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 3.10.0-1160.25.1.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 4.18.0-240.22.1.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 3.10.0-1160.25.1.el7 (RHEL7.9)
27 4.12.14-95.48 (SLES12 SP4)
28 4.12.14-122.66 (SLES12 SP5)
29 4.12.14-197.75 (SLES15 SP1)
30 5.3.18-24.61 (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 4.18.0-240.22.1.el8 (RHEL8.3)
41 4.18.0-305.3.1.el8 (RHEL8.4)
42 4.4.120-92.70 (SLES12 SP2)
43 4.4.180-94.100 (SLES12 SP3)
44 4.4.0-131 (Ubuntu 16.04)
45 5.4.0-37 (Ubuntu 20.04)
46 5.4.0 (vanilla kernel.org)
47 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
48 * Recommended ZFS version: 2.0.0
49 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
50 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
51 stack overflow. For more information, please refer to bugzilla 17630.
52 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
53 the re-exporting nodes to be mounted with "32bitapi" mount option
55 --------------------------------------------------------------------------------
59 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
60 for currently supported client and server kernel versions.
61 * Server primary kernels built and tested during release cycle:
62 3.10.0-1062.18.1.el7 (RHEL7.7)
63 3.10.0-1127.19.1.el7 (RHEL7.8)
64 3.10.0-1160.6.1.el7 (RHEL7.9)
65 * Other server kernels known to build and work at some point (others may also work):
66 3.10.0-862.14.4.el7 (RHEL7.5)
67 3.10.0-957.27.2.el7 (RHEL7.6)
68 4.18.0-80.11.2.el8 (RHEL8.0)
69 4.18.0-147.8.1.el8 (RHEL8.1)
70 4.18.0-193.28.1.el8 (RHEL8.2)
71 4.18.0-240.1.1.el8 (RHEL8.3)
72 4.4.120-92.70 (SLES12 SP2)
73 4.4.180-94.100 (SLES12 SP3)
74 4.4.0-131 (Ubuntu 16.04)
75 4.15.0-32 (Ubuntu 18.04)
76 5.4.0-48 (Ubuntu 20.04)
77 vanilla linux 5.4.0 (ZFS + ldiskfs)
78 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
79 * Client primary kernels built and tested during release cycle:
80 3.10.0-1062.18.1.el7 (RHEL7.7)
81 3.10.0-1127.19.1.el7 (RHEL7.8)
82 3.10.0-1160.6.1.el7 (RHEL7.9)
83 4.12.14-95.48 (SLES12 SP4)
84 4.12.14-122.20 (SLES12 SP5)
85 4.12.14-197.75 (SLES15 SP1)
86 5.3.18-24.24 (SLES15 SP2)
87 4.15.0-48 (Ubuntu 18.04)
88 * Other clients known to build on these kernels at some point (others may also work):
89 3.10.0-862.14.4.el7 (RHEL7.5)
90 3.10.0-957.27.2.el7 (RHEL7.6)
91 4.14.0-49.13.1.el7a (RHEL7.5)
92 4.14.0-115.2.2.el7a (RHEL7.6)
93 4.18.0-80.11.2.el8 (RHEL8.0)
94 4.18.0-147.8.1.el8 (RHEL8.1)
95 4.18.0-193.28.1.el8 (RHEL8.2)
96 4.18.0-240.1.1.el8 (RHEL8.3)
97 4.4.120-92.70 (SLES12 SP2)
98 4.4.180-94.100 (SLES12 SP3)
99 4.4.0-131 (Ubuntu 16.04)
100 5.4.0-37 (Ubuntu 20.04)
101 5.4.0 (vanilla kernel.org)
102 * Recommended e2fsprogs version: 1.45.6.wc5 or newer
103 * Recommended ZFS version: 2.0.0
104 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
105 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
106 stack overflow. For more information, please refer to bugzilla 17630.
107 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
108 the re-exporting nodes to be mounted with "32bitapi" mount option
110 --------------------------------------------------------------------------------
114 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
115 for currently supported client and server kernel versions.
116 * Server primary kernels built and tested during release cycle:
117 3.10.0-1062.1.1.el7 (RHEL7.7)
118 * Other server kernels known to build and work at some point (others may also work):
119 3.10.0-862.14.4.el7 (RHEL7.5)
120 3.10.0-957.27.2.el7 (RHEL7.6)
121 4.18.0-80.7.1.el8 (RHEL8.0)
122 4.4.120-92.70 (SLES12 SP2)
123 4.4.180-94.100 (SLES12 SP3)
124 4.4.0-131 (Ubuntu 16.04)
125 4.15.0-32 (Ubuntu 18.04)
126 vanilla linux 4.6.7 (ZFS only)
127 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
128 * Client primary unpatched kernels built and tested during release cycle:
129 3.10.0-1062.1.1.el7 (RHEL7.7)
130 4.12.14-95.29 (SLES12 SP4)
131 4.15.0-48 (Ubuntu 18.04)
132 * Other clients known to build on unpatched kernels at some point (others may also work):
133 3.10.0-862.14.4.el7 (RHEL7.5)
134 3.10.0-957.27.2.el7 (RHEL7.6)
135 4.14.0-49.13.1.el7a (RHEL7.5)
136 4.14.0-115.2.2.el7a (RHEL7.6)
137 4.18.0-80.7.1.el8 (RHEL8.0)
138 4.4.120-92.70 (SLES12 SP2)
139 4.4.180-94.100 (SLES12 SP3)
140 4.4.0-131 (Ubuntu 16.04)
141 vanilla linux 4.15.0 (ZFS only)
142 * Recommended e2fsprogs version: 1.45.2.wc1 or newer
143 * Recommended ZFS / SPL version: 0.7.13
144 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
145 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
146 stack overflow. For more information, please refer to bugzilla 17630.
147 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
148 the re-exporting nodes to be mounted with "32bitapi" mount option
149 * Note for Ubuntu 14.04.5:
150 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
151 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
153 --------------------------------------------------------------------------------
157 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
158 for currently supported client and server kernel versions.
159 * Server known to build on the following kernels (others may also work):
160 2.6.32-696.18.7.el6 (RHEL6.9)
161 3.10.0-862.14.4.el7 (RHEL7.5)
162 3.10.0-957.el7 (RHEL7.6)
163 4.4.120-92.70 (SLES12 SP2)
164 4.4.162-94.69 (SLES12 SP3)
165 4.4.0-131 (Ubuntu 16.04)
166 4.15.0-32 (Ubuntu 18.04)
167 vanilla linux 4.6.7 (ZFS only)
168 * ldiskfs needs an ldiskfs patch series for that kernel, ZFS does not
169 * Client known to build on unpatched kernels (others may also work):
170 2.6.32-696.18.7.el6 (RHEL6.9)
171 2.6.32-754.6.3.el6 (RHEL6.10)
172 3.10.0-862.14.4.el7 (RHEL7.5)
173 3.10.0-957.el7 (RHEL7.6)
174 4.14.0-49.13.1.el7a (RHEL7.5)
175 4.14.0-115.2.2.el7a (RHEL7.6)
176 4.4.120-92.70 (SLES12 SP2)
177 4.4.162-94.69 (SLES12 SP3)
178 4.4.0-131 (Ubuntu 16.04)
179 4.15.0-32 (Ubuntu 18.04)
180 vanilla linux 4.15.0 (ZFS only)
181 * Recommended e2fsprogs version: 1.44.3.wc1 or newer
182 * Recommended ZFS / SPL version: 0.7.9
183 * Tested with ZFS / SPL version: 0.7.9
184 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
185 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
186 stack overflow. For more information, please refer to bugzilla 17630.
187 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
188 the re-exporting nodes to be mounted with "32bitapi" mount option
189 * Note for Ubuntu 14.04.5:
190 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
191 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
193 --------------------------------------------------------------------------------
195 2018-04-03 Intel Corporation
197 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
198 for currently supported client and server kernel versions.
199 * Server known to build on patched kernels:
200 2.6.32-431.29.2.el6 (RHEL6.5)
201 2.6.32-504.30.3.el6 (RHEL6.6)
202 2.6.32-573.26.1.el6 (RHEL6.7)
203 2.6.32-642.15.1.el6 (RHEL6.8)
204 2.6.32-696.18.7.el6 (RHEL6.9)
205 3.10.0-693.21.1.el7 (RHEL7.4)
206 3.0.101-0.47.71 (SLES11 SP3)
207 3.0.101-107 (SLES11 SP4)
208 3.12.74-60.64.40 (SLES12 SP1)
209 4.4.114-92.64 (SLES12 SP2)
210 4.4.114-94.11 (SLES12 SP3)
211 3.13.0-101 (Ubuntu 14.04, ZFS only)
212 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
213 4.4.0-116 (Ubuntu 16.04)
214 vanilla linux 4.6.7 (ZFS only)
215 * Client known to build on unpatched kernels:
216 2.6.32-431.29.2.el6 (RHEL6.5)
217 2.6.32-504.30.3.el6 (RHEL6.6)
218 2.6.32-573.26.1.el6 (RHEL6.7)
219 2.6.32-642.15.1.el6 (RHEL6.8)
220 2.6.32-696.18.7.el6 (RHEL6.9)
221 3.10.0-693.21.1.el7 (RHEL7.4)
222 3.0.101-0.47.71 (SLES11 SP3)
223 3.0.101-107 (SLES11 SP4)
224 3.12.74-60.64.40 (SLES12 SP1)
225 4.4.114-92.64 (SLES12 SP2)
226 4.4.114-94.11 (SLES12 SP3)
227 3.13.0-101 (Ubuntu 14.04)
228 4.4.0-85.108 (Ubuntu 14.04.5 LTS)
229 4.4.0-116 (Ubuntu 16.04)
231 * Recommended e2fsprogs version: 1.42.13.wc6 or newer
232 * Recommended ZFS / SPL version: 0.7.6
233 * Tested with ZFS / SPL version: 0.7.6
234 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
235 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
236 stack overflow. For more information, please refer to bugzilla 17630.
237 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
238 the re-exporting nodes to be mounted with "32bitapi" mount option
239 * Note for Ubuntu 14.04.5:
240 The 4.4.0 series is the HWE Kernel derived from 16.04 (Xenial).
241 See: https://wiki.ubuntu.com/Kernel/LTSEnablementStack
243 --------------------------------------------------------------------------------
245 2017-07-05 Intel Corporation
247 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
248 for currently supported client and server kernel versions.
249 * Server known to build on patched kernels:
250 2.6.32-431.29.2.el6 (RHEL6.5)
251 2.6.32-504.30.3.el6 (RHEL6.6)
252 2.6.32-573.26.1.el6 (RHEL6.7)
253 2.6.32-642.15.1.el6 (RHEL6.8)
254 2.6.32-696.3.1.el6 (RHEL6.9)
255 3.10.0-514.21.1.el7 (RHEL7.3)
256 3.0.101-0.47.71 (SLES11 SP3)
257 3.0.101-100 (SLES11 SP4)
258 3.12.74-60.64.40 (SLES12 SP1)
259 4.4.59-92.17 (SLES12 SP2)
260 vanilla linux 4.6.7 (ZFS only)
261 * Client known to build on unpatched kernels:
262 2.6.32-431.29.2.el6 (RHEL6.5)
263 2.6.32-504.30.3.el6 (RHEL6.6)
264 2.6.32-573.26.1.el6 (RHEL6.7)
265 2.6.32-642.15.1.el6 (RHEL6.8)
266 2.6.32-696.3.1.el6 (RHEL6.9)
267 3.10.0-514.21.1.el7 (RHEL7.3)
268 3.0.101-0.47.71 (SLES11 SP3)
269 3.0.101-100 (SLES11 SP4)
270 3.12.74-60.64.40 (SLES12 SP1)
271 4.4.59-92.17 (SLES12 SP2)
273 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
274 * Recommended ZFS / SPL version: 0.6.5.9
275 * Tested with ZFS / SPL version: 0.6.5.9
276 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
277 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
278 stack overflow. For more information, please refer to bugzilla 17630.
279 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
280 the re-exporting nodes to be mounted with "32bitapi" mount option
282 --------------------------------------------------------------------------------
284 12-07-2016 Intel Corporation
286 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
287 for currently supported client and server kernel versions.
288 * Server known to build on patched kernels:
289 2.6.32-431.29.2.el6 (RHEL6.5)
290 2.6.32-504.30.3.el6 (RHEL6.6)
291 2.6.32-573.26.1.el6 (RHEL6.7)
292 2.6.32-642.11.1.el6 (RHEL6.8)
293 3.10.0-514.el7 (RHEL7.3)
294 3.0.101-0.47.71 (SLES11 SP3)
295 3.0.101-80 (SLES11 SP4)
296 3.12.62-60.64.8 (SLES12 SP1)
297 vanilla linux 4.5.7 (ZFS only)
298 * Client known to build on unpatched kernels:
299 2.6.32-431.29.2.el6 (RHEL6.5)
300 2.6.32-504.30.3.el6 (RHEL6.6)
301 2.6.32-573.26.1.el6 (RHEL6.7)
302 2.6.32-642.11.1.el6 (RHEL6.8)
303 3.10.0-514.el7 (RHEL7.3)
304 3.0.101-0.47.71 (SLES11 SP3)
305 3.0.101-80 (SLES11 SP4)
306 3.12.62-60.64.8 (SLES12 SP1)
308 * Recommended e2fsprogs version: 1.42.13.wc5 or newer
309 * Recommended ZFS / SPL version: 0.6.5.8
310 * Tested with ZFS / SPL version: 0.6.5.8
311 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
312 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
313 stack overflow. For more information, please refer to bugzilla 17630.
314 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
315 the re-exporting nodes to be mounted with "32bitapi" mount option
317 --------------------------------------------------------------------------------
319 02-29-2016 Intel Corporation
321 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
322 for currently supported client and server kernel versions.
323 * Server known to build on patched kernels:
324 2.6.32-431.29.2.el6 (RHEL6.5)
325 2.6.32-504.30.3.el6 (RHEL6.6)
326 2.6.32-573.12.1.el6 (RHEL6.7)
327 3.10.0-327.3.1.el7 (RHEL7.2)
328 3.0.101-0.47.71 (SLES11 SP3)
329 3.0.101-68 (SLES11 SP4)
330 vanilla linux 4.2.1 (ZFS only)
331 * Client known to build on unpatched kernels:
332 2.6.32-431.29.2.el6 (RHEL6.5)
333 2.6.32-504.30.3.el6 (RHEL6.6)
334 2.6.32-573.12.1.el6 (RHEL6.7)
335 3.10.0-327.3.1.el7 (RHEL7.2)
336 3.0.101-0.47.71 (SLES11 SP3)
337 3.0.101-68 (SLES11 SP4)
340 * Recommended e2fsprogs version: 1.42.13.wc4 or newer
341 * Recommended ZFS / SPL version: 0.6.4.2
342 * Tested with ZFS / SPL version: 0.6.4.2
343 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
344 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
345 stack overflow. For more information, please refer to bugzilla 17630.
346 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
347 the re-exporting nodes to be mounted with "32bitapi" mount option
349 --------------------------------------------------------------------------------
351 03-10-2015 Intel Corporation
353 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
354 for currently supported client and server kernel versions.
355 * Server known to build on patched kernels:
356 2.6.32-431.29.2.el6 (RHEL6.5)
357 2.6.32-504.8.1.el6 (RHEL6.6)
358 3.0.101-0.46 (SLES11 SP3)
359 * Client known to build on unpatched kernels:
360 2.6.32-431.29.2.el6 (RHEL6.5)
361 2.6.32-504.8.1.el6 (RHEL6.6)
362 3.10.0-123.20.1.el7 (RHEL7)
363 3.0.101-0.46 (SLES11 SP3)
364 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
365 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
366 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
367 stack overflow. For more information, please refer to bugzilla 17630.
368 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
369 the re-exporting nodes to be mounted with "32bitapi" mount option
371 Severity : enhancement
373 Description: control OST-index in IDIF via ROCOMPAT flag.
374 Details : Introduce new flag OBD_ROCOMPAT_IDX_IN_IDIF that is stored in the
375 last_rcvd file. For new formatted OST device, it will be auto set;
376 for the case of upgrading from old OST device, you can enable it
377 via the lproc interface osd-ldiskfs.index_in_idif. With such flag
378 enabled, for new created OST-object, its IDIF-in-LMA will contain
379 the OST-index; for the existing OST-object, the OSD will convert
380 old format IDIF as new format IDIF with OST-index stored in the
381 LMA EA when accessing such OST-object or via OI scrub. Once such
382 flag is enabled, it cannot be reverted back, so the system cannot
383 be downgraded to the orignal incompatible version.
385 --------------------------------------------------------------------------------
387 07-30-2014 Intel Corporation
389 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
390 for currently supported client and server kernel versions.
391 * Server known to build on patched kernels:
392 2.6.32-431.20.3.el6 (RHEL6.5)
393 3.0.101-0.31 (SLES11 SP3)
394 * Client known to build on unpatched kernels:
395 2.6.32-431.20.3.el6 (RHEL6.5)
396 3.0.101-0.31 (SLES11 SP3)
399 * Recommended e2fsprogs version: 1.42.9.wc1 or newer
400 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
401 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
402 stack overflow. For more information, please refer to bugzilla 17630.
403 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
404 the re-exporting nodes to be mounted with "32bitapi" mount option
406 --------------------------------------------------------------------------------
408 10-31-2013 Intel Corporation
410 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
411 * Server builds for kernels:
412 2.6.32-358.18.1.el6 (RHEL6.4)
413 * Client builds for unpatched kernels:
414 2.6.32-358.18.1.el6 (RHEL6.4)
415 3.0.92-0.8 (SLES11 SP3)
417 * Recommended e2fsprogs version: 1.42.7.wc2 or newer
418 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
419 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
420 stack overflow. For more information, please refer to bugzilla 17630.
421 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
422 the re-exporting nodes to be mounted with "32bitapi" mount option
424 --------------------------------------------------------------------------------
426 05-30-2013 Intel Corporation
428 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
429 * Server builds for kernels:
430 2.6.32-358.6.2.el6 (RHEL6.4)
431 * Client builds for unpatched kernels:
432 2.6.32-358.6.2.el6 (RHEL6.4)
433 3.0.74-0.6.6 (SLES11 SP2)
435 * Recommended e2fsprogs version: 1.42.7.wc1 or newer
436 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
437 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
438 stack overflow. For more information, please refer to bugzilla 17630.
439 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
440 the re-exporting nodes to be mounted with "32bitapi" mount option
442 --------------------------------------------------------------------------------
444 9-30-2012 Whamcloud, Inc.
446 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
447 * Server support for kernels:
448 2.6.32-279.5.1.el6 (RHEL6)
449 * Client support for unpatched kernels:
450 2.6.32-279.5.1.el6 (RHEL6)
452 2.6.18-238.19.1.el5 (RHEL5)
453 2.6.16.60-0.69.1 (SLES10)
454 * Recommended e2fsprogs version: 1.41.90.wc4
455 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
456 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
457 stack overflow. For more information, please refer to bugzilla 17630.
