fixed (again) crash in sync_io_timeout (214)
6630 - probably fixed callback-related race (385)
6632 - Add capability to MDS protocol
6633 - LDLM cancellations and callbacks on different portals
6635 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6638 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6639 - defer open object destruction to close time (601981, 138)
6640 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6641 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6642 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6643 - don't call obd_setattr in ll_file_release for destroyed objects
6644 * protocol change to lustre_msg: move |version| and add |flags|
6645 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6646 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6647 * added replay of create, unlink, link and rename operations during
6648 MDS failover; recovery should be much more robust now
6649 * remove failed OSCs from LOVs (only lov_create uses this so far)
6650 * the lustre-HOWTO was brought (more) up to date (582544)
6652 2002-10-23 Phil Schwan <phil@clusterfs.com>
6655 - in-use dentries weren't being reused properly (617851)
6656 - prevent multiple LDLM setup (599178)
6657 - fix LOV size calculations for truncate (617853)
6658 - fix client handling of MDS intent errors (POSIX)
6659 - fix permission bug in lovstripe.c test (624321)
6660 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6661 - truncate past end of file could corrupt data
6662 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6663 - a race in recovery could return ETIMEDOUT to apps (623947)
6664 - building outside the source directory was fixed
6665 * the lustre-HOWTO was brought (more) up to date (582544)
6666 * major progress was made on recovery functionality
6668 2002-10-10 Phil Schwan <phil@clusterfs.com>
6671 - recovery deadlock fix
6672 - rm -rf causes LBUG fix (617817)
6673 - file open by multiple tasks fix (618962)
6674 - directory permissions bugs (602707 and 620007)
6675 - journal_stop fixed with locking (611313)
6676 - O_APPEND failures resolved (618273, perhaps 614459)
6677 - lconf PATH fix (619770)
6678 - IA64 build fix (621450)
6679 - RPC buffer sizes scale with amount of memory
6681 2002-10-01 Phil Schwan <phil@clusterfs.com>
6684 - locks would be cancelled without throwing away data pages,
6685 resulting in inconsistent data (605627)
6686 - inode attributes were not always being refreshed (605627, 612449)
6687 - lconf now continues to cleanup after lctl reports an error
6688 - MDS now enforces user permissions (602707)
6689 - lprocfs cleanup fixed, but not yet enabled (614157)
6690 - fixed infinite server hang, should a client not respond to an AST
6691 - avoid going into recovery if user calls readlink() with a buffer
6692 that's too small (613941)
6693 - AST RPCs no longer require replies (614867) -- this may be changed
6694 - don't crash server if client sends an IOV that's too big (611336)
6695 - fixed lock conversion deadlock (611892)
6696 - fixed the following of symlinks (614622)
6697 * recovery: the server can remove locks from a client that dies, other
6698 clients can make progress
6699 * more extN patch fixes
6700 * compile-time configurable ptlrpc buffer allocations
6702 - collaborative read cache document
6703 - Lustre Lite Performance CDR document-in-progress
6705 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6708 - fix typo in patch-2.4.18
6710 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6713 - clear ptlrpc request each time in handle_incoming_request()
6714 - unlink of files now destroys the object on the OST
6716 2002-09-19 Peter Braam <braam@clusterfs.com>
6718 * add hard link support
6719 * change obdfile creation method
6720 * kernel patch changed
6722 2002-09-19 Peter Braam <braam@clusterfs.com>
6725 - stack overflow bug in extN fixed
6727 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6729 * documentation updates
6730 - add man pages for config tools
6731 - update tests/README to describe testing with new config tools
6732 - finish metadata API descriptions
6733 * bug fixes and cleanups
6734 - statfs workaround for 16TB limit
6735 - LOV stripe allocation improved, can stripe on subset of OSTs
6736 - LOV file size/IO offset was wrong for files > 4GB in size
6737 - object EA data was being dropped, caused files to be unreadable
6738 - memory overflow with non-LOV OST caused memory corruption
6739 - fixed regression tests to work with new config tools, obdfilter
6740 - fixed bug when directory size became larger than 1 block
6741 - fixed bug (for single client case) when PWD was deleted
6742 - invalidate local directory pages when doing intent-based ops
6743 - avoid LDLM oops when lock callback contained bad data
6745 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6747 * documentation updates
6748 * bug fixes and cleanups
6749 - configuration tools
6752 - 64-bit compile warnings
6753 - 64-bit internal statfs data
6755 * test_brw on persistent OST devices
6757 * lprocfs (disabled)
6759 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6761 * documentation updates
6762 * bug fixes and cleanups
6763 * configuration tools
6765 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6768 * many small fixes to 0_5_4
6769 * io/network handling
6770 * thinkos in MDS operations
6772 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6775 * crucial basic fixes to 0.5.3
6776 * IOR, Iozone work over Elan
6779 2002-08-07 Phil Schwan <phil@clusterfs.com>
6780 * version 0_5_3, our first alpha
6781 * we use the new Portals iovs
6782 * documentation updates
6783 * bug fixes and cleanups
6784 * small changes in the DLM wire protocol
6786 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6787 * version 0_5_1 with some initial stability,
6788 * locking on MD and file I/O.
6789 * documentation updates
6790 * several bug fixes since 0.5.0
6791 * small changes in wire protocol
6793 2002-07-18 Phil Schwan <phil@clusterfs.com>
6795 * delivered as Lustre Light Alpha
6796 * fixed a crash after handling invalid MDS requests
6797 * fixed directory pages for architectures with non-4k pages sizes
6799 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6800 * release version v0_4_4
6801 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6803 * Fixes a number of interruption problems with OST operations.
6804 * Update documentation for portals header changes
6805 * Move all wire protocol structs/defines to lustre_idl.h
6806 * Fixes symlink length bug.
6807 * Add tcpdump to repository.
6809 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6810 * release version v0_4_3
6811 * Fixes statfs for inodes on extN.
6812 * Fixes bug in runtests which would delete /etc/hosts.
6813 * Use 64-bit object IDs wherever possible (not into VFS though)
6814 Remove ost_get_info, which is unused by lustre, and out of date.
6816 2002-07-03 Peter Braam <braam@clusterfs.com>
6817 * release version v0_4_2 Fixes a lookup error (type not passed)
6818 * move forward to head of Portals
6819 * move forward to latest Lustre kernel
6821 2002-06-25 Peter Braam <braam@clusterfs.com>
6822 * release version v0_4_1. Hopefully stable on single node use.