458 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
459 the re-exporting nodes to be mounted with "32bitapi" mount option
461 Severity : enhancement
463 Description: Update RHEL6.3 kernel to 2.6.32-279.5.1.el6.
465 Severity : enhancement
467 Description: Update RHEL6.3 kernel to 2.6.32-279.2.1.el6.
469 Severity : enhancement
471 Description: Update RHEL6.2 kernel to 2.6.32-220.17.1.el6
473 Severity : enhancement
475 Description: Update RHEL6.2 kernel to 2.6.32-220.13.1.el6
477 Severity : enhancement
479 Description: Update RHEL6.2 kernel to 2.6.32-220.7.1.el6
481 --------------------------------------------------------------------------------
483 04-30-2012 Whamcloud, Inc.
485 * See https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
486 * Server support for kernels:
487 2.6.32-220.4.2.el6 (RHEL6)
488 * Client support for unpatched kernels:
489 2.6.32-220.4.2.el6 (RHEL6)
491 2.6.18-238.19.1.el5 (RHEL5)
492 2.6.16.60-0.69.1 (SLES10)
493 * Recommended e2fsprogs version: 1.41.90.wc4
494 * NFS export disabled when stack size < 8192 (32-bit Lustre clients),
495 since the NFSv4 export of Lustre filesystem with 4K stack may cause a
496 stack overflow. For more information, please refer to bugzilla 17630.
497 * NFSv4 reexport to 32-bit NFS client nodes requires Lustre client on
498 the re-exporting nodes to be mounted with "32bitapi" mount option
500 Severity : enhancement
502 Description: Update RHEL6.2 kernel to 2.6.32-220.4.2.el6
504 Severity : enhancement
506 Description: Update RHEL6.2 kernel to 2.6.32-220.4.1.el6
508 Severity : enhancement
510 Description: Update RHEL6.2 kernel to 2.6.32-220.el6
512 Severity : enhancement
514 Description: Update RHEL6.1 kernel to 2.6.32-131.17.1.el6
516 --------------------------------------------------------------------------------
518 09-30-2011 Whamcloud, Inc.
520 * Server support for kernels:
521 2.6.18-238.19.1.el5 (RHEL5)
522 2.6.32-131.6.1.el6 (RHEL6)
523 * Client support for unpatched kernels:
524 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
525 2.6.18-238.19.1.el5 (RHEL5)
526 2.6.32-131.6.1.el6 (RHEL6)
527 2.6.29.4-167.fc11 (FC11)
528 2.6.16.60-0.69.1 (SLES10)
530 * Recommended e2fsprogs version: 1.41.90.wc3
531 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
532 Lustre filesystem with 4K stack may cause a stack overflow. For more
533 information, please refer to bugzilla 17630.
534 * NFSv4 reexport requires Lustre client on the re-exporting nodes to be
535 mounted with "32bitapi" mount option
537 Severity : enhancement
538 Description: Update RHEL6.1 kernel to 2.6.32-131.6.1.el6
540 Severity : enhancement
541 Description: Update RHEL5.6 kernel to 2.6.18-238.19.1.el5
543 Severity : enhancement
544 Description: Update RHEL5.6 kernel to 2.6.18-238.9.1.el5
547 Frequency : only with SLES10
549 Description: Use OFED "KMP" provided by Novell
550 Details : SLES10 SP3 ships with OFED in a separate "KMP" package.
551 Lustre is now built against this package. That means you need to
552 install the ofed-kmp package from Novell for the patchless client.
553 Note that the ofed-kmp that Novell ships may not exactly match the
554 kernel version but should still be compatible.
556 Severity : enhancement
558 Description: Update RHEL5.5 kernel to 2.6.18-194.17.1.el5
559 Update OEL5.5 kernel to 2.6.18-194.17.1.0.1.el5.
561 --------------------------------------------------------------------------------
563 2010-07-15 Oracle, Inc.
565 * Server support for kernels:
566 2.6.18-164.11.1.el5 (RHEL 5)
567 2.6.18-164.11.1.0.1.el5 (OEL 5)
568 * Client support for unpatched kernels:
569 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
570 2.6.18-164.11.1.el5 (RHEL 5),
571 2.6.18-164.11.1.0.1.el5 (OEL 5)
572 2.6.16.60-0.42.8 (SLES 10),
574 2.6.29.4-167.fc11 (FC11)
575 * Recommended e2fsprogs version: 1.41.10-sun2
577 Severity : enhancement
579 Description: Distribute lfs_migrate script with lustre
580 Details : The lfs_migrate mentioned in the manual is now included in the
581 lustre tree in lustre/scripts and will distributed in lustre RPMs.
583 Severity : enhancement
585 Description: Failover nid list added to import proc file.
587 Severity : enhancement
589 Description: "som_preview" mount option required to use SOM
590 Details : The clients will need to use som_preview mount option
594 Description: Lustre does not do 1MB IOs to HW RAID
595 Details : Bump MAX_PHYS/HW_SEGMENTS and SG_ALL to 256 in the RHEL5 kernel.
597 Severity : enhancement
599 Description: lfs quota output is not very convenient for awk/sed-parsing
600 Details : Some positions in lfs quota output table could be empty or
601 non-empty which made it hard to parse it with scripts, now
602 a dash is put instead of space where there is not supposed
607 Description: Update to OFED 1.5.1
608 Details : OFED 1.5.1 will be used in environments that don't use
609 vendor supplied OFED.
613 Description: Remove lnet. prefix from lctl params display
614 Details : Also renamed lnet.memused to lnet_memused to
619 Description: Use conf_param -d to remove permanant settings.
623 Description: Add list_param -R
624 Details : lctl list_param now supports recursive mode.
627 Bugzilla : 21829, 22312
628 Description: Fix llobdstat.
629 Details : 1. Now updating the write_bytes counter
630 2. Port script fixes from 1.8.
634 Description: decrease the usage of memory on clients.
635 Details : 1. On clients, recycle dentries and inodes unused.
636 2. Delete the code related to ll_deathrow(att 6215 in bug 1443). It
641 Description: llobdstat fix and enhancement
642 Details : add a counter to set a limit to how many samples will be returned
643 fix a wildcard in the path to limit to obdfilter stats only
647 Description: lfs check only allowed for root
648 Details : Code cleanups and added a regresion test of non-root use of
653 Description: MDS Oopses in ptlrpc_dispatch_difficult_reply()/__list_add()
654 Details : Fix hr->hr_index overload.
658 Description: Fix FIEMAP functionality.
662 Description: FC11 patchless client support added
666 Description: Update RHEL5.4 kernel to 2.6.18-164.11.1.el5 and
667 OEL5.4 kernel to 2.6.18-164.11.1.0.1.el5.
671 Description: New acc-sm script and various test-framework enhancments.
675 Description: Previously landed patch was reverted.
679 Description: Update SLES10 kernel to 2.6.16.60-0.42.8 (patchless)
683 Description: Add ldiskfs maxdirsize mount option.
687 Description: Fix mkfs.lustre support for 16TB LUNs.
691 Description: Print mmp_check_interval
692 Details : print mmp_check_interval and make it possible to abort
693 mount operation in case it takes too long.
697 Description: Add cascading_rw to lustre mpi tests.
701 Description: Store SOM data in EA.
702 Details : Also several SOM related cleanups.
706 Description: Add refcount to md_open_data
710 Description: Use refcount on lov_request_set to prevent memory corruption.
714 Description: Statahead: use parent dir's i_mutex for synchronization.
718 Description: loadgen improvements.
722 Description: Enable peer health detection in ksocklnd and ko2iblnd.
726 Description: Control DCACHE_LUSTRE_INVALID flag with MDS_INODE_LOOKCUP lock
730 Description: Fix race on obd_devcie:md_stats
734 Description: Fix SLES 11 client builds.
735 Details : Resolved uninitalized pointer gcc warnings.
739 Description: Allow non-root access for "lfs check".
740 Details : Added a check in obd_class_ioctl() for OBD_IOC_PING_TARGET.
744 Description: Clear imp_force_reconnect correctly in ptlrpc_connect_interpret()
746 Severity : enhancement
748 Description: Update kernel to RHEL5.4 2.6.18-164.6.1.el5 and
749 OEL5 2.6.18-164.6.1.0.1.el5(Both in-kernel OFED enabled).
753 Description: Sanity test_27p fails.
754 Details : OBD_FAIL_OST_ENOSPC has been moved to filter_statfs().
755 Sanity tests 27 were cleaned up and improved.
759 Description: Unify req format on client/servers
760 Details : Use new req_capsule API [almost] everywhere instead of old PTLRPC
761 buffers and swabbers approach..
766 Description: drop join file stuff.
768 Severity : enhancement
770 Description: Add 2.6.27 and SLES11 (patchless client) support
774 Description: Fix in ptlrpc_expire_one_request() to print the signed time
777 Severity : enhancement
779 Description: Build kernels (RHEL5, OEL5 and SLES10/11) using the vendor's own
782 Severity : enhancement
784 Description: Remove set_info(KEY_UNLINKED) from MDS/OSC
786 Severity : enhancement
788 Description: correctly handle big reply message.
789 Details : send LNet event if reply is bigger than buffer and adjust this buffer
794 Description: Drop unnecessary __GFP_NOMEMALLOC flag from filter_get_page()
796 Severity : enhancement
798 Description: Add support for OEL5.
802 Description: Conf-sanity.sh 50g test - deactivated OST should not cause a panic.
804 Severity : enhancement
806 Description: "weak-modules" support
807 Details : Implement "weak-modules" support which enables kernel modules
808 to be used with any kernel that implements the same kABI. In
809 order to achieve this modules are now installed in
810 /lib/modules/$(uname -r)/updates/kernel on all distributions.
814 Description: client could not reconnect to OST because of an active request.
815 Details : The client now retries to reconnect to the same server, if a
816 connect request failed with EBUSY or -EAGAIN. Test case 35b
817 conf-sanity.sh has been created.
821 Description: Don't allow make backward step on assignin osc next id.
822 Details : race between allocation next id and ll_sync thread can be cause
823 of set wrong osc next id and can be kill valid ost objects.
827 Description: Changes in raid5-large-io-rhel5.patch to calculate sectors properly
831 Description: Increase the default BLK_DEF_MAX_SECTORS value for RHEL5 and SLES11
835 Description: Error handling in osc_statfs_interpret() has been improved.
836 Details : Check in osc_statfs_interpret() for EBADR.
840 Description: Do not send statfs() requests to OSTs disabled by administrator.
841 Details : Check in lov_prep_statfs_set() for non-NULL ltd_exp.
843 Severity : enhancement
845 Description: Added sanity-quota test to for lfs(1) to work with numeric uid/gid.
847 Severity : enhancement
849 Description: Update kernel to RHEL5 2.6.18-128.7.1.el5.
851 Severity : enhancement
853 Description: Enhancement for lfs(1) command to use numeric uid/gid.
857 Description: Improved cleanup procedure for conf-sanity test_46a.
861 Description: Do not start bulk transfer, if client tried to reconnect.
862 Details : Do not start bulk transfer, if client tried to reconnect,
863 but failed due to active rpcs. Check for exp_abort_active_req
864 has been implemented.
868 Description: Prevent inconsistences between linux and lustre mount structures.
869 Details : Wait indefinitely in server_wait_finished() until mnt_count drops.
870 Make the sleep interruptible.
872 Severity : enhancement
874 Description: provide server to client comms path
875 Details : server to client communications path via new LDLM_SET_INFO rpc
879 Description: truncate starts GFP_FS allocation under transaction causing deadlock
880 Details : ldiskfs_truncate calls grab_cache_page which may start page
881 allocation under an open transaction. This may lead to
882 calling prune_icache with consequent lustre reentrance.
886 Description: send correctly lsm on open replay
887 Details : MDS is trust to LSM size on replay open, but client can set wrong size
892 Description: enable client interface failover
893 Details : When a child reconnects from another NID, properly update export
894 nid hash position and ldlm reverse import.
896 Severity : enhancement
898 Description: Communicate OST degraded/readonly state via statfs to MDS
899 Details : Flags in the statfs returned from OSTs indicate whether the
900 OST is in a degraded RAID state, or if the filesystem has
901 turned read-only after a filesystem error is detected.
906 Description: don't panic if EPROTO was hit when reading symlink
907 Details : correctly handling request reference in error cases.
909 Severity : enhancement
911 Description: Add LustreNetLink, a kernel-userspace communcation path.
913 Severity : enhancement
915 Description: Update kernel to SLES10 SP2 2.6.16.60-0.39.3.
920 Description: access to llog context before init.
921 Details : move handling CATALOGS file at osc layer and forbid access to llog
925 Frequency : rare, only when client is writing to OST on local node
927 Description: Avoid deadlock for local client writes
928 Details : Use new OBD_BRW_MEMALLOC flag to notify OST about writes in the
929 memory freeing context. This allows OST threads to set the
930 PF_MEMALLOC flag on task structures in order to allocate memory
931 from reserved pools and complete IO.
932 Use GFP_HIGHUSER for OST allocations for non-local client writes,
933 so that the OST threads generate memory pressure and allow
934 inactive pages to be reclaimed.
936 Severity : enhancement
938 Description: Update kernel to RHEL5.3 2.6.18-128.1.14.el5.
943 Description: lock ordering violation between &cli->cl_sem and _lprocfs_lock
944 Details : move ldlm namespace creation in setup phase to avoid grab
945 _lprocfs_lock with cli_sem held.
949 Description: Temporarily disable grant shrink.
950 Details : Disable the feature for debugging.
953 Frequency : only when formatting test filesystems
955 Description: Unable to run several mkfs.lustre on loop devices at the same time
956 Details : mkfs.lustre returns error 256 on the concurrent loop devices
957 formatting. The solution is to proper handle the error.
959 Severity : enhancement
961 Description: Update kernel to RHEL5.3 2.6.18-128.1.6.el5.
963 Severity : enhancement
965 Description: Update kernel to SLES10 SP2 2.6.16.60-0.37.
970 Description: resolve race between obd_disconnect and class_disconnect_exports
971 Details : if obd_disconnect will be called to already disconnected export he
972 forget release one reference and osc module can't unloaded.
974 Severity : enhancement
976 Description: Allow tuning service thread via /proc
977 Details : For each service a new
978 /proc/fs/lustre/{service}/*/thread_{min,max,started} entry is
979 created that can be used to set min/max thread counts, and get the
980 current number of running threads.
985 Descriptoin: don't return error if have partially created objects for file.
986 Details : lov_update_create_set uses set->set_success as index for created
987 objects, so if some requests will be failed, they will have hole
988 at end of array and qos_shrink_lsm for allocate correct lsm.
990 Severity : enhancement
992 Description: Update OFED support to 1.4.1
997 Description: Reduce small size read RPC
998 Details : Set read-ahead limit for every file and only do read-ahead when
999 available read-ahead pages are bigger than 1M to avoid small size
1002 Severity : enhancement
1004 Description: move AT tunable parameters for more consistent usage
1005 Details : add AT tunables under /proc/sys/lustre, add to conf_param parsing
1007 Severity : enhancement
1009 Description: add lazystatfs mount option to allow statfs(2) to skip down OSTs
1010 Details : allow skip disconnected ost for send statfs request and hide error
1016 Description: fix racy locking of mballoc block bitmaps causing BUG
1017 Details : The locking of the mballoc buddy bitmap and the in-memory
1018 block bitmap was using two different spin locks in some
1019 cases. This made it possible to incorrectly access the
1020 mballoc bitmap while another process was modifying it,
1021 causing a sanity assertion to fail. While no on-disk corruption
1022 was reported, there was some risk of this happening.
1025 Frequency : rare, on llog test 6
1027 Descriptoin: don't allow connect to already connected import
1028 Details : allowing connect to already connected import is hide connecting problem.
1033 Descriptoin: allow kill process which wait statahead result
1034 Details : for some reasons 'ls' can stick in waiting result from statahead,
1035 in this case need way for kill this process.
1037 Severity : enhancement
1039 Description: Add state history info file, enhance import info file
1040 Details : Track import connection state changes in a new osc/mdc proc file;
1041 add overview-type data to the osc/mdc import proc file.
1043 Severity : enhancement
1045 Description: MDS create should not wait for statfs RPC while holding DLM lock.
1047 Severity : enhancement
1049 Description: Update to RHEL5U3 kernel-2.6.18-128.1.1.el5.
1054 Descriptoin: OST grant too much space to client even there are not enough space.
1055 Details : Client will shrink its grant cache to OST if there are no write
1056 activity over 6 mins (GRANT_SHRINK_INTERVAL), and OST will retrieve
1057 this grant cache if there are already not enough avaible space
1058 (left_space < total_clients * 32M).
1061 Frequency : start MDS on uncleanly shutdowned MDS device
1063 Descriptoin: ll_sync thread stay in waiting mds<>ost recovery finished
1064 Details : stay in waiting mds<>ost recovery finished produce random bugs
1065 due race between two ll_sync thread for one lov target. send
1066 ACTIVATE event only if connect realy finished and import have
1070 Frequency : rare, connect and disconnect target at same time
1072 Descriptoin: ASSERTION(atomic_read(&imp->imp_inflight) == 0
1073 Details : don't call obd_disconnect under lov_lock. this long time
1074 operation and can block ptlrpcd which answer to connect request.
1079 Descriptoin: don't lose wakeup for imp_recovery_waitq
1080 Details : recover_import_no_retry or invalidate_import and import_close can
1081 both sleep on imp_recovery_waitq, but we was send only one wakeup
1085 Frequency : always with long access acl
1087 Descriptoin: mds can't pack reply with long acl.
1088 Details : mds don't control size of acl but they limited by reint/getattr
1091 Severity : enhancement
1093 Description: Update to SLES10 kernel-2.6.16.60-0.33.
1095 Severity : enhancement
1097 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
1100 Frequency : start MDS on uncleanly shutdowned MDS device
1102 Descriptoin: aborting recovery hang on MDS
1103 Details : don't throttle destroy RPCs for the MDT.
1106 Frequency : on remount
1108 Description: external journal device not working after the remount
1109 Details : clear dev_rdonly flag for external journal devices in
1115 Description: shutdown vs evict race
1116 Details : client_disconnect_export vs connect request race.
1117 if client will evicted at this time - we start invalidate
1118 thread without referece to import and import can be freed
1122 Frequency : rare, need acl's on inode.
1124 Description: client can't handle ost additional correctly
1125 Details : if ost was added after client connected to mds client can have
1126 hit lnet_try_match_md ... to big messages to wide striped files.
1127 in this case need teach client to handle config events about add
1128 lov target and update client max ea size at that event.
1130 Severity : enhancement
1132 Description: Changelogs
1133 Details : Changelogs are a lightweight mechanism to track filesystem
1134 metadata and namespace changes. The changelog is recorded
1135 permanently on the MDTs, and is periodically "consumed" / purged
1136 when records are no longer needed.
1138 Severity : enhancement
1140 Description: compact fld format with extents
1141 Details : Store range of seq rather than every seq in FLD. Seq
1142 controller update FLD rather than clients. In Case of CMD, mdt0
1143 has FLD, all other metadata server act as non persistent proxy
1144 for FLD queries and cache fld entries in fld cache.
1149 Description: don't skip ost target if they assigned to file
1150 Details : Drop slow OSCs if we can, but not for requested start idx.
1151 This means "if OSC is slow and it is not the requested
1152 start OST, then it can be skipped, otherwise skip it only
1153 if it is inactive/recovering/out-of-space.
1157 Description: more cleanup in mds_lov
1158 Details : not send LOV EA under replay, we can't know about they size at this
1159 time. Don't allow client connect to mds before any ost connected,
1160 for avoid problems with LOV EA size and returning EIO to client.
1162 Severity : enhancement
1164 Description: Interoperability at server side (Disk interoperability)
1166 Severity : enhancement
1168 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
1170 Severity : enhancement
1172 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
1174 Severity : enhancement
1176 Description: New client IO stack (CLIO).
1178 Severity : enhancement
1180 Description: Commit on sharing. Eliminate inter-client dependencies between
1181 uncommitted transactions by doing transaction commits.
1182 Thereby clients may recovery independently.
1185 Frequency : Create a symlink file with a very long name
1187 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
1188 Details : If there is no extra space in the request for early cancels,
1189 ldlm_req_handles_avail() returns 0 instead of a negative value.
1191 Severity : enhancement
1193 Description: Add /proc entry for import status
1194 Details : The mdc, osc, and mgc import directories now have
1195 an import directory that contains useful import data for debugging
1196 connection problems.
1198 Severity : enhancement
1200 Description: Re-disable certain /proc logging
1201 Details : Enable and disable client's offset_stats, extents_stats and
1202 extents_stats_per_process stats logging on the fly.
1205 Frequency : Only on FC kernels 2.6.22+
1207 Description: oops in statahead
1208 Details : Do not drop reference count for the dentry from VFS when lookup,
1209 VFS will do that by itself.
1211 Severity : enhancement
1213 Description: Generic /proc file permissions
1214 Details : Set /Proc file permissions in a more generic way to enable non-
1215 root users operate on some /proc files.
1219 Description: Hitting mdc_commit_close() ASSERTION
1220 Details : Properly handle request reference release in
1221 ll_release_openhandle().
1225 Description: quota recovery deadlock during mds failover
1226 Details : This patch includes att18982, att18236, att18237 in bz14840.
1228 1. fix osts hang when mds does failover with quotaon
1229 2. prevent watchdog storm when osts threads wait for the
1234 Frequency : only patchless client
1235 Description: add workaround for race between add/remove dentry from hash
1237 Severity : enhancement
1239 Description: Allow OST glimpses to return PW locks
1243 Description: LBUG when llog conf file is full
1244 Details : When llog bitmap is full, ENOSPC should be returned for plain log.
1248 Description: Prevent import from entering FULL state when server in recovery
1252 Description: service mount cannot take device name with ":"
1253 Details : Only when device name contains ":/" will mount treat it as
1259 Description: replace ptlrpcd with the statahead thread to interpret the async
1260 statahead RPC callback
1264 Frequency : on recovery
1265 Description: I/O failures after umount during fail back
1266 Details : if client reconnected to restarted server we need join to recovery
1267 instead of find server handler is changed and process self
1268 eviction with cancel all locks.
1270 Severity : enhancement
1272 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
1274 Severity : enhancement
1276 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
1278 Severity : enhancement
1280 Description: Upcall on Lustre log has been dumped
1281 Details : Allow for a user mode script to be called once a Lustre log has
1282 been dumped. It passes the filename of the dumped log to the
1283 script, the location of the script can be specified via
1284 /proc/sys/lnet/debug_log_upcall.
1289 Description: avoid idr_remove called for id which is not allocated.
1290 Details : Move assigment s_dev for clustered nfs to end of initialization,
1291 to avoid problem with error handling.
1296 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
1297 Details : When connection is reused this not moved from CONN_UNUSED_HASH
1298 into CONN_USED_HASH and this prodice warning when put connection
1299 again in unused hash.
1302 Severity : enhancement
1304 Description: File striping can now be set to use an arbitrary pool of OSTs.
1306 Severity : enhancement
1308 Description: Export bytes_read/bytes_write count on OSC/OST.
1312 Description: Early reply size mismatch, MGC loses connection
1313 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
1314 the connect flags are properly negotiated.
1319 Description: quotas are not honored with O_DIRECT
1320 Details : all writes with the flag O_DIRECT will use grants which leads to
1321 this problem. Now using OBD_BRW_SYNC to guard this.
1325 Description: add quota statistics
1326 Details : 1. sort out quota proc entries and proc code.
1327 2. add quota statistics
1329 Severity : enhancement
1331 Description: enable quota support for HEAD.
1335 Description: Properly propagate oinfo flags from lov to osc for statfs
1336 Details : restore missing copy oi_flags to lov requests.
1338 Severity : enhancement
1340 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
1341 : l_getgroups(8), lst(8), routerstat(8)
1342 Details : included man pages for llobdstat(8), llstat(8),
1343 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
1345 Severity : enhancement
1347 Description: configure's --enable-quota should check the
1348 : kernel .config for CONFIG_QUOTA
1349 Details : configure is terminated if --enable-quota is passed but
1350 : no quota support is in kernel
1354 Description: Remove portals compatibility
1355 Details : Remove portals compatibility, not interoperable with releases
1360 Description: Resolve device initialization race
1361 Details : Prevent proc handler from accessing devices added to the
1362 obd_devs array but yet be intialized.
1364 Severity : enhancement
1366 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
1368 Severity : enhancement
1370 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
1375 Description: Using wrong pointer in osc_brw_prep_request
1376 Details : Access to array[-1] can produce panic if kernel compiled with
1377 CONFIG_PAGE_ALLOC enabled
1381 Description: Client runs out of low memory
1382 Details : Consider only lowmem when counting initial number of llap pages
1386 Description: *optional* service tags registration
1387 Details : if the "service tags" package is installed on a Lustre node
1388 When the filesystem is mounted, a local-node service tag will
1389 be created. See http://inventory.sun.com/ for more information
1390 about the Service Tags asset management system.
1394 Description: Kernel BUG tries to release flock
1395 Details : Lustre does not destroy flock lock before last reference goes
1396 away. So always drop flock locks when client is evicted and
1397 perform unlock regardless of successfulness of speaking to MDS.
1401 Description: add refcount protection for osc callbacks, avoid panic on shutdown
1405 Description: sanity test 65a fails if stripecount of -1 is set
1406 Details : handle -1 striping on filesystem in ll_dirstripe_verify
1411 Description: ASSERTION(CheckWriteback(page,cmd)) failed
1412 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
1415 Severity : enhancement
1417 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
1421 Description: do not process already freed flock
1422 Details : flock can possibly be freed by another thread before it reaches
1423 to ldlm_flock_completion_ast.
1427 Description: LBUG during stress test
1428 Details : Need properly lock accesses the flock deadlock detection list.
1432 Description: oops in page fault handler
1433 Details : kernel page fault handler can return two special 'pages' in error
1434 case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
1438 Description: timeout with invalidate import.
1439 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
1440 should be handled by ptlrpcd. This produce long age waiting and
1441 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
1443 Severity : enhancement
1445 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
1449 Description: enable MGS and MDT services start separately
1450 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
1451 a co-located MGS without starting the MGS, which is a complement
1452 to 'nosvc' mount option.
1456 Frequency : after recovery
1457 Description: precreate to many object's after del orphan.
1458 Details : del orphan st in oscc last_id == next_id and this triger growing
1459 count of precreated objects. Set flag LOW to skip increase count
1460 of precreated objects.
1464 Frequency : rare, on clear nid stats
1465 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
1466 Details : when clean nid stats sometimes try destroy live entry,
1467 and this produce panic in free.
1471 Description: Stack overflow during MDS log replay
1472 ease stack pressure by using a thread dealing llog_process.
1476 Description: wait until IO finished before start new when do lock cancel.
1477 Details : VM protocol want old IO finished before start new, in this case
1478 need wait until PG_writeback is cleared until check dirty flag and
1479 call writepages in lock cancel callback.
1481 Severity : enhancement
1483 Description: using special macro for print time and cleanup in includes.
1487 Description: mds_mfd_close() ASSERTION(rc == 0)
1488 Details : In mds_mfd_close(), we need protect inode's writecount change
1489 within its orphan write semaphore to prevent possible races.
1493 Description: Obsolete CURRENT_SECONDS and use cfs_time_current_sec() instead.
1497 Frequency : rare, on shutdown ost
1498 Description: don't hit live lock with umount ost.
1499 Details : shrink_dcache_parent can be in long loop with destroy dentries,
1500 use shrink_dcache_sb instead.
1504 Description: don't panic with use echo client
1505 Details : echo client pass NULL as client nid pointer and this produce null
1506 pointer dereference.
1510 Description: fix build on ppc32
1511 Details : compile code with -m64 flag produce wrong object file for ppc32.
1515 Description: add message levels for liblustreapi
1519 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
1520 Details : Move the 'good_osts' check before the 'total_bavail' check. This
1521 will result in an -EAGAIN and in the exit call path we call
1522 alloc_rr() which will with increasing aggressiveness attempt to
1523 aquire precreated objects on the minimum number of required OSCs.
1527 Description: Use old size assignment to avoid deadlock
1528 Details : Reverts the changes in bugs 2369 and bug 14138 that introduced
1529 the scheduling while holding a spinlock. We do not need locking
1530 for size in ll_update_inode() because size is only updated from
1531 the MDS for directories or files without objects, so there is no
1532 other place to do the update, and concurrent access to such inodes
1533 are protected by the inode lock.
1537 Description: resolve "_IOWR redefined" build error on SLES10
1541 Description: dump the memory debugging after all modules are unloaded to
1542 suppress false negative in conf_sanity test 39
1544 Severity : enhancement
1546 Description: build kernel-ib packages for OFED 1.3 in our release cycle
1551 Description: fix SLES kernel versioning
1552 Details : the kernel version for our SLES 10 kernel did not include a "-"
1553 before the "smp" at the end. while this was not a problem in
1554 general, it did mean that software trying to use the kernel
1555 version to try to detect a vendor specific kernel would fail.
1556 this was most evident by the OFED build scripts.
1560 Description: Don't update lov_desc members until making sure they are valid
1561 Details : When updating lov_desc members via proc fs, need fix their
1562 validities before doing the real update.
1566 Description: don't put request into delay list while invalidate in flight.
1567 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
1568 import in flight. this produce timeout for invalidate and sometimes
1569 can cause stale data.
1573 Frequency : on ppc only
1574 Description: not convert ost objects for directory because it's not exist.
1575 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1579 Description: Add FMODE_EXEC file flag for SLES10 SP1 kernel.
1581 Severity : enhancement
1583 Description: Update to support 2.6.22.14 vanilla kernel.
1587 Frequency : rare, on recovery
1588 Description: read procfs can produce deadlock in some situation
1589 Details : Holding lprocfs lock which send rpc can produce block for destroy
1590 obd objects and this also block reconnect with -EALREADY. This
1591 isn't fix all lprocfs bugs - but make it rare.
1593 Severity : enhancement
1595 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1598 Frequency : frequent on X2 node
1600 Description: mdc_set_open_replay_data LBUG
1601 Details : Set replay data for requests that are eligible for replay.
1605 Description: lustre_mgs: operation 101 on unconnected MGS
1606 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1607 MGC, and late on MGC cannot successfully connect to MGS and a lot
1608 of the error messages complaining that MGS is not connected.
1612 Frequency : on network error
1613 Description: panic with double free request if network error
1614 Details : mdc_finish_enqueue is finish request if any network error ocuring,
1615 but it's true only for synchronus enqueue, for async enqueue
1616 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1619 Severity : enhancement
1621 Description: client-side metadata stat-ahead during readdir(directory readahead)
1622 Details : perform client-side metadata stat-ahead when the client detects
1623 readdir and sequential stat of dir entries therein
1626 Frequency : on start mds
1628 Description: Implement get_info(last_id) in obdfilter.
1631 Frequency : occasional
1633 Description: Correctly check stale fid, not start epoch if ost not support SOM
1634 Details : open with flag O_CREATE need set old fid in op_fid3 because
1635 op_fid2 was overwritten with new generated fid, but mds can answer
1636 with one of these two fids and both is not stale. Setattr
1637 incorrectly started an epoch and assume will be called
1638 done_writing, but without SOM done_writing ever being called.
1641 Frequency : rare, depends on device drivers and load
1643 Description: MDS or OSS nodes crash due to stack overflow
1644 Details : Code changes in 1.8.0 increased the stack usage of some functions.
1645 In some cases, in conjunction with device drivers that use a lot
1646 of stack the MDS (or possibly OSS) service threads could overflow
1647 the stack. One change which was identified to consume additional
1648 stack has been reworked to avoid the extra stack usage.
1651 Frequency : occasional
1653 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1654 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1655 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1656 oscc_has_objects and see OSC has no objests and call
1657 oscc_internal_create to resend create request.
1659 Severity : enhancement
1661 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1663 Severity : enhancement
1665 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1668 Frequency : very rare
1670 Description: Fix replay if there is an un-replied request and open
1671 Details : In some cases, older replay request will revert the
1672 mcd->mcd_last_xid on MDS which is used to record the client's
1673 latest sent request.
1675 Severity : enhancement
1677 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1679 Severity : enhancement
1681 Description: Add rhel5 support to HEAD.
1683 Serverity : enhancement
1685 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1690 Description: Don't allow skipping OSTs if index has been specified.
1691 Details : Don't allow skipping OSTs if index has been specified, make
1692 locking in internal create lots better.
1696 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1697 Details : access to bitfield in structure is always rounded to long
1698 and this produce problem with not atomic change any bit.
1702 Description: Lustre make rpms failed.
1703 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1704 builds Lustre rpms from tarball.
1707 Frequency : rare on shutdown ost
1709 Description: If llog cancel was not send before clean_exports phase, this can
1710 produce deadlock in llog code.
1711 Details : If llog thread has last reference to obd and call class_import_put
1712 this produce deadlock because llog_cleanup_commit_master wait when
1713 last llog_commit_thread exited, but this never success because was
1714 called from llog_commit_thread.
1718 Description: allow userland application know is lost one of stripes.
1719 Details : fill lvb_blocks with error code on ost and return it to
1720 application if error flag found.
1724 Description: NULL lov_tgts causing MDS oops
1725 Details : more safe checks for NULL lov_tgts for avoid oops.
1727 Severity : enhancement
1729 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1733 Descriptoin: make lov_create() will not stuck in obd_statfs_rqset()
1734 Details : If an OST is down the MDS will hang indefinitely in
1735 obd_statfs_rqset() waiting for the statfs data. While for
1736 MDS QOS usage of statfs, it should not stuck in waiting.
1738 Severity : enhancement
1740 Description: remote_acl support
1741 Details : Support ACL-based permission check for remote user.
1742 Support setfacl/getfacl for remote user with the utils
1743 "lfs {l,r}{s,g}etfacl" which follow the same parameter format as
1744 the system "{s,g}etfacl" utils.
1746 Severity : enhancement
1748 Description: Update to RHEL4 U6 kernel-2.6.9-67.EL.
1750 Severity : enhancement
1752 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1756 Description: make mgs_setparam() handle fsname containing dash
1757 Details : fsname containing a dash does not work with lctl conf_param
1759 Severity : enhancement
1761 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1763 Severity : enhancement
1765 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1769 Frequency : rare, at shutdown
1770 Description: access already free / zero obd_namespace.
1771 Details : if client_disconnect_export was called without force flag set,
1772 and exist connect request in flight, this can produce access to
1773 NULL pointer (or already free pointer) when connect_interpret
1774 store ocd flags in obd_namespace.
1778 Frequency : only at startup
1779 Description: not alloc memory with spinlock held.
1780 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1781 if any spinlock held.
1783 Severity : enhancement
1785 Description: make lustre randomly fail allocating memory
1786 Details : Make lustre randomly failed allocating memory for testing purpose.
1788 Severity : enhancement
1790 Description: lost problems with lov objid file
1791 Details : Fixes some scability and access to not inited memory problems
1792 in work with lov objdid file.
1797 Description: lfs find does not continue on file error
1798 Details : Continue other files processing when a file/dir is absent.
1802 Description: Inconsistent usage of lustre_pack_reply()
1803 Details : Standardize the usage of lustre_pack_reply() such that it
1804 always generate a CERROR on failure.
1807 Frequency : occasional
1809 Description: MDS hang or stay in waiting lock
1810 Details : If client receive lock with CBPENDING flag ldlm need send lock
1811 cancel as separate rpc, to avoid situation when cancel request
1812 can't processed due all i/o threads stay in wait lock.
1816 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1820 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1824 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1828 Description: Update to RHEL4 latest kernel.
1830 Severity : enhancement
1832 Description: Build SLES10 patchless client fails
1833 Details : The configure was broken by run ./configure with
1834 --with-linux-obj=.... argument for patchless client. When the
1835 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1836 can't search header adequately. Use absolute path such as
1837 -I($LINUX)/include instead.
1841 Description: interrupt oig_wait produce painc on resend.
1842 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1843 private set, and this produce situation when rq_ptlrpcd_data not
1844 copyed to new allocated request and triggered LBUG on assert
1845 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1846 ptlrpcd set if request is changed and can be safety changed to use
1851 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1852 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1853 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1854 these macros and replace them with their difinitions instead.
1858 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1859 Details : Update which_patch & target file for SLES10 latest kernel.
1862 Description: Remove client patches from SLES 10 kernel.
1863 Details : This causes SLES 10 clients to behave as patchless clients
1864 even on a Lustre-patched (server) kernel.
1866 Severity : enhancement
1868 Description: self-adjustable client's lru lists
1869 Details : use adaptive algorithm for managing client cached locks lru
1870 lists according to current server load, other client's work
1871 pattern, memory activities, etc. Both, server and client
1872 side namespaces provide number of proc tunables for controlling
1875 Severity : enhancement
1877 Description: light-weight GSS support
1878 Details : Support krb5n and krb5a mode, which keep Kerberos 5 authentication
1879 and reduce performance overhead.
1881 Severity : enhancement
1883 Description: Linux keyring support
1884 Details : Support using service of Linux keyring for Lustre GSS internal
1885 context refresh/cache mechanism.
1889 Description: Fix errors in lfs documentation
1890 Details : Fixes man pages
1894 Description: don't use GFP_* in generic Lustre code.
1895 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
1899 Description: obdclass is limited by single OBD_ALLOC(idarray)
1900 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
1904 Description: warnings with build patchless client with vanila 2.6.19 and up
1905 Details : change old ctl_table style and replace ctl_table/ctl_table_header
1906 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
1910 Frequency : Only for SLES
1911 Description: sanity_quota fail test_1
1912 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
1913 file, which makes TSTID[2] inunique.
1917 Frequency : Only for SLES9
1918 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
1919 Details : Update target/ChangeLog/which_patch .
1923 Frequency : Only for test_10 in sanity-quota.sh
1924 Description: a bug in quota test code
1925 Details : checking a test flag in a wrong place
1929 Frequency : Only for test_13 in sanity-quota.sh
1930 Description: a bug in quota test script
1931 Details : shouldn't check overfull allocated quota
1935 Description: jbd statistics
1936 Details : Port older jbd statistics patch for sles10
1938 Severity : enhancement
1940 Description: Add printing inode info into message about error in writepage.
1945 Description: reply_lock_interpret crash due to race with it and lock cancel.
1946 Details : Do not replay locks that are being cancelled. Do not reference
1947 locks by their address during replay, just by their handle.
1951 Frequency : When flocks are used.
1952 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
1953 Details : Flock locks might destroy just granted lock if it could be merged
1954 with another existing flock, this is done in completion handler,
1955 so teach ldlm_cli_enquque_fini that this is a valid case for
1961 Description: Oops in read and write path when failing to allocate lock.
1962 Details : Check if lock allocation failed and return error back.
1966 Description: lstripe command fails for valid OST index
1967 Details : The stripe offset is compared to lov->desc.ld_tgt_count
1968 instead of lov->desc.ld_active_tgt_count.
1972 Description: sanity.sh failed test 103
1973 Details : RHEL mis-interpret setfacl "-X" param, so we won't test setfacl
1978 Description: df doesn't work properly if diskfs blocksize != 4K
1979 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1983 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1984 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1986 Severity : enhancement
1988 Description: Adaptive timeouts
1989 Details : RPC timeouts adapt to changing server load and network
1990 conditions to reduce resend attempts and improve recovery time.
1994 Description: llapi_file_create() does not allow some changes
1995 Details : add llapi_file_open() that allows specifying the mode and
1996 open flags, and also returns an open file handle.
2000 Description: merge and cleanup kernel patches.
2001 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
2005 Description: Add journal checksum support.(Kernel part)
2006 Details : The journal checksum feature adds two new flags i.e
2007 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT and
2008 JBD2_FEATURE_COMPAT_CHECKSUM. JBD2_FEATURE_CHECKSUM flag
2009 indicates that the commit block contains the checksum for
2010 the blocks described by the descriptor blocks. Now commit
2011 record can be sent to disk without waiting for descriptor
2012 blocks to be written to disk. This behavior is controlled
2013 using JBD2_FEATURE_ASYNC_COMMIT flag.
2017 Description: OSS needs mutliple precreate threads
2018 Details : Add ability to start more than one create thread per OSS.
2022 Description: Sanity.sh test_65k failed: setstripe should have succeeded
2023 Details : execute lfs setstripe on client
2027 Description: mds_obd_create error creating tmp object
2028 Details : When the user sets quota on root, llog will be affected and can't
2029 create files and write files.
2032 Frequency : Always on ia64 patchless client, and possibly others.
2034 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
2035 Details : This allows the patchless client to be loaded on architectures
2036 without this export.
2040 Description: RedHat Update kernel for RHEL5
2041 Details : Add the kernel config file for RHEL5.
2045 Description: RedHat Update kernel for RHEL5
2046 Details : Modify the kernel config file more closer RHEL5.
2050 Description: Build failure against Centos5 (RHEL5)
2051 Details : Define PAGE_SIZE when it isn't present.
2055 Description: test_7 on sanity-quota.sh can't be used on separate machines
2056 Details : add facet to handle it
2060 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
2061 Details : with NFS, the anon dentry's parent was set to itself in
2062 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
2063 corresponding dentry other than use rec->ur_name.
2065 Severity : enhancement
2067 Description: lfs setstripe enhancement
2068 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2072 Description: enable data checksumming by default
2073 Details : enable checksum by default, allow --disable-checksum
2074 configure option and "-o nochecksum" mount option. Checksums
2075 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
2079 Description: lustre support for RHEL5
2080 Details : Add support for RHEL5.
2084 Description: Client eviction due to failover config
2085 Details : after a connection loss, the lustre client should attempt to
2086 reconnect to the last active server first before trying the
2087 other potential connections.
2091 Description: when mds and osts use different quota unit(32bit and 64bit),
2092 quota will be released repeatly.
2093 Details : void sending multiple quota reqs to mds, which will keep the
2094 status between the reqs.
2098 Description: osts not allocated evenly to files
2099 Details : change the condition to increase offset_idx
2102 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
2104 Description: Data corruption for OSTs that are formatted larger than 2TB
2106 Details : When generating the bio request for lustre file writes the
2107 sector number would overflow a temporary variable before being
2108 used for the IO. The data reads correctly from Lustre (which
2109 will overflow in a similar manner) but other file data or
2110 filesystem metadata may be corrupted in some cases.
2114 Description: Tune the kernel for good SCSI performance.
2115 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
2116 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
2121 Description: rewrite ext2-derived code in obdclass/uuid.c
2122 Details : rewrite inherited code (uuid parsing code from ext2 utils)
2123 from scratch preserving functionality.
2127 Description: "lfs find -obd UUID" prints directories
2128 Details : "lfs find -obd UUID" will return all directory names instead
2129 of just file names. It is incorrect because the directories
2130 do not reside on the OSTs.
2134 Description: use i_size_read and i_size_write in 2.6 port
2135 Details : replace inode->i_size access with i_size_read/write()
2138 Frequency : when removing large files
2140 Description: scheduling issue during removal of large Lustre files
2141 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
2142 It causes scheduling issues when removing large files (17TB in the
2146 Frequency : only with liblustre clients on XT3
2148 Description: evictions taking too long
2149 Details : allow llrd to evict clients directly on OSTs
2152 Frequency : only on ppc
2154 Description: /proc/fs/lustre/devices broken on ppc
2155 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
2156 We should make sure the type of 'index' is loff_t and then cast
2157 explicitly as needed below. Do not assign an explicitly cast
2161 Frequency : only for rhel5
2163 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
2164 Details : Modify the target file & which_kernel.
2166 Severity : enhancement
2168 Description: omit set fsid for export NFS
2169 Details : fix set/restore device id for avoid EMFILE error and mark lustre
2170 fs as FS_REQUIRES_DEV for avoid problems with generate fsid.
2174 Frequency : Always, for kernels after 2.6.16
2175 Description: Fix warning idr_remove called for id=.. which is not allocated.
2176 Details : Last kernels save old s_dev before kill super and not allow
2177 to restore from callback - restore it before call kill_anon_super.
2181 Description: buffer overruns could theoretically occur
2182 Details : llapi_semantic_traverse() modifies the "path" argument by
2183 appending values to the end of the origin string, and a
2184 overrun may occur. Adding buffer overrun check in liblustreapi.
2188 Description: Fix error on 'ls .' at the top of the Lustre mount.
2189 Details : Don't revalidate dentry if it is a root dentry.
2193 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
2194 Details : Modify vm-tunables-rhel4.patch.
2198 Description: Kernel config for 2.6.18-vanilla.
2199 Details : Modify targets/2.6-vanilla.target.in.
2200 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
2201 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
2202 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
2203 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
2207 Description: improve handling recoverable errors
2208 Details : if request processig with error which can be recoverable on server
2209 request should be resend, otherwise page released from cache and
2214 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
2215 Details : Modify target file & which_patch.
2216 A flaw was found in the IA32 system call emulation provided
2217 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
2218 value could be stored in the %RAX register, which could trigger an
2219 out-of-bounds system call table access. An untrusted local user
2220 could exploit this flaw to run code in the kernel
2221 (ie a root privilege escalation). (CVE-2007-4573).
2225 Description: O_DIRECT bypasses client statistics.
2226 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2227 read_ahead_stats were updated. copied the stats section from
2228 osc_send_oap_rpc() into async_internal().
2232 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla
2236 Description: change order of libsysio includes
2237 Details : '#include sysio.h' should always come before '#include xtio.h'
2241 Description: handle "serious error: objid * already exists" more gracefully
2242 Details : If LAST_ID value on disk is smaller than the objects existing in
2243 the O/0/d* directories, it indicates disk corruption and causes an
2244 LBUG(). If the object is 0-length, then we should use the existing
2245 object. This will help to avoid a full fsck in most cases.
2247 Severity : enhancement
2249 Description: adapt the lustre_config script to support the upgrade case
2250 Details : Add "-u" option for lustre_config script to support upgrading 1.4
2251 server targets to 1.6 in parallel.
2255 Description: To avoid grant space > avaible space when the disk is almost
2256 full. Without this patch you might see the error "grant XXXX >
2257 available" or some LBUG about grant, when the disk is almost
2259 Details : In filter_check_grant, for non_grant cache write, we should
2260 check the left space by if (*left > ungranted + bytes), instead
2261 of (*left > ungranted), because only we are sure the left space
2262 is enough for another "bytes", then the ungrant space should be
2263 increase. In client, we should update cl_avail_grant only there
2264 is OBD_MD_FLGRANT in the reply.
2268 Description: Update RHEL 4 kernel to fix local root privilege escalation.
2269 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
2270 described in CVE-2007-4573. This problem could allow untrusted
2271 local users to gain root access.
2274 Frequency : when using O_DIRECT and quotas
2276 Description: Incorrect file ownership on O_DIRECT output files
2277 Details : block usage reported by 'lfs quota' does not take into account
2278 files that have been written with O_DIRECT.
2283 Description: touch file failed when fs is not full
2284 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
2285 otherwise we can get ENOSP while fs is not full.
2289 Description: parallel lock callbacks
2290 Details : Instead of sending blocking and completion callbacks as separated
2291 requests, adding them to a set and sending in parallel.
2294 Frequency : only for Cray XT3
2295 Bugzilla : 12829/13455
2296 Description: Changing primary group doesn't change the group lustre assigns to
2298 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
2299 overridden with the primary group provided by the group upcall,
2300 whereas the supplied fsgid can be trusted if it is in the list of
2301 supplementary groups returned by the group upcall.
2303 Severity : enhancement
2305 Description: Allow masking D_WARNING, D_ERROR messages from console
2306 Details : Console messages can now be disabled via lnet.printk.
2310 Description: User code with malformed file open parameter crashes client node
2311 Details : Before packing join_file req, all the related reference should be
2312 checked carefully in case some malformed flags cause fake
2313 join_file req on client.
2317 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
2318 Details : ldlm_completion_ast() assumes that a lock is granted when the req
2319 mode is equal to the granted mode. However, it should also check
2320 that LDLM_FL_CP_REQD is not set.
2324 Description: Heavy nfs access might result in deadlocks
2325 Details : After ELC code landed, it is now improper to enqueue any mds
2326 locks under och_sem, because enqueue might want to decide to
2327 cancel open locks for same inode we are holding och_sem for.
2331 Description: Client eviction while running blogbench
2332 Details : A lot of unlink operations with concurrent I/O can lead to a
2333 deadlock causing evictions. To address the problem, the number of
2334 oustanding OST_DESTROY requests is now throttled to
2335 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
2336 callbacks are processed in priority.
2340 Description: enable ACLs on MDS by default
2341 Details : ACLs must be enabled on MDS by default.
2344 Frequency : PPC/PPC64 only
2346 Description: conflicts between asm-ppc64/types.h and lustre_types.h
2347 Details : fix duplicated definitions between asm-ppc64/types.h and
2348 lustre_types.h on PPC.
2351 Frequency : PPC/PPC64 only
2353 Description: asm-ppc/segment.h does not exist
2354 Details : fix compile issue on PPC.
2358 Description: better handle error messages in extents code
2361 Frequency : RHEL4 only
2363 Description: mkfs is very slow on IA64/RHEL4
2364 Details : A performance regression has been discovered in the MPT Fusion
2365 driver between versions 3.02.73rh and 3.02.99.00rh. As a
2366 consequence, we have downgraded the MPT Fusion driver in the RHEL4
2367 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
2369 Severity : enhancement
2371 Description: SNMP support enhancement
2372 Details : Adding total number of sampled request for an MDS node in snmp
2375 Severity : enhancement
2377 Description: Optimize ldlm waiting list processing for PR extent locks
2378 Details : When processing waiting list for read extent lock and meeting read
2379 lock that is same or wider to it that is not contended, skip
2380 processing rest of the list and immediatelly return current
2381 status of conflictness, since we are guaranteed there are no
2382 conflicting locks in the rest of the list.
2386 Description: Time out and refuse to reconnect
2387 Details : When the failover node is the primary node, it is possible
2388 to have two identical connections in imp_conn_list. We must
2389 compare not conn's pointers but NIDs, otherwise we can defeat
2390 connection throttling.
2394 Description: port llog fixes from b1_6 into HEAD
2395 Details : Port llog reference couting and some llog cleanups from b1_6
2396 (bug 10800) into HEAD, for protect from panic and access to already
2397 free llog structures.
2401 Description: Detect stride IO mode in read-ahead
2402 Details : When a client does stride read, read-ahead should detect that and
2403 read-ahead pages according to the detected stride pattern.
2407 Description: data checksumming impacts single node performance
2408 Details : add support for several checksum algorithm. Currently, only CRC32
2409 and Adler-32 are supported. The checksum type can be changed on
2410 the fly via /proc/fs/lustre/osc/*/checksum_type.
2414 Description: use adler32 for page checksums
2415 Details : when available, use the Adler-32 algorithm instead of CRC32 for
2420 Description: build for x2 fails
2421 Details : fix compile issue on Cray systems.
2425 Description: Properly match for duplicate locks
2426 Details : Due to different lock order from skiplists code, we need to
2427 traverse entire list for now
2430 Frequency : only on PPC/SLES10
2432 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
2433 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
2434 be defined. Add a hack in mkfs_lustre.c to work around this compile
2439 Description: LASSERT on MDS when client holding flock lock dies
2440 Details : ldlm pool logic depends on number of granted locks equal to
2441 number of released locks which is not true for flock locks, so
2442 just exclude such locks from consideration.
2446 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
2447 Details : Use fsfilt_sync() for both the whole filesystem sync and
2448 individual file sync to eliminate dangerous inode locking
2449 with I_LOCK that can lead to a deadlock.
2453 Description: performance in 1.6.3
2454 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
2455 sure that 1MB requests are not fragmented by the block layer.
2457 Severity : enhancement
2459 Description: organize the server-side client stats on per-nid basis
2460 Details : Change the structure of stats under obdfilter and mds to
2468 The "uuid"s file would list the uuids of _active_ exports.
2469 And the clear entry is to clear all stats and stale nids.
2471 Severity : enhancement
2473 Description: eliminate client locks in face of contention
2474 Details : file contention detection and lockless i/o implementation
2475 for contended files.
2479 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
2482 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
2484 Description: MDS may lose file striping (and hence file data) in some cases
2485 Details : If there are additional extended attributes stored on the MDS,
2486 in particular ACLs, SELinux, or user attributes (if user_xattr
2487 is specified for the client mount options) then there is a risk
2488 of attribute loss. Additionally, the Lustre file striping
2489 needs to be larger than default (e.g. striped over all OSTs),
2490 and an additional attribute must be stored initially in the
2491 inode and then increase in size enough to be moved to the
2492 external attribute block (e.g. ACL growing in size) for file
2497 Description: skiplist implementation simplification
2498 Details : skiplists are used to group compatible locks on granted list
2499 that was implemented as tracking first and last lock of each
2500 lock group the patch changes that to using doubly linked lists
2504 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
2505 Details : Change LASSERTs to client eviction (i.e. abort client's recovery)
2506 because LASSERT on both the data supplied by a client, and the
2507 data on disk is dangerous and incorrect.
2509 Severity : enhancement
2511 Description: Slow truncate/writes to huge files at high offsets.
2512 Details : Directly associate cached pages to lock that protect those pages,
2513 this allows us to quickly find what pages to write and remove
2514 once lock callback is received.
2518 Description: more ldlm soft lockups
2519 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
2520 starve other threads from the resource lock for a long time in
2521 case of long waiting queue, so change the debug level from
2522 D_OTHER to the less frequently used D_INFO.
2524 Severity : enhancement
2526 Description: add -gid, -group, -uid, -user options to lfs find
2530 Description: Hung threads in invalidate_inode_pages2_range
2531 Details : The direct IO path doesn't call check_rpcs to submit a new RPC
2532 once one is completed. As a result, some RPCs are stuck in the
2533 queue and are never sent.
2537 Description: filter threads hungs on waiting journal commit
2538 Details : Cleanup filter group llog code, then only filter group llog will
2539 be only created in the MDS/OST syncing process.
2543 Description: Procfs and llog threads access destoryed import sometimes.
2544 Details : Sync the import destoryed process with procfs and llog threads by
2545 the import refcount and semaphore.
2547 Severity : enhancement
2549 Description: openlock cache of b1_6 port to HEAD
2554 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
2555 Details : If insertion of an extent fails, then discard the inode
2556 preallocation and free data blocks else it can lead to duplicate
2561 Description: don't always update ctime in ext3_xattr_set_handle()
2562 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle.
2563 In some cases the ctime should not be updated, for example for
2564 2.0->1.8 compatibility it is necessary to delete an xattr and it
2565 should not update the ctime.
2569 Bugzilla : 15713/16362
2570 Description: Assertion in iopen_connect_dentry in 1.6.3
2571 Details : looking up an inode via iopen with the wrong generation number can
2572 populate the dcache with a disconneced dentry while the inode
2573 number is in the process of being reallocated. This causes an
2574 assertion failure in iopen since the inode's dentry list contains
2575 both a connected and disconnected dentry.
2579 Description: assertion failure in ldlm_handle2lock()
2580 Details : fix a race between class_handle_unhash() and class_handle2object()
2581 introduced in lustre 1.6.5 by bug 13622.
2586 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
2587 Details : remove the limit of 256 scsi disks in the sd_iostat patch
2592 Description: oops in sd_iostats_seq_show()
2593 Details : unloading/reloading the scsi low level driver triggers a kernel
2594 bug when trying to access the sd iostat file.
2599 Description: Kernel panics during QLogic driver reload
2600 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
2601 patch, causing memory corruption.
2606 Description: journal_dev option does not work in b1_6
2607 Details : pass mount option during pre-mount.
2609 Severity : enhancement
2611 Description: Add a FIEMAP(FIle Extent MAP) ioctl
2612 Details : FIEMAP ioctl will allow an application to efficiently fetch the
2613 extent information of a file. It can be used to map logical blocks
2614 in a file to physical blocks in the block device.
2618 Description: LDLM soft lockups - improvement
2619 Details : It is be possible to send the lock handle along with each read
2620 or write request because the client is already doing a lock match
2621 itself so there isn't any reason the OST should have to re-do that
2625 Frequency : only on Cray X2
2627 Description: X2 build failures
2628 Details : fix build failures on Cray X2.
2632 Description: xid & resent requests
2633 Details : Initialize RPC XID from clock at startup (randomly if clock is
2636 Severity : enhancement
2638 Description: Add lustre_start utility to start or stop multiple Lustre servers
2643 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
2644 Details : In case of memory pressure, list_del() can be called twice on
2645 req->rq_history_list, causing a kernel oops.
2649 Description: kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
2650 Details : fix stack overflow in the distributed lock manager by defering
2651 export eviction after a failed AST to the elt thread instead of
2652 handling it in the dlm interpret routine.
2656 Description: Convert some comments to new format.
2657 Details : Update documenting comments to match doxygen conventions.
2661 Description: Grammar fixes.
2662 Details : A couple of trivial spelling fixes.
2666 Description: OSD_COUNTERS-mandatory
2667 Details : Make previously optional ->oti_{w,r}_locks sanity checks mandatory
2668 to simplify the code.
2672 Description: simplify cmm_device freeing logic.
2673 Details : Call cmm_device_free() in the failure path of cmm_device_alloc().
2677 Description: Add lockdep support to dt_object_operations locking interface.
2678 Details : Augment ->do_{read,write}_lock() prototypes with a `role'
2679 parameter indicating lock ordering. Update mdd code to use new
2684 Description: Introduce failloc constants for lockless IO tests.
2685 Details : Add two new failloc constants to test lockless IO. Only one of
2686 them in implemented---another is checked in yet to be landed
2691 Description: Add lockdep support for inode mutex.
2692 Details : Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used
2693 in the situations where Lustre has to lock more than one inode
2698 Description: Add optional invariants checking support.
2699 Details : Add new LINVRNT() macro, optional on new --enable-invariants
2700 configure switch. This macro is to be used for consistency and
2701 sanity checks that are too expensive to be left in `production'
2706 Description: Zap lock->l_granted_mode with explicit LCK_MINMODE.
2707 Details : Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to
2712 Description: Add lockdep support for ldlm_lock and ldlm_resource.
2713 Details : Use spin_lock_nested() in (the only) situation where more than
2714 one ldlm_lock is locked simultaneously. Also, fix possible
2715 dead-lock in ldlm_lock_change_resource() by enforcing particular
2720 Description: Use struct ldlm_callback_suite in ldlm_lock_create().
2721 Details : Instead of specifying each ldlm_lock call-back through separate
2722 parameter, wrap them into struct ldlm_callback_suite.
2726 Description: Kill join_lru obd method and its callers.
2727 Details : CLIO uses lock weighting policy to keep locks over mmapped regions
2728 in memory---a requirement implemented through ->o_join_lru() obd
2729 method in HEAD. Remove this method and its users.
2733 Description: Add asynchronous ldlm ENQUEUE completion handler.
2734 Details : CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a
2735 case that stock ldlm_completion_ast() cannot handle as it waits
2736 until lock is granted. Introduce new ldlm_completion_ast_async()
2737 for this. Also comment ldlm_completion_ast().
2741 Description: ldlm_error <-> errno conversion.
2742 Details : Add functions to map (rather arbitrary) between LDLM error codes
2743 and standard errno values. CLIO needs this to prevent LDLM specific
2744 constants from escaping ldlm and osc.
2748 Description: Kill unused ldlm_handle2lock_ns() function.
2749 Details : Kill unused ldlm_handle2lock_ns() function.
2753 Description: Add lu_ref support to ldlm_lock
2754 Details : lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
2755 lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
2756 and ldlm_resource. LDLM interface has to be changed, because code that
2757 releases a reference on a lock, has to "know" what reference this is.
2758 In the most frequent case
2760 lock = ldlm_handle2lock(handle);
2762 LDLM_LOCK_PUT(lock);
2764 no changes are required. When any other reference (received _not_ from
2765 ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
2766 instead of LDLM_LOCK_PUT().
2768 Arguably, changes are pervasive, and interface requires some discipline
2769 for proper use. On the other hand, it was very instrumental in finding
2770 a few leaked lock references.
2774 Description: Add ldlm_lock_addref_try().
2775 Details : Introduce ldlm_lock_addref_try() function (used by CLIO) that
2776 attempts to addref a lock that might be being canceled
2781 Description: Add ldlm_weigh_callback().
2782 Details : Add new ->l_weigh_ast() call-back to ldlm_lock. It is called
2783 by ldlm_cancel_shrink_policy() to estimate lock "value", instead of
2784 hard-coded `number of pages' logic.
2788 Description: Add lockdep annotations to llog code.
2789 Details : Use appropriately tagged _nested() locking calls in the places
2790 where llog takes more than one ->lgh_lock lock.
2794 Description: Add loi_kms_set().
2795 Details : Wrap kms updates into a helper function.
2799 Description: Constify instances of struct lsm_operations.
2800 Details : Constify instances of struct lsm_operations.
2804 Description: lu_conf support.
2805 Details : On a server, a file system object is uniquely identified
2806 by a fid, which is sufficient to locate and load all object
2807 state (inode). On a client, on the other hand, more data are
2808 necessary instantiate an object. Change lu_object_find() and
2809 friends to take additional `lu_conf' argument describing object.
2810 Typically this includes layout information.
2814 Description: lu_context fixes.
2815 Details : Introduce new lu_context functions that are needed on the client
2816 side, where some system threads (ptlrpcd) are shared by multiple
2817 modules, and so cannot be stopped during module shutdown.
2821 Description: Add start and stop methods to lu_device_type_operations.
2822 Details : Introduce two new methods in lu_device_type_operations, that are
2823 invoked when first instance of a given type is created and last one
2824 is destroyed respectively. This is need by CLIO.
2828 Description: Add lu_ref support to struct lu_device.
2829 Details : Add lu_ref support to lu_object and lu_device. lu_ref is used to
2830 track leaked references.
2834 Description: Introduce lu_kmem_descr.
2835 Details : lu_kmem_descr and its companion interface allow to create
2836 and destroy a number of kmem caches at once.
2840 Description: Fix lu_object finalization race.
2841 Details : Fix a race between lu_object_find() finding an object and its
2842 concurrent finalization. This race is (most likely) not possible
2843 on the server, but might happen on the client.
2847 Description: Introduce lu_ref interface.
2848 Details : lu_ref is a debugging module allowing to track references to
2849 a given object. It is quite cpu expensive, and has to be
2850 explicitly enabled with --enable-lu_ref. See usage description
2855 Description: Factor lu_site procfs stats into a separate function.
2856 Details : Separate lu_site stats printing code into a separate function
2857 that can be reused on a client.
2861 Description: Constify instances of struct {lu,dt,md}_device_operations.
2862 Details : Constify instances of struct {lu,dt,md}_device_operations.
2866 Description: Introduce struct md_site and move meta-data specific parts of
2867 struct lu_site here.
2868 Details : Move md-specific fields out of struct lu_site into special struct
2869 md_site, so that lu_site can be used on a client.
2873 Description: Kill mdd_lov_destroy().
2874 Details : Remove unused mdd code.
2878 Description: Add st_block checking to multistat.c.
2879 Details : Add st_block checking to multistat.c.
2883 Description: Add lu_ref support to struct obd_device.
2884 Details : Add lu_ref tracking to obd_device.
2888 Description: Kill obd_set_fail_loc().
2889 Details : Remove unused code.
2893 Description: Add special type for ptlrpc_request interpret functions.
2894 Details : Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
2895 there. Actual usage will be in CLIO.
2899 Description: Replace RW_LOCK_UNLOCKED() macro with rwlock_init().
2900 Details : Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former
2901 doesn't work with lockdep.
2905 Description: Add rwv.c test program.
2906 Details : New testing program exercising readv(2) and writev(2) (Qian).
2910 Description: Add sendfile.c test program.
2911 Details : New testing program exercising sendfile(2) (Jay).
2915 Description: Ratelimit a message that can be very frequent.
2916 Details : Ratelimit a memory allocation failure message that can
2921 Description: Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2922 Details : Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
2926 Description: Liblustre build fixes.
2927 Details : Liblustre build fixes.
2931 Description: libcfs: add cfs_{need,cond}_resched() interface.
2932 Details : libcfs: add cfs_{need,cond}_resched() definition and
2933 implementations for Linux, NT, and liblustre.
2935 Severity : enhancement
2937 Description: More exported tunables for mballoc
2938 Details : Add support for tunable preallocation window and new tunables for
2939 large/small requests
2943 Description: Detect corruption of block bitmap and checking for preallocations
2944 Details : Checks validity of on-disk block bitmap. Also it does better
2945 checking of number of applied preallocations. When corruption is
2946 found, it turns filesystem readonly to prevent further corruptions.
2950 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
2951 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
2952 window to 0. If it does hit cache miss, and read-pattern is still
2953 stride-io mode, does not reset the stride window, but also does not
2954 increase the stride window length in this case.
2958 Frequency : only for big-endian servers
2959 Description: Check if system is big-endian while mounting fs with extents feature
2960 Details : Mounting a filesystem with extents feature will fail on big-endian
2961 systems since ext3-based ldiskfs is not supported on big-endian
2962 systems. This can be over-riden with "bigendian_extents" mount option.
2964 Severity : enhancement
2966 Description: The root squash functionality
2967 Details : A security feature, which is to prevent users from being able
2968 to mount lustre on their desktop, run as root, and delete
2969 all of the files in the filesystem. The goal is accomplished by
2970 remapping user id (UID) and group id (GID) of the root user to
2971 a UID and GID specified by the system administartor via Lustre
2972 configuration management server (MGS). The functionality also
2973 allows to specify sets of clients for which the remapping does
2978 Description: Excessive recovery window
2979 Details : With AT enabled, the recovery window can be excessively long (6000+
2980 seconds). To address this problem, we no longer use
2981 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
2982 timeout no longer depends on the service time, it is set to
2983 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
2984 time via pb_service_time.
2988 Description: Watchdog triggered on MDS failover
2989 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
2990 the OSTs know that the MDS "UUID" can be reused for the same export
2991 from a different NID, so we do not need to wait for the export to be
2995 Frequency : rare, only if using MMP with Linux RAID
2997 Description: MMP doesn't work with Linux RAID
2998 Details : While using HA for Lustre servers with Linux RAID, it is possible
2999 that MMP will not detect multiple mounts. To make this work we
3000 need to unplug the device queue in RAID when the MMP block is being
3001 written. Also while reading the MMP block, we should read it from
3002 disk and not the cached one.
3004 Severity : enhancement
3006 Description: open file using fid
3007 Details : A file can be opened using just its fid, like
3008 <mntpt>/.lustre/fid/SEQ:OID:VER - this is needed for HSM and replication
3011 Frequency : Only in RHEL5 when mounting multiple ext3 filesystems
3014 Description: "kmem_cache_create: duplicate cache jbd_4k" error message
3015 Details : add proper locking for creation of jbd_4k slab cache
3019 Description: MMP check in ext3_remount() fails without displaying any error
3020 Details : When multiple mount protection fails during remount, proper error
3023 Severity : enhancement
3025 Description: Allow stripe size to be up to 4G-64k
3026 Details : Fix math logic to allow large stripe sizes.
3030 Description: add check for >8TB ldiskfs filesystems
3031 Details : ext3-based ldiskfs does not support greater than 8TB LUNs.
3032 Don't allow >8TB ldiskfs filesystems to be mounted without
3033 force_over_8tb mount option
3038 Descriptoin: OSS complains "too many blocks" and crashes
3039 Details : In filter_commitrw_write():
3040 1. lquota_chkquota();
3041 2. obdo_from_inode();
3042 3. lquota_pending_commit();
3043 "2." has changed uid and gid in this case, then caused
3044 lquota_chkquota() and lquota_pending_commit() to use
3045 different uid and gid. That is the root of the bug.
3049 Descriptoin: getstripe doesn't report stripe count = -1 sanely
3050 Details : Detect and print out the 'stripe count = -1' case properly.
3054 Description: Out or order replies might be lost on replay
3055 Details : In ptlrpc_retain_replayable_request if we cannot find retained
3056 request with tid smaller than one currently being added, add it
3057 to the start, not end of the list.
3059 --------------------------------------------------------------------------------
3061 2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
3063 * Support for kernels:
3064 2.6.5-7.283 (SLES 9),
3065 2.6.9-55.EL (RHEL 4),
3066 2.6.16.46-0.14 (SLES 10),
3067 2.6.18.8 vanilla (kernel.org)
3068 * Client support for unpatched kernels:
3069 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
3070 2.6.16 - 2.6.22 vanilla (kernel.org)
3071 * Due to recently discovered recovery problems, we do not recommend
3072 using patchless RHEL 4 clients with this or any earlier release.
3073 * Recommended e2fsprogs version: 1.39.cfs8
3074 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
3075 * Starting with this release, the ldiskfs backing filesystem required
3076 by Lustre is now in its own package, lustre-ldiskfs. This package
3077 should be installed. It is versioned separately from Lustre and
3078 may be released separately in future.
3082 Description: block reactivating mgc import until all deactivates complete
3083 Details : Fix race when failing back MDT/MGS to itself (testing)
3086 Frequency : at statup only
3088 Description: mds_lov_synchronize race leads to various problems
3089 Details : simultaneous MDT->OST connections at startup can cause the
3090 sync to abort, leaving the OSC in a bad state.
3092 Severity : enhancement
3094 Description: add optional extra BUILD_VERSION info
3095 Details : add a new environment variable (namely LUSTRE_VERS) which allows
3096 to override the lustre version.
3099 Frequency : 2.6.18 servers only
3101 Description: ll_kern_mount() doesn't release the module reference
3102 Details : The ldiskfs module reference count never drops down to 0
3103 because ll_kern_mount() doesn't release the module reference.
3108 Description: server LBUG when using old ost_num_threads parameter
3109 Details : Accept the old ost_num_threads parameter but warn that it
3110 is deprecated, and fix an off-by-one error that caused an LBUG.
3115 Description: Transient SCSI error results in persistent IO issue
3116 Details : iobuf->dr_error is not reinitialized to 0 between two
3120 Frequency : sometimes when underlying device returns I/O errors
3122 Description: OSTs not going read-only during write failures
3123 Details : OSTs are not remounted read-only when the journal commit threads
3124 get I/O errors because fsfilt_ext3 calls journal_start/stop()
3125 instead of the ext3 wrappers.
3129 Description: poor connect scaling with increasing client count
3130 Details : Don't run filter_grant_sanity_check for more than 100 exports
3131 to improve scaling for large numbers of clients.
3134 Frequency : SLES10 only
3136 Description: sanity-quota.sh quotacheck failed: rc = -22
3137 Details : Quotas cannot be enabled on SLES10.
3140 Frequency : liblustre clients only
3142 Description: getdirentries does not give error when run on compute nodes
3143 Details : getdirentries does not fail when the size specified as an argument
3144 is too small to contain at least one entry
3146 Severity : enhancement
3148 Description: Add LNET router traceability for debug purposes
3149 Details : If a checksum failure occurs with a router as part of the
3150 IO path, the NID of the last router that forwarded the bulk data
3151 is printed so it can be identified.
3156 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
3157 Details : Due to a race condition, liblustre clients were occasionally
3158 evicted incorrectly.
3160 Severity : enhancement
3162 Description: lfs setstripe use optional parameters instead of postional
3165 Severity : enhancement
3167 Description: Nanosecond timestamp support for ldiskfs
3168 Details : The on-disk ldiskfs filesystem has added support for nanosecond
3169 resolution timestamps. There is not yet support for this at
3170 the Lustre filesystem level.
3173 Frequency : during server recovery
3175 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
3176 Details : request with rq_no_resend flag not awake l_wait_event if they get
3180 Frequency : nfs export on patchless client
3182 Description: connectathon hang when test nfs export over patchless client
3183 Details : Disconnected dentry cannot be found with lookup, so we do not need
3184 to unhash it or make it invalid
3187 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
3188 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
3189 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
3194 Description: open req refcounting wrong on reconnect
3195 Details : If reconnect happened between getting open reply from server and
3196 call to mdc_set_replay_data in ll_file_open, we will schedule
3197 replay for unreferenced request that we are about to free.
3198 Subsequent close will crash in variety of ways.
3199 Check that request is still eligible for replay in
3200 mdc_set_replay_data().
3205 Description: disable writes to filesystem when reading health_check file
3206 Details : the default for reading the health_check proc file has changed
3207 to NOT do a journal transaction and write to disk, because this
3208 can cause reads of the /proc file to hang and block HA state
3209 checking on a healthy but otherwise heavily loaded system. It
3210 is possible to return to the previous behaviour during configure
3211 with --enable-health-write.
3213 Severity : enhancement
3215 Description: 64-bit inode version
3216 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
3217 to the inode. This will be required for version-based recovery.
3222 Description: MDS fails to start if a duplicate client export is detected
3223 Details : in some rare cases it was possible for a client to connect to
3224 an MDS multiple times. Upon recovery the MDS would detect this
3225 and fail during startup. Handle this more gracefully.
3227 Severity : enhancement
3229 Description: Add -o localflock option to simulate old noflock
3231 Details : This will achieve local-only flock/fcntl locks
3237 Description: log_commit_thread vs filter_destroy race leads to crash
3238 Details : Take import reference before releasing llog record semaphore
3243 Description: Wrong request locking in request set processing
3244 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
3245 imp_delayed_list, in this place should be used imp_lock.
3248 Frequency : when reconnection
3250 Description: Grant Leak when osc reconnect to OST
3251 Details : When osc reconnect ost, OST(filter) should check whether it
3252 should grant more space to client by comparing fed_grant and
3253 cl_avail_grant, and return the granted space to client instead
3254 of "new granted" space, because client will call osc_init_grant
3255 to update the client grant space info.
3258 Frequency : when client reconnect to OST
3260 Description: Grant Leak when osc do resend and replay bulk write
3261 Details : When osc reconnect to OST, OST(filter)should clear grant info of
3262 bulk write request, because the grant info while be sync between
3263 OSC and OST when reconnect, and we should ignore the grant info
3264 these of resend/replay write req.
3269 Description: Grant space more than available left space sometimes.
3270 Details : When then OST is about to be full, if two bulk writing from
3271 different clients came to OST. Accord the avaliable space of the
3272 OST, the first req should be permitted, and the second one
3273 should be denied by ENOSPC. But if the seconde arrived before
3274 the first one is committed. The OST might wrongly permit second
3275 writing, which will cause grant space > available space.
3278 Frequency : when client is evicted
3280 Description: Grant might be wrongly erased when osc is evicted by OST
3281 Details : when the import is evicted by server, it will fork another
3282 thread ptlrpc_invalidate_import_thread to invalidate the
3283 import, where the grant will be set to 0. While the original
3284 thread will update the grant it got when connecting. So if
3285 the former happened latter, the grant will be wrongly errased
3286 because of this race.
3291 Description: Checking Stale with correct fid
3292 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
3293 to check whether it is stale, because sometimes, we want the
3294 enqueue happened anyway, and op_data.fid2 will not be initialized.
3296 Severity : enhancement
3298 Description: update patchless client
3299 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
3302 Frequency : only with 2.4 kernel
3304 Description: random memory corruption
3305 Details : size of struct ll_inode_info is to big for union inode.u and this
3306 can be cause of random memory corruption.
3311 Description: Memory leak in recovery
3312 Details : Lov_mds_md was not free in an error handler in mds_create_object.
3313 It should also check obd_fail before fsfilt_start, otherwise if
3314 fsfilt_start return -EROFS,(failover mds during mds recovery).
3315 then the req will return with repmsg->transno = 0 and rc = EROFS.
3316 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
3317 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
3318 be freed no matter whether fsfilt_commit success or not.
3321 Frequency : only with huge count clients
3323 Description: Prevents from taking the superblock lock in llap_from_page for
3325 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
3326 for llap_from_page call in ll_removepage prevents from taking the
3327 superblock lock for a soon died page.
3332 Description: Not check open intent error before release open handle
3333 Details : in some rare cases, the open intent error is not checked before
3334 release open handle, which may cause
3335 ASSERTION(open_req->rq_transno != 0), because it tries to release
3336 the failed open handle.
3341 Description: Set cat log bitmap only after create log success.
3342 Details : in some rare cases, the cat log bitmap is set too early. and it
3343 should be set only after create log success.
3347 Description: Accessing a block bevice can re-enable I/O when Lustre is
3348 tearing down a device.
3349 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
3354 Description: service threads may hog cpus when there are a lot of requests
3355 Details : Insert cond_resched to give other threads a chance to use some CPU
3360 Description: the cat log was not initialized in recovery
3361 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
3362 unlink log on mds will not be initialized until mds post
3363 recovery. And also in mds post recovery, the unlink log will
3364 initialization will be done asynchronausly, so there will be race
3365 between add unlink log and unlink log initialization.
3369 Description: brw_stats were being printed incorrectly
3370 Details : brw_stats were being printed as log2 but all of them were not
3371 recorded as log2. Also remove some code duplication arising from
3372 filter_tally_{read,write}.
3376 Frequency : rare, only in recovery.
3377 Description: ASSERTION(req->rq_type != LI_POISON) failed
3378 Details : imp_lock should be held while iterating over imp_sending_list for
3379 prevent destroy request after get timeout in ptlrpc_queue_wait.
3383 Description: replay-single.sh test 52 fails
3384 Details : A lock's skiplist need to be cleanup when it being unlinked
3385 from its resource list.
3389 Description: Short directio read returns full requested size rather than
3391 Details : Direct I/O operations should return actual amount of bytes
3392 transferred rather than requested size.
3394 Severity : enhancement
3396 Description: metadata RPC reduction (e.g. for rm performance)
3397 Details : decrease the amount of synchronous RPC between clients and servers
3398 by canceling conflicing lock before the operation on the client
3399 and packing thier handles into the main operation RPC to server.
3401 Severity : enhancement
3403 Description: Async OSC create to avoid the blocking unnecessarily.
3404 Details : If an OST has no remaining object, system will block on the
3405 creation when it needs to create a new object on this OST. Now,
3406 ways use pre-created objects when available, instead of blocking on
3407 an empty osc while others are not empty. If we must block, we
3408 block for the shortest possible period of time.
3412 Description: improve handling recoverable errors
3413 Details : if request processig with error which can be recoverable on server
3414 request should be resend, otherwise page released from cache and
3417 Severity : enhancement
3419 Description: refine locking for avoid write wrong info into lov_objid file
3420 Details : fix possible races with add new target and write/update data in
3423 --------------------------------------------------------------------------------
3425 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
3430 Frequency : on some architectures
3432 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
3433 Details : Uninitialized flags sometimes cause configuration commands to
3437 Frequency : patchless clients only
3439 Description: missing __iget() symbol export
3440 Details : The __iget() symbol export is missing. To avoid the need for
3441 this on patchless clients the deathrow inode reaper is turned
3442 off, and we depend on the VM to clean up old inodes. This
3443 dependency was during via the fix for bug 12181.
3445 --------------------------------------------------------------------------------
3447 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
3449 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
3450 INTEROPERATE with older versions automatically. In many cases a
3451 special upgrade step is needed. Please read the
3452 user documentation before upgrading any part of a live system.
3453 * WIRE PROTOCOL CHANGE from previous 1.6 beta versions. This
3454 version will not interoperate with 1.6 betas before beta5 (1.5.95).
3455 * WARNING: Lustre configuration and startup changes are required with
3456 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
3460 Severity : enhancement
3462 Description: MountConf
3463 Details : Lustre configuration is now managed via mkfs and mount
3464 commands instead of lmc and lconf. New obd types (MGS, MGC)
3465 are added for dynamic configuration management. See
3466 https://mail.clusterfs.com/wikis/lustre/MountConf for
3469 Severity : enhancement
3471 Description: dynamic OST addition
3472 Details : OSTs can now be added to a live filesystem
3474 Severity : enhancement
3476 Description: startup order invariance
3477 Details : MDTs and OSTs can be started in any order. Clients only
3478 require the MDT to complete startup.
3480 Severity : enhancement
3482 Description: parallel, asynchronous orphan cleanup
3483 Details : orphan cleanup is now performed in separate threads for each
3484 OST, allowing parallel non-blocking operation.
3486 Severity : enhancement
3488 Description: optimized stripe assignment
3489 Details : stripe assignments are now made based on ost space available,
3490 ost previous usage, and OSS previous usage, in order to try
3491 to optimize storage space and networking resources.
3493 Severity : enhancement
3495 Description: Permanently set tunables
3496 Details : All writable /proc/fs/lustre tunables can now be permanently
3497 set on a per-server basis, at mkfs time or on a live system.
3499 Severity : enhancement
3501 Description: Lustre message v2
3502 Details : Add lustre message format v2.
3504 Severity : enhancement
3506 Description: client OST exclusion list
3507 Details : Clients can be started with a list of OSTs that should be
3508 declared "inactive" for known non-responsive OSTs.
3512 Description: SPEC SFS validation failure on NFS v2 over lustre.
3513 Details : Changes the blocksize for regular files to be 2x RPC size,
3514 and not depend on stripe size.
3516 Severity : enhancement
3518 Description: Multiple MD RPCs in flight.
3519 Details : Further unserialise some read-only MDS RPCs - learn about intents.
3520 To avoid overly-overloading MDS, introduce a limit on number of
3521 MDS RPCs in flight for a single client and add /proc controls
3522 to adjust this limit.
3524 Severity : enhancement
3526 Description: client read/write statistics
3527 Details : Add client read/write call usage stats for performance
3528 analysis of user processes.
3529 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
3530 file access. extents_stats shows chunk size distribution.
3531 extents_stats_per_process show chunk size distribution per
3534 Severity : enhancement
3536 Description: mds statistics
3537 Details : Add detailed mds operations statistics in
3538 /proc/fs/lustre/mds/*/stats.
3542 Description: Failure of copying files with lustre special EAs.
3543 Details : Client side always return success for setxattr call for lustre
3544 special xattr (currently only "trusted.lov").
3549 Description: Refcount LNET uuids
3550 Details : The global LNET uuid list grew linearly with every startup;
3551 refcount repeated list entries instead of always adding to
3554 Severity : enhancement
3556 Description: Dynamic service threads
3557 Details : Within a small range, start extra service threads
3558 automatically when the request queue builds up.
3561 Frequency : mixed-endian client/server environments
3563 Description: mixed-endian crashes
3564 Details : The new msg_v2 system had some failures in mixed-endian
3567 Severity : enhancement
3569 Description: Easy OST removal
3570 Details : OSTs can be permanently deactivated with e.g. 'lctl
3571 conf_param lustre-OST0001.osc.active=0'
3573 Severity : enhancement
3575 Description: MGS proc entries
3576 Details : Added basic proc entries for the MGS showing what filesystems
3579 Severity : enhancement
3581 Description: provide MGS failover
3582 Details : Added config lock reacquisition after MGS server failover.
3584 Severity : enhancement
3586 Description: add Linux 2.4 support
3587 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3591 Description: a large application tries to do I/O to the same resource and dies
3592 in the middle of it.
3593 Details : Check the req->rq_arrival time after the call to
3594 ost_brw_lock_get(), but before we do anything about
3595 processing it & sending the BULK transfer request. This
3596 should help move old stale pending locks off the queue as
3597 quickly as obd_timeout.
3600 Frequency : when an incorrect nid is specified during startup
3602 Description: ptlrpc connect to non-existant node causes kernel crash
3603 Details : LNET can't be re-entered from an event callback, which
3604 happened when we expire a message after the export has been
3605 cleaned up. Instead, hand the zombie cleanup off to another
3608 Severity : enhancement
3610 Description: plain/inodebits lock performance improvement
3611 Details : Grouping plain/inodebits in granted list by their request modes
3612 and bits policy, thus improving the performance of search through
3616 Frequency : only if OST filesystem is corrupted
3618 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3619 Details : for a short time RPCs with bulk IO are in the replay list,
3620 but replay of bulk IOs is unimplemented. If the OST filesystem
3621 is corrupted due to disk cache incoherency and then replay is
3622 started it is possible to trip an assertion. Avoid putting
3623 committed RPCs into the replay list at all to avoid this issue.
3626 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3628 Description: System hang on startup
3629 Details : This bug allowed the liblustre (e.g. catamount) client to
3630 return to the app before handling all startup RPCs. This
3631 could leave the node unresponsive to lustre network traffic
3632 and manifested as a server ptllnd timeout.
3634 Severity : enhancement
3636 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3637 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3638 boolean to control whether to print peer debug info when a
3639 client's RPC times out.
3642 Frequency : only for kernels with patches from Lustre below 1.4.3
3644 Description: Remove old rdonly API
3645 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3648 Frequency : only for devices with external journals
3650 Description: Set external device read-only also
3651 Details : During a commanded failover stop, we set the disk device
3652 read-only while the server shuts down. We now also set any
3653 external journal device read-only at the same time.
3656 Frequency : when upgrading from 1.4 while trying to change parameters
3658 Description: The wrong (new) MDC name was used when setting parameters for
3659 upgraded MDT's. Also allows changing of OSC (and MDC)
3660 parameters if --writeconf is specified at tunefs upgrade time.
3663 Frequency : when setting specific ost indicies
3665 Description: QOS code breaks on skipped indicies
3666 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3667 created with --index need not be sequential.
3669 Severity : enhancement
3671 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3672 Details : The uninit_groups feature works in conjunction with the kernel
3673 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3674 up the pass1 processing of e2fsck. This is a read-only feature
3675 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3676 mount filesystems that have had this feature enabled.
3678 Severity : enhancement
3680 Description: Improve multi-block allocation algorithm to avoid fragmentation
3681 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3682 allocation locality and avoid filesystem fragmentation.
3684 ------------------------------------------------------------------------------
3686 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3688 * Support for kernels:
3689 2.4.21-47.0.1.EL (RHEL 3)
3690 2.6.5-7.283 (SLES 9)
3691 2.6.9-42.0.10.EL (RHEL 4)
3692 2.6.12.6 vanilla (kernel.org)
3693 2.6.16.27-0.9 (SLES 10)
3694 * Recommended e2fsprogs version: 1.39.cfs5
3696 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3700 Frequency : occasional, depends on client load and configuration
3701 Bugzilla : 12181, 12203
3702 Description: data loss for recently-modified files
3704 Details : In some cases it is possible that recently written or created
3705 files may not be written to disk in a timely manner (this should
3706 normally be within 30s unless client IO load is very high).
3707 The problem appears as zero-length files or files that are a
3708 multiple of 1MB in size after a client crash or client eviction
3709 that are missing data at the end of the file.
3711 This problem is more likely to be hit on clients where files are
3712 repeatedly created and unlinked in the same directory, clients
3713 have a large amount of RAM, have many CPUs, the filesystem has
3714 many OSTs, the clients are rebooted frequently, and/or the files
3715 are not accessed by other nodes after being written.
3717 The presence of the problem can be detected by looking at
3718 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3719 smaller than the second (nr_unused) then dirty files will not
3720 be flushed automatically to disk. "sync; sleep 10" should be
3721 run several times on the node before unmounting it to update
3722 Lustre (this is also safe to run on nodes without this problem).
3724 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3725 that can cause this same problem, so customers using that kernel
3726 also need to update the kernel in addition to Lustre. In order
3727 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3729 It is normal that files written just before a client crash (less
3730 than 30s) may not yet have been flushed to disk, even for local
3734 Frequency : frequent on thin XT3 nodes
3736 Description: UUID collision on thin XT3 Linux nodes
3737 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3738 randomly, since we relied on an insufficiently-seeded PRNG.
3743 Description: OSS hangs after "All ost request buffers busy"
3744 Details : A deadlock between quota and journal operations caused OSS
3745 hangs after printing "All ost request buffers busy."
3748 Frequency : always on liblustre builds
3750 Description: Cleanup compiler warnings on liblustre
3753 Frequency : always on liblustre builds on XT3
3755 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3759 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3760 > 32000 subdirectories
3761 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3762 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3763 subdirectory count crosses 32000. This will aid e2fsck to
3764 correctly handle more than 32000 subdirectories.
3767 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3769 Description: System hang on startup
3770 Details : This bug allowed the liblustre (e.g. catamount) client to
3771 return to the app before handling all startup RPCs. This
3772 could leave the node unresponsive to lustre network traffic
3773 and manifested as a server ptllnd timeout.
3775 Severity : enhancement
3777 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3778 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3779 boolean to control whether to print peer debug info when a
3780 client's RPC times out.
3785 Description: make O_SYNC working on 2.6 kernels
3786 Details : 2.6 kernels use different method for mark pages for write,
3787 so need add a code to lustre for O_SYNC work.
3792 Description: Failure to close file and release space on NFS
3793 Details : Put inode details into lock acquired in ll_intent_file_open.
3794 Use mdc_intent_lock in ll_intent_open to properly
3795 detect all kind of errors unhandled by mdc_enqueue.
3800 Description: proc file read during shutdown sometimes raced obd removal,
3802 Details : Add lock to prevent obd access after proc file removal.
3805 Frequency : Only for files larger than 4GB on 32-bit clients.
3807 Description: improperly doing page alignment of locks
3808 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3809 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3814 Description: RPCs being resent when they shouldn't be
3815 Details : Some RPCs that should not be resent are being resent. This
3816 can cause inconsistencies in the RPC state machine. Do not
3817 resend such requests.
3820 Frequency : rare, only with NFS export
3822 Description: Crash on NFS re-export node
3823 Details : under very unusual load conditions an assertion is hit in
3824 ll_intent_file_open()
3827 Frequency : only if OST filesystem is corrupted
3829 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3830 Details : for a short time RPCs with bulk IO are in the replay list,
3831 but replay of bulk IOs is unimplemented. If the OST filesystem
3832 is corrupted due to disk cache incoherency and then replay is
3833 started it is possible to trip an assertion. Avoid putting
3834 committed RPCs into the replay list at all to avoid this issue.
3839 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3840 Details : Large single O_DIRECT read and write calls can fail to allocate
3841 a sufficiently large buffer to process the request. In case of
3842 allocation failure the allocation is retried with a smaller
3843 buffer and broken into smaller requests.
3845 Severity : enhancement
3847 Description: Add -o localflock option to simulate old noflock behaviour.
3848 Details : This will achieve local-only flock/fcntl locks coherentness.
3853 Description: versioning check is incomplete
3854 Details : Checking the version difference of client vs. server, report
3855 error if the gap is too big.
3860 Description: add support PG_writeback bit
3861 Details : add support for PG_writeback bit for Lustre, for more carefull
3862 work with page cache in 2.6 kernel. This also fix some deadlocks
3863 and remove hack for work O_SYNC with 2.6 kernel.
3865 Severity : enhancement
3867 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3868 Details : The uninit_groups feature works in conjunction with the kernel
3869 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3870 up the pass1 processing of e2fsck. This is a read-only feature
3871 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3872 mount filesystems that have had this feature enabled.
3874 Severity : enhancement
3876 Description: Improve multi-block allocation algorithm to avoid fragmentation
3877 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3878 allocation locality and avoid filesystem fragmentation.
3880 ------------------------------------------------------------------------------
3882 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3884 * Support for kernels:
3885 2.6.9-42.0.3.EL (RHEL 4)
3886 2.6.5-7.276 (SLES 9)
3887 2.4.21-47.0.1.EL (RHEL 3)
3888 2.6.12.6 vanilla (kernel.org)
3889 2.6.16.21-0.8 (SLES10)
3890 * Recommended e2fsprogs version: 1.39.cfs2-0
3892 * The backwards-compatible /proc/sys/portals symlink has been removed
3893 in this release. Before upgrading, please ensure that you change
3894 any configuration scripts or /etc/sysctl.conf files that access
3895 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3896 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3897 in advance of the upgrade on any system running Lustre 1.4.6 or
3898 newer, since /proc/sys/lnet was added in that version.
3899 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3903 Frequency : only when quota is used
3905 Description: avoid scanning export list for quota master
3906 Details : Change the algorithms to avoid scanning export list in order
3907 to improve the efficiency.
3910 Frequency : MDS failover only, very rarely
3912 Description: "went back in time" messages on mds failover
3913 Details : The greatest transno may be lost when the current operation
3914 finishes with an error (transno==0) and the client's last_rcvd
3915 record is over-written. Save the greatest transno in the
3916 mds_last_transno for this case.
3919 Frequency : always for specific kernels and striping counts
3921 Description: client may get "Matching packet too big" without ACL support
3922 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3923 when trying to access files in certain configurations. The
3924 clients should in fact be denied when mounting because they do
3925 not understand ACLs.
3928 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3930 Description: many clients connecting with IO in progress causes connect timeouts
3931 Details : Avoid synchronous journal commits to avoid delays caused by many
3932 clients connecting/disconnecting when bulk IO is in progress.
3933 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3934 OST_IO_PORTAL to avoid delays behind potentially many pending
3938 Frequency : occasionally with multiple writers to a single file
3940 Description: shared writes to file may result in wrong size reported by stat()
3941 Details : Allow growing of kms when extent lock is cancelled
3944 Frequency : always with random mmap IO to multi-striped file
3946 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3947 Details : Only if the hole is at the end of OST object so that kms is too
3948 small. Fix is to increase kms accordingly in ll_nopage.
3951 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3953 Description: writes to a missing object would leak memory on the OST
3954 Details : If there is an inconsistency between the MDS and OST filesystems,
3955 such that the MDS references an object that doesn't exist, writes
3956 to that object will leak memory due to incorrect cleanup in the
3957 error handling path, eventually running out of memory on the OST.
3962 Description: Creating too long symlink causes lustre errors
3963 Details : Check symlink and name lengths before sending requests to MDS.
3966 Frequency : only if flock is enabled (not on by default)
3968 Description: posix locks not released on fd closure on 2.6.9+
3969 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3970 kernels, this caused such locks not to be released on fd close and
3971 then assertions on fs unmount about still used locks.
3974 Frequency : MDS failover only, very rarely
3976 Description: clients may get ASSERTION(granted_lock != NULL)
3977 Details : When request was taking a long time, and a client was resending
3978 a getattr by name lock request. The were multiple lock requests
3979 with the same client lock handle and
3980 mds_getattr_name->fixup_handle_for_resent_request found one of the
3981 lock handles but later failed with ASSERTION(granted_lock != NULL).
3986 Description: handle->h_buffer_credits > 0, assertion failure
3987 Details : h_buffer_credits is zero after truncate, causing assertion
3988 failure. This patch extends the transaction or creates a new
3992 Frequency : NFS re-export or patchless client
3993 Bugzilla : 11179, 10796
3994 Description: Crash on NFS re-export node (__d_move)
3995 Details : We do not want to hash the dentry if we don't have a lock.
3996 But if this dentry is later used in d_move, we'd hit uninitialised
3997 list head d_hash, so we just do this to init d_hash field but
3998 leave dentry unhashed.
4001 Frequency : NFS re-export or patchless client
4003 Description: NFS exports has problem with symbolic link
4004 Details : lustre client didn't properly install dentry when re-exported
4005 to NFS or running patchless client.
4008 Frequency : NFS re-export or patchless client
4010 Description: Various nfs/patchless fixes.
4011 Details : fixes reuse disconected alias for lookup process - this fixes
4012 warning "find_exported_dentry: npd != pd",
4013 fix permission error with open files at nfs.
4014 fix apply umask when do revalidate.
4017 Frequency : occasional
4019 Description: Crash on NFS re-export node
4020 Details : calling clear_page() on the wrong pointer triggered oops in
4021 generic_mapping_read().
4024 Frequency : rarely, using O_DIRECT IO
4026 Description: unaligned directio crashes client with LASSERT
4027 Details : check for unaligned buffers before trying any requests.
4030 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
4032 Description: stale data returned from RAID cache
4033 Details : If only a small amount of IO is done to the RAID device before
4034 reading it again it is possible to get stale data from the RAID
4035 cache instead of reading it from disk.
4038 Frequency : always for sles10 kernel
4040 Description: sles10 support
4041 Details : ll_follow_link: compile fixes and using of nd_set_link
4042 under newer kernels.
4045 Frequency : depends on arch, kernel and compiler version, always on sles10
4048 Description: recursive or deep enough symlinks cause stack overflow
4049 Details : getting rid of large stack-allocated variable in
4053 Frequency : depends on hardware
4055 Description: lustre write performance loss in the SLES10 kernel
4056 Details : the performance loss is caused by using of write barriers in the
4057 ext3 code. The SLES10 kernel turns barrier support on by
4058 default. The fix is to undo that change for ldiskfs.
4060 ------------------------------------------------------------------------------
4062 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
4064 * Support for kernels:
4065 2.6.9-42.0.3EL (RHEL 4)
4066 2.6.5-7.276 (SLES 9)
4067 2.4.21-47.0.1.EL (RHEL 3)
4068 2.6.12.6 vanilla (kernel.org)
4072 Frequency : quota enabled and large files being deleted
4074 Description: releasing more than 4GB of quota at once hangs OST
4075 Details : If a user deletes more than 4GB of files on a single OST it
4076 will cause the OST to spin in an infinite loop. Release
4077 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4082 Description: statfs data retrieved from /proc may be stale or zero
4083 Details : When reading per-device statfs data from /proc, in the
4084 {kbytes,files}_{total,free,avail} files, it may appear
4085 as zero or be out of date.
4088 Frequency : systems with MD RAID1 external journal devices
4090 Description: lconf's call to blkid is confused by RAID1 journal devices
4091 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4092 of returning all block devices that match the journal UUID.
4095 Frequency : always, for aggregate stripe size over 4GB
4097 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
4098 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
4099 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
4102 Frequency : always if "lfs find" used on a local file/directory
4104 Description: "lfs find" segfaults if used on a local file/directory
4105 Details : The case where a directory component was not specified wasn't
4106 handled correctly. Handle this properly.
4109 Frequency : always on ppc64
4111 Description: the write to an ext3 filesystem mounted with mballoc got stuck
4112 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4113 perform endianness conversion.
4116 Frequency : rarely (truncate to non-zero file size after write under load)
4117 Bugzilla : 10730, 10687
4118 Description: Files padded with zeros to next 4K multiple
4119 Details : With filesystems mounted using the "extents" option (2.6 kernels)
4120 it is possible that files that are truncated to a non-zero size
4121 immediately after being written are filled with zero bytes beyond
4122 the truncated size. No file data is lost.
4124 Severity : enhancement
4126 Description: Allow recovery/failover for liblustre clients.
4127 Details : liblustre clients were unaware of failover configurations until
4130 Severity : enhancement
4132 Description: user file locks should fail when not mounting with flock option
4133 Details : Set up an error-returning stub in ll_file_operations.lock field
4134 to prevent incorrect behaviour when client is mounted without
4135 flock option. Also, set up properly f_op->flock field for
4139 Frequency : always on ia64
4141 Description: "lfs df" loops on printing out MDS statfs information
4142 Details : The obd_ioctl_data was not initialized and in some systems
4143 this caused a failure during the ioctl that did not return
4144 an error. Initialize the struct and return an error on failure.
4147 Frequency : SLES 9 only
4149 Description: Error of copying files with lustre special EAs as root
4150 Details : Client side always return success for setxattr call for lustre
4151 special xattr (currently only "trusted.lov").
4154 Frequency : rarely on clusters with both ia64+i386 clients
4156 Description: ia64+i686 clients doing shared IO on the same file may LBUG
4157 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
4158 clients are doing concurrent writes to the same file it is
4159 possible that the ia64 clients may LASSERT because the OST
4160 extent locks are not PAGE_SIZE aligned. Ensure that grown
4161 locks are always aligned on the request boundary.
4164 Frequency : specific use, occasional
4166 Description: Overwriting in use executable truncates on-disk binary image
4167 Details : If one node attempts to overwrite an executable in use by
4168 another node, we now correctly return ETXTBSY instead of
4169 truncating the file.
4171 Severity : enhancement
4173 Description: Async OSC create to avoid the blocking unnecessarily.
4174 Details : If an OST has no remaining object, system will block on the creating
4175 when need to create a new object on this OST. Now, ways use
4176 pre-created objects when available, instead of blocking on an
4177 empty osc while others are not empty. If we must block, we block
4178 for the shortest possible period of time.
4183 Description: chmod on Lustre root is propagated to other clients
4184 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
4185 invalid by the follow_mount time.
4188 Frequency : liblustre clients only
4190 Description: Race in 'instant cancel' lock handling could lead to such locks
4191 never to be granted in case of SMP MDS
4192 Details : Do not destroy not yet granted but cbpending locks in
4196 Frequency : replay/resend of open
4198 Description: non null lock assetion failure in mds_intent_policy
4199 Details : Trying to replay/resend lockless open requests resulted in
4200 mds_open() returning 0 with no lock. Now it sets a flag if
4201 a lock is going to be returned.
4203 Severity : enhancement
4205 Description: Checksum enhancements
4206 Details : New checksum enhancements allow for resending RPCs that failed
4209 Severity : enhancement
4211 Description: Tunables on number of dirty pages in cacche
4212 Details : Allow to set limit on number of dirty pages cached.
4217 Description: client crash on unmount - lock still has references
4218 Details : In some error handling cases it was possible to leak a lock
4219 reference on a client while accessing a file. This was not
4220 harmful to the client during operation, but would cause the
4221 client to crash when the filesystem is unmounted.
4224 Frequency : specific case, rare
4226 Description: ETXTBSY on mds though file not in use by client
4227 Details : ETXTBSY is no longer incorrectly returned when attempting to
4228 chmod or chown a directory that the user previously tried to
4229 execute or a currently-executing binary.
4232 Frequency : extremely rare except on liblustre-based clients
4234 Description: Lustre space not freed when files are deleted
4235 Details : Clean up open-unlinked files after client eviction. Previously
4236 the unlink was skipped and the files remained as orphans.
4241 Description: OST failure "would be an LBUG" in waiting_locks_callback()
4242 Details : In some cases it was possible to send a blocking callback to
4243 a client doing a glimpse, even though that client didn't get
4244 a lock granted. When the glimpse lock is cancelled on the OST
4245 the freed lock is left on the waiting list and corrupted the list.
4248 Frequency : all core dumps
4250 Description: Broke core dumps to lustre
4251 Details : Negative dentry may be unhashed if parent does not have UPDATE
4252 lock, but some callers, e.g. do_coredump, expect dentry to be
4253 hashed after successful create, hash it in ll_create_it.
4255 ------------------------------------------------------------------------------
4257 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
4259 * Support for kernels:
4260 2.6.9-42.0.2.EL (RHEL 4)
4261 2.6.5-7.276 (SLES 9)
4262 2.4.21-40.EL (RHEL 3)
4263 2.6.12.6 vanilla (kernel.org)
4267 Frequency : always on RHEL 3
4269 Description: Number of open files grows over time
4270 Details : The number of open files grows over time, whether or not
4271 Lustre is started. This was due to a filp leak introduced
4272 by one of our kernel patches.
4274 ------------------------------------------------------------------------------
4276 08-20-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4278 * Support for kernels:
4279 2.6.9-42.EL (RHEL 4)
4280 2.6.5-7.276 (SLES 9)
4281 2.4.21-40.EL (RHEL 3)
4282 2.6.12.6 vanilla (kernel.org)
4287 Bugzilla : 5719, 9635, 9792, 9684
4288 Description: OST (or MDS) trips assertions in (re)connection under heavy load
4289 Details : If a server is under heavy load and cannot reply to new
4290 connection requests before the client resends the (re)connect,
4291 the connection handling code can behave badly if two service
4292 threads are concurrently handing separate (re)connections from
4293 the same client. Add better locking to the connection handling
4294 code, and ensure that only a single connection will be processed
4295 for a given client UUID, even if the lock is dropped.
4297 Severity : enhancement
4299 Description: add TCP zero-copy support to kernel
4300 Details : Add support to the kernel TCP stack to allow zero-copy bulk
4301 sends if the hardware supports scatter-gather and checksumming.
4302 This allows socklnd to do client-write and server-read more
4303 efficiently and reduce CPU utilization from skbuf copying.
4306 Frequency : only if NFS exporting from client
4308 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
4309 Details : If setting attributes on a file created under NFS that had
4310 never been opened it would be possible to oops the client
4311 if the file had no objects.
4314 Frequency : always for liblustre
4316 Description: liblustre client does MDS+OSTs setattr RPC for each write
4317 Details : When doing a write from a liblustre client, the client
4318 incorrectly issued an RPC to the MDS and each OST the file was
4319 striped over in order to update the timestamps. When writing
4320 with small chunks and many clients this could overwhelm the MDS
4321 with RPCs. In all cases it would slow down the write because
4322 these RPCs are unnecessary.
4324 Severity : enhancement
4326 Description: allow number of MDS service threads to be changed at module load
4327 Details : It is now possible to change the number of MDS service threads
4328 running. Adding "options mds mds_num_threads={N}" to the MDS's
4329 /etc/modprobe.conf will set the number of threads for the next
4330 time Lustre is restarted (assuming the "mds" module is also
4331 reloaded at that time). The default number of threads will
4332 stay the same, 32 for most systems.
4337 Description: OST crash if filesystem is unformatted or corrupt
4338 Details : If an OST is started on a device that has never been formatted
4339 or if the filesystem is corrupt and cannot even mount then the
4340 error handling cleanup routines would dereference a NULL pointer.
4345 Description: NULL pointer deref in llap_from_page.
4346 Details : get_cache_page_nowait can return a page with NULL (or otherwise
4347 incorrect) mapping if the page was truncated/reclaimed while it was
4348 searched for. Check for this condition and skip such pages when
4349 doing readahead. Introduce extra check to llap_from_page() to
4350 verify page->mapping->host is non-NULL (so page is not anonymous).
4353 Frequency : Sometimes when using sys_sendfile
4355 Description: "page not covered by a lock" warnings from ll_readpage
4356 Details : sendfile called ll_readpage without right page locks present.
4357 Now we introduced ll_file_sendfile that does necessary locking
4358 around call to generic_file_sendfile() much like we do in
4362 Frequency : with certain MDS communication failures at client mount time
4364 Description: NULL pointer deref after failed client mount
4365 Details : a client connection request may delayed by the network layer
4366 and not be sent until after the PTLRPC layer has timed out the
4367 request. If the client fails the mount immediately it will try
4368 to clean up before the network times out the request. Add a
4369 reference from the request import to the obd device and delay
4370 the cleanup until the network drops the request.
4373 Frequency : occasionally during client (re)connect
4375 Description: assertion failure during client (re)connect
4376 Details : processing a client connection request may be delayed by the
4377 client or server longer than the client connect timeout. This
4378 causes the client to resend the connection request. If the
4379 original connection request is replied in this interval, the
4380 client may trip an assertion failure in ptlrpc_connect_interpret()
4381 which thought it would be the only running connect process.
4384 Frequency : only with obd_echo servers and clients that are rebooted
4386 Description: kernel BUG accessing uninitialized data structure
4387 Details : When running an obd_echo server it did not start the ping_evictor
4388 thread, and when a client was evicted an uninitialized data
4389 structure was accessed. Start the ping_evictor in the RPC
4390 service startup instead of the OBD startup.
4392 Severity : enhancement
4393 Bugzilla : 10193 (patchless)
4394 Description: Remove dependency on various unexported kernel interfaces.
4395 Details : No longer need reparent_to_init, exit_mm, exit_files,
4396 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
4399 Frequency : rare (only users of deprecated and unsupported LDAP config)
4401 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
4402 Details : LDAP apparently contains 'lustreName' attributes instead of
4403 'name'. A simple remapping of the name is sufficient.
4406 Frequency : rare (only with non-default dump_on_timeout debug enabled)
4408 Description: waiting_locks_callback trips kernel BUG if client is evicted
4409 Details : Running with the dump_on_timeout debug flag turned on makes
4410 it possible that the waiting_locks_callback() can try to dump
4411 the Lustre kernel debug logs from an interrupt handler. Defer
4412 this log dumping to the expired_lock_main() thread.
4414 Severity : enhancement
4416 Description: Support NFS exporting on 2.6 kernels.
4417 Details : Implement non-rawops metadata methods for NFS server to use without
4418 changing NFS server code.
4421 Frequency : very rare (synthetic metadata workload only)
4423 Description: two racing renames might cause an MDS thread to deadlock
4424 Details : Running the "racer" program may cause one MDS thread to rename
4425 a file from being the source of a rename to being the target of
4426 a rename at exactly the same time that another thread is doing
4427 so, and the second thread has already enqueued these locks after
4428 doing a lookup of the target and is trying to relock them in
4429 order. Ensure that we don't try to re-lock the same resource.
4432 Frequency : only very large systems with liblustre clients
4434 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
4435 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
4436 all OSTs in parallel. This allows the eviction of stale liblustre
4437 clients to proceed much faster than if they were done in series,
4438 and also offers similar improvements for other set_info RPCs.
4443 Description: excessive CPU usage during initial read phase on client
4444 Details : During the initial read phase on a client, it would agressively
4445 retry readahead on the file, consuming too much CPU and impacting
4446 performance (since 1.4.5.8). Improve the readahead algorithm
4447 to avoid this, and also improve some other common cases (read
4448 of small files in particular, where "small" is files smaller than
4449 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
4454 Description: MDS crash when receiving packet with unknown intent.
4455 Details : Do not LBUG in unknown intent case, just return -EFAULT
4457 Severity : enhancement
4458 Bugzilla : 9293, 9385
4459 Description: MDS RPCs are serialised on client. This is unnecessary for some.
4460 Details : Do not serialize getattr (non-intent version) and statfs.
4463 Frequency : occasional, when OST network is overloaded/intermittent
4465 Description: client evicted by OST after bulk IO timeout
4466 Details : If a client sends a bulk IO request (read or write) the OST
4467 may evict the client if it is unresposive to its data GET/PUT
4468 request. This is incorrect if the network is overloaded (takes
4469 too long to transfer the RPC data) or dropped the OST GET/PUT
4470 request. There is no need to evict the client at all, since
4471 the pinger and/or lock callbacks will handle this, and the
4472 client can restart the bulk request.
4475 Frequency : Always when mmapping file with no objects
4477 Description: client crashes when mmapping file with no objects
4478 Details : Check that we actually have objects in a file before doing any
4479 operations on objects in ll_vm_open, ll_vm_close and
4485 Description: Request leak when working with deleted CWD
4486 Details : Introduce advanced request refcount tracking for requests
4487 referenced from lustre intent.
4489 Severity : Enhancement
4491 Description: Cache open file handles on client.
4492 Details : MDS now will return special lock along with openhandle, if
4493 requested and client is allowed to hold openhandle, even if unused,
4494 until such a lock is revoked. Helps NFS a lot, since NFS is opening
4495 closing files for every read/write openration.
4497 Severity : Enhancement
4499 Description: Cache open negative dentries on client when possible.
4500 Details : Guard negative dentries with UPDATE lock on parent dir, drop
4501 negative dentries on lock revocation.
4506 Description: Remounting a client read-only wasn't possible with a zconf mount
4507 Details : It wasn't possible to remount a client read-only with llmount.
4509 Severity : enhancement
4510 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
4511 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
4512 MPICH is included to add Lustre-specific ADIO interfaces.
4513 This is based closely on the UFS ADIO layer and only differs
4514 in file creation, in order to allow the OST striping to be set.
4515 This is user-contributed code and not supported by CFS.
4520 Description: extended inode attributes (immutable, append-only) work improperly
4521 when 2.4 and 2.6 kernels are used on client/server or vice versa
4522 Details : Introduce kernel-independent values for these flags.
4524 Severity : enhancement
4527 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
4528 Details : Many of the /proc/ tunables can only be tuned at a megabyte
4529 granularity. Now, Fractional MB granularity is be supported,
4530 this is very useful for low memory system.
4532 Severity : enhancement
4534 Description: Getattr by fid
4535 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
4536 locks, avoids extra getattr rpc requests to MDS, allows '/' to
4537 have locks and avoids getattr rpc requests for it on every stat.
4540 Frequency : Always, for filesystems larger than 2TB
4542 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4543 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4544 filesystems larger than 2TB to be mounted with mballoc enabled.
4547 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4549 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
4550 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4551 filesystems larger than 2TB is mounted then the kernel will
4552 silently corrupt the start of the filesystem. CONFIG_LBD is
4553 enabled for all CFS-supported kernels, but the possibility of
4554 this happening with a modified kernel config exists.
4556 Severity : enhancement
4558 Description: add client O_DIRECT support for 2.6 kernels
4559 Details : It is now possible to do O_DIRECT reads and writes to files
4560 in the Lustre client mountpoint on 2.6 kernel clients.
4562 Severity : enhancement
4564 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4565 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4566 parallel, not waiting for response from every OST before sending
4567 a rpc to the next OST.
4572 Description: setattr vs write race when updating file timestamps
4573 Details : Client processes that update a file timestamp into the past
4574 right after writing to the file (e.g. tar) it is possible that
4575 the updated file modification time can be reset to the current
4576 time due to a race between processing the setattr and write RPC.
4578 Severity : enhancement
4580 Description: Bring 'lfs find' closer in line with regular Linux find.
4581 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4582 -print0 options and obtains all the needed info through the lustre
4585 Severity : enhancement
4587 Description: support up to 1024 configured devices on one node
4588 Details : change obd_dev array from statically allocated to dynamically
4589 allocated structs as they are first used to reduce memory usage
4594 Description: Flush dirty partially truncated pages during truncate
4595 Details : Immediatelly flush partially truncated pages in filter_setattr,
4596 this way we completely avoid having any pages in page cache on OST
4597 and can retire ugly workarounds during writes to flush such pages.
4602 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4603 Details : protect inode from truncation within vfs_unlink() context
4604 just take a reference before calling vfs_unlink() and release it
4605 when parent's i_sem is free.
4610 Description: last_id value checked outside lock on OST caused LASSERT failure
4611 Details : If there were multiple MDS->OST object precreate requests in
4612 flight, it was possible that the OST's last object id was checked
4613 outside a lock and incorrectly tripped an assertion. Move checks
4614 inside locks, and discard old precreate requests.
4617 Frequency : always, if extents are used on OSTs
4619 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4620 Details : OSTs using the extents format would not zero the high 16 bits of
4621 the index physical block number. This is not a problem for any
4622 OST filesystems smaller than 16TB, and no kernels support ext3
4623 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4624 new/modified files) and can be fixed for existing filesystems
4625 with e2fsprogs-1.39-cfs1.
4630 Description: import connection selection may be incorrect if timer wraps
4631 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4632 import connections to be ignored if the 32-bit jiffies counter
4633 wraps. Use a 64-bit jiffies counter.
4636 Frequency : very large clusters immediately after boot
4638 Description: LNET request buffers exhausted under heavy short-term load
4639 Details : If a large number of client requests are generated on a service
4640 that has previously never seen so many requests it is possible
4641 that the request buffer growth cannot keep up with the spike in
4642 demand. Instead of dropping incoming requests, they are held in
4643 the LND until the RPC service can accept more requests.
4646 Frequency : Sometimes during replay
4648 Description: Assertion failure in ll_local_open after replay.
4649 Details : If replay happened on an open request reply before we were able
4650 to set replay handler, reply will become not swabbed tripping the
4651 assertion in ll_local_open. Now we set the handler right after
4652 recognising of open request
4655 Frequency : very rare
4657 Description: kernel reports "badness in vsnprintf"
4658 Details : Reading from the "recovery_status" /proc file in small chunks
4659 may cause a negative length in lprocfs_obd_rd_recovery_status()
4660 call to vsnprintf() (which is otherwise harmless). Exit early
4661 if there is no more space in the output buffer.
4663 Severity : enhancement
4665 Description: clear OBD RPC statistics by writing to them
4666 Details : It is now possible to clear the OBD RPC statistics by writing
4667 to the "stats" file.
4672 Description: Client mtime is not the same on different clients after utimes
4673 Details : In some cases, the client was using the utimes() syscall on
4674 a file cached on another node. The clients now validate the
4675 ctime from the MDS + OSTs to determine which one is right.
4680 Description: Inability to activate failout mode
4681 Details : lconf script incorrectly assumed that in pythong string's numeric
4682 value is used in comparisons.
4685 Frequency : always with multiple stripes per file
4687 Description: Inefficient object allocation for mutli-stripe files
4688 Details : When selecting which OSTs to stripe files over, for files with
4689 a stripe count that divides evenly into the number of OSTs,
4690 the MDS is always picking the same starting OST for each file.
4691 Return the OST selection heuristic to the original design.
4696 Description: mount failures may take full timeout to return an error
4697 Details : Under some heavy load conditions it is possible that a
4698 failed mount can wait for the full obd_timeout interval,
4699 possibly several minutes, before reporting an error.
4700 Instead return an error as soon as the status is known.
4702 Frequency : quota enabled and large files being deleted
4704 Description: releasing more than 4GB of quota at once hangs OST
4705 Details : If a user deletes more than 4GB of files on a single OST it
4706 will cause the OST to spin in an infinite loop. Release
4707 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
4712 Description: statfs data retrieved from /proc may be stale or zero
4713 Details : When reading per-device statfs data from /proc, in the
4714 {kbytes,files}_{total,free,avail} files, it may appear
4715 as zero or be out of date.
4718 Frequency : systems with MD RAID1 external journal devices
4720 Description: lconf's call to blkid is confused by RAID1 journal devices
4721 Details : Use the "blkid -l" flag to locate the MD RAID device instead
4722 of returning all block devices that match the journal UUID.
4725 Frequency : always, for aggregate stripe size over 4GB
4727 Description: assertion fails when trying to use 4GB stripe size
4728 Details : Use "setstripe" to set stripe size over 4GB will fail the kernel,
4729 complaining "ASSERTION(lsm->lsm_xfersize != 0)"
4732 Frequency : always on ppc64
4734 Description: the first write on an ext3 filesystem with mballoc got stuck
4735 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
4736 perform endianness conversion.
4738 ------------------------------------------------------------------------------
4740 02-14-2006 Cluster File Systems, Inc. <info@clusterfs.com>
4742 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4743 INTEROPERATE with older versions automatically. Please read the
4744 user documentation before upgrading any part of a live system.
4745 * WARNING: Lustre networking configuration changes are required with
4746 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4749 * Support for kernels:
4750 2.6.9-22.0.2.EL (RHEL 4)
4751 2.6.5-7.244 (SLES 9)
4752 2.6.12.6 vanilla (kernel.org)
4755 Severity : enhancement
4756 Bugzilla : 7981/8208
4757 Description: Introduced Lustre Networking (LNET)
4758 Details : LNET is new networking infrastructure for Lustre, it includes
4759 a reorganized network configuration mode (see the user
4760 documentation for full details) as well as support for routing
4761 between different network fabrics. Lustre Networking Devices
4762 (LNDs) for the supported network fabrics have also been
4763 created for this new infrastructure.
4765 Severity : enhancement
4766 Description: Introduced Access control lists
4767 Details : clients can set ACLs on files and directories in order to have
4768 more fine-grained permissions than the standard Unix UGO+RWX.
4769 The MDS must be started with the "-o acl" mount option.
4771 Severity : enhancement
4772 Description: Introduced filesystem quotas
4773 Details : Administrators may now establish per-user quotas on the
4776 Severity : enhancement
4778 Description: Configuration change for the XT3
4779 The PTLLND is now used to run Lustre over Portals on the XT3
4780 The configure option(s) --with-cray-portals are no longer used.
4781 Rather --with-portals=<path-to-portals-includes> is used to
4782 enable building on the XT3. In addition to enable XT3 specific
4783 features the option --enable-cray-xt3 must be used.
4788 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4789 Details : A race between allocating a new llcd and re-getting the llcd_lock
4790 allowed another thread to grab newly-allocated llcd.
4792 Severity : enhancement
4794 Description: 2.6 OST async journal commit and locking fix to improve performance
4795 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4796 2.6 kernels are now async as they already were in 2.4 kernels so
4797 that they can commit concurrently with the network bulk transfer.
4798 For block-allocated files the filter allocation semaphore is held
4799 to avoid filesystem fragmentation during allocation. BKL lock
4800 removed for 2.6 xattr operations where it is no longer needed.
4805 Description: lconf incorrectly determined whether two IP networks could talk
4806 Details : In some more complicated routing and multiple-network
4807 configurations, lconf will avoid trying to make a network
4808 connection to a disjoint part of the IP space. It was doing the
4809 math incorrectly for one set of cases.
4814 Description: Fix for potential infinite loop processing records in an llog.
4815 Details : If an llog record is corrupted/zeroed, it is possible to loop
4816 forever in llog_process(). Validate the llog record length
4817 and skip the remainder of the block on error.
4820 Frequency : occasional (liblustre only)
4822 Description: liblustre could not open files whose last component is a symlink
4823 Details : sysio_path_walk() would incorrectly pass the open intent to
4824 intermediate path components.
4827 Frequency : rare (liblustre only with non-standard tuning)
4828 Bugzilla : 7201 (7350)
4829 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4830 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4831 after it has been released, only do this for referenced locks
4833 Severity : enhancement
4835 Description: specifying an (invalid) directory default stripe_size of -1
4836 would reset the directory default striping
4837 Details : stripe_size -1 was used internally to signal directory stripe
4838 removal, now use "all default" to signal dir stripe removal
4839 as a directory striping of "all default" is not useful
4842 Frequency : common for large clusters running liblustre clients
4844 Description: doing an ls when liblustre clients are running is slow
4845 Details : sending a glimpse AST to a liblustre client waits for every AST
4846 to time out, as liblustre clients will not respond. Since they
4847 cannot cache data we refresh the OST lock LVB from disk instead.
4849 Severity : enhancement
4851 Description: doing an ls at the same time as file IO can be slow
4852 Details : enqueue and other "small" requests can be blocked behind many
4853 large IO requests. Create a new OST IO portal for non-IO
4854 requests so they can be processed faster.
4857 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4859 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4860 created file with mode 000
4861 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4862 allow an ftruncate() as a non-root user to a file with mode 000.
4863 Lustre now respects this flag to disable mode checks when
4864 truncating a file owned by the user
4867 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4870 Description: Revoking locks from clients that went dead or catatonic might take
4872 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4873 cancellation of such locks instant on servers without waiting for
4874 any reply from clients. Clients drops these locks when cancel
4875 notification from server is received without replying.
4878 Frequency : liblustre-only, when liblustre client dies or becomes busy
4880 Description: Doing ls on Linux clients can take a long time with active
4882 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4883 granting such locks to it in the first place if possible. Locks
4884 are taken by proxy on the OST during the read or write and
4885 dropped immediately afterward. Add connect flags handling, do
4886 not grant locks to liblustre clients for glimpse ASTs.
4888 Severity : enhancement
4890 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4891 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4892 sequential reads and behaves badly with applications doing random
4893 reads. Improve it by reading ahead at least read region, and
4894 avoiding excessive large RPC for small reads.
4896 Severity : enhancement
4898 Description: Creating more than 1000 files for a single job may cause a load
4899 imbalance on the OSTs if there are also a large number of OSTs.
4900 Details : qos_prep_create() uses an OST index reseed value that is an
4901 even multiple of the number of available OSTs so that if the
4902 reseed happens in the middle of the object allocation it will
4903 still utilize the OSTs as uniformly as possible.
4908 Description: OST or MDS may oops in ping_evictor_main()
4909 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4910 but doesn't restart at beginning of obd_exports_timed list
4913 Severity : enhancement
4915 Description: improve by-nid export eviction on the MDS and OST
4916 Details : allow multiple exports with the same NID to be evicted at one
4917 time without re-searching the exports list.
4920 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4922 Description: MDS may oops in groups_free()
4923 Details : in rare race conditions a newly allocated group_info struct is
4924 freed again, and this can be NULL. The 2.4 compatibility code
4925 for groups_free() checked for a NULL pointer, but 2.6 did not.
4928 Frequency : common for liblustre clients doing little filesystem IO
4929 Bugzilla : 9352, 7313
4930 Description: server may evict liblustre clients accessing contended locks
4931 Details : if a client is granted a lock or receives a completion AST
4932 with a blocking AST already set it would not reply to the AST
4933 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4936 Frequency : lfs setstripe, only systems with more than 160 OSTs
4938 Description: unable to set striping with a starting offset beyond OST 160
4939 Details : llapi_create_file() incorrectly limited the starting stripe
4940 index to the maximum single-file stripe count.
4943 Frequency : LDAP users only
4945 Description: lconf did not handle in-kernel recovery with LDAP properly
4946 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4948 Severity : enhancement
4950 Description: bind OST threads to NUMA nodes to improve performance
4951 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4952 node and do their allocations there to localize memory access
4954 Severity : enhancement
4956 Description: llmount can determine client NID directly from Myrinet (GM)
4957 Details : the client NID code from gmnalnid was moved directly into
4958 llmount, removing the need to use this or specifying the
4959 client NID explicitly when mounting GM clients with zeroconf
4962 Frequency : if client is started with down MDS
4964 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4965 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4966 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4967 ptlrpc_import_delayed_req() because we didn't check if the
4968 request was interrupted, and we also didn't break out of the
4969 event loop if there was no timeout
4974 Description: data loss during non-page-aligned writes to a single file from
4975 both multiple nodes and multiple threads on one node at same time
4976 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4977 inconsistency led to false short-reads, that were cached and later
4978 used by ->prepare_write() to fill in partially written page,
4979 leading to data loss.
4982 Frequency : always, if lconf --abort_recovery used
4984 Description: lconf --abort_recovery fails with 'Operation not supported'
4985 Details : lconf was attempting to abort recovery on the MDT device and not
4988 Severity : enhancement
4990 Description: remove cleanup logs
4991 Details : replace lconf-generated cleanup logs with lustre internal
4992 cleanup routines. Eliminates the need for client-cleanup and
4995 Severity : enhancement
4997 Description: add support for EAs (user and system) on lustre filesystems
4998 Details : it is now possible to store extended attributes in the Lustre
4999 client filesystem, and with the user_xattr mount option it
5000 is possible to allow users to store EAs on their files also
5002 Severity : enhancement
5004 Description: Add possibility (config option) to show minimal available OST free
5006 Details : When compiled with --enable-mindf configure option, statfs(2)
5007 (and so, df) will return least minimal free space available from
5008 all OSTs as amount of free space on FS, instead of summary of
5009 free spaces of all OSTs.
5011 Severity : enhancement
5013 Description: do not expand extent locks acquired on OST-side
5014 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
5015 by server: they are not cached anyway.
5018 Frequency : when mmap is used/binaries executed from Lustre
5020 Description: Unmmap pages before throwing them away from read cache.
5021 Details : llap_shrink cache now attempts to unmap pages before discarding
5022 them (if unmapping failed - do not discard). SLES9 kernel has
5023 extra checks that trigger if this unmapping is not done first.
5028 Description: lconf didn't resolve symlinks before checking to see whether a
5029 given mountpoint was already in use
5032 Frequency : when migrating failover services
5033 Bugzilla : 6395, 9514
5034 Description: When migrating a subset of services from a node (e.g. failback
5035 from a failover service node) the remaining services would
5036 time out and evict clients.
5037 Details : lconf --force (implied by --failover) sets the global obd_timeout
5038 to 5 seconds in order to quickly disconnect, but this caused
5039 other RPCs to time out too quickly. Do not change the global
5040 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
5042 Severity : enhancement
5043 Frequency : if MDS is started with down OST
5044 Bugzilla : 9439,5706
5045 Description: Allow startup/shutdown of an MDS without depending on the
5046 availability of the OSTs.
5047 Details : Asynchronously call mds_lov_synchronize during MDS startup.
5048 Add appropriate locking and lov-osc refcounts for safe
5049 cleaning. Add osc abort_inflight calls in case the
5050 synchronize never started.
5053 Frequency : occasional (Cray XT3 only)
5055 Description: root not authorized to access files in CRAY_PORTALS environment
5056 Details : The client process capabilities were not honoured on the MDS in
5057 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
5058 been accessed by an authorized user then root was able to access
5059 the file on the local client also. The root user capabilities
5060 are now allowed on the MDS, as this environment has secure UID.
5063 Frequency : occasional
5065 Description: ldiskfs "too long searching" message happens too often
5066 Details : A debugging message (otherwise harmless) prints too often on
5067 the OST console. This has been reduced to only happen when
5068 there are fragmentation problems on the filesystem.
5073 Description: Division by zero in statfs when all OSCs are inactive
5074 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
5075 lov_statfs divides by value returned by lov_get_stripecnt().
5079 Bugzilla : 9489, 3273
5080 Description: First write from each client to each OST was only 4kB in size,
5081 to initialize client writeback cache, which caused sub-optimal
5082 RPCs and poor layout on disk for the first written file.
5083 Details : Clients now request an initial cache grant at (re)connect time
5084 and so that they can start streaming writes to the cache right
5085 away and always do full-sized RPCs if there is enough data.
5086 If the OST is rebooted the client also re-establishes its grant
5087 so that client cached writes will be honoured under the grant.
5092 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
5093 Details : Now I/O RPCs go to different portal number and (presumably) fast
5094 lock requests (and glimses) and other RPCs get their own service
5095 threads pool that should be able to service those RPCs
5098 Severity : enhancement
5100 Description: Ability to exchange lustre version between client and servers and
5101 issue warnings at client side if client is too old. Also for
5102 liblustre clients there is ability to refuse connection of too old
5104 Details : New 'version' field is added to connect data structure that is
5105 filled with version info. That info is later checked by server and
5109 Frequency : rare, liblustre only.
5110 Bugzilla : 9296, 9581
5111 Description: Two simultaneous writes from liblustre at offset within same page
5112 might proceed at the same time overwriting eachother with stale
5114 Details : I/O lock withing llu_file_prwv was released too early, before data
5115 actually was hitting the wire. Extended lock-holding time until
5116 server acknowledges receiving data.
5119 Frequency : extremely rare. Never observed in practice.
5121 Description: avoid generating lustre_handle cookie of 0.
5122 Details : class_handle_hash() generates handle cookies by incrementing
5123 global counter, and can hit 0 occasionaly (this is unlikely, but
5124 not impossible, because initial value of cookie counter is
5125 selected randonly). Value of 0 is used as a sentinel meaning
5126 "unassigned handle" --- avoid it. Also coalesce two critical
5127 sections in this function into one.
5129 Severity : enhancement
5131 Description: allow liblustre clients to delegate truncate locking to OST
5132 Details : To avoid overhead of locking, liblustre client instructs OST to
5133 take extent lock in ost_punch() on client's behalf. New connection
5134 flag is added to handle backward compatibility.
5136 Severity : enhancement
5137 Bugzilla : 4928, 7341, 9758
5138 Description: allow number of OST service threads to be specified
5139 Details : a module parameter allows the number of OST service threads
5140 to be specified via "options ost ost_num_threads={N}" in the
5141 OSS's /etc/modules.conf or /etc/modprobe.conf.
5145 Bugzilla : 6146, 9635, 9895
5146 Description: servers crash with bad pointer in target_handle_connect()
5147 Details : In rare cases when a client is reconnecting it was possible that
5148 the connection request was the last reference for that export.
5149 We would temporarily drop the export reference and get a new
5150 one, but this may have been the last reference and the export
5151 was just destroyed. Get new reference before dropping old one.
5153 Severity : enhancement
5154 Frequency : if client is started with failover MDS
5156 Description: Allow multiple MDS hostnames in the mount command
5157 Details : Try to read the configuration from all specified MDS
5158 hostnames during a client mount in case the "primary"
5161 Severity : enhancement
5163 Description: Stop sending data to evicted clients as soon as possible.
5164 Details : Check if the client we are about to send or are sending data to
5165 was evicted already. (Check is done every second of waiting,
5166 for which l_wait_event interface was extended to allow checking
5167 of exit condition at specified intervals).
5170 Frequency : rare, normally only when NFS exporting is done from client
5172 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
5174 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
5177 Frequency : always, if srand() is called before liblustre initialization
5179 Description: Liblustre uses system PRNG disturbing its usage by user application
5180 Details : Introduce internal to lustre fast and high-quality PRNG for
5181 lustre usage and make liblustre and some other places in generic
5182 lustre code to use it.
5184 Severity : enhancement
5185 Bugzilla : 9477, 9557, 9870
5186 Description: Verify that the MDS configuration logs are updated when xml is
5187 Details : Check if the .xml configuration logs are newer than the config
5188 logs stored on the MDS and report an error if this is the case.
5189 Request --write-conf, or allow starting with --old_conf.
5191 Severity : enhancement
5193 Description: Handle symlinks in the path when checking if Lustre is mounted.
5194 Details : Resolve intermediate symlinks when checking if a client has
5195 mounted a filesystem to avoid duplicate client mounts.
5200 Description: lconf can hit an error exception but still return success.
5201 Details : The lconf command catches the Command error exception at the top
5202 level script context and will exit with the associated exit
5203 status, but doesn't ensure that this exit status is non-zero.
5208 Description: failure of ptlrpc thread startup can cause oops
5209 Details : Starting a ptlrpc service thread can fail if there are a large
5210 number of threads or the server memory is very fragmented.
5211 Handle this without oopsing.
5214 Frequency : always, only if liblustre and non-default acceptor port was used
5216 Description: liblustre cannot connect to servers with non-default acceptor port
5217 Details : tcpnal_set_default_params() was not called and was therefore
5218 ignoring the environment varaible TCPNAL_PORT, as well as other
5219 TCPNAL_ environment variables
5224 Description: two objects could be created on the same OST for a single file
5225 Details : If an OST is down, in some cases it was possible to create two
5226 objects on a single OST for a single file. No problems other
5227 than potential performance impact and spurious error messages.
5231 Bugzilla : 5681, 9562
5232 Description: Client may oops in ll_unhash_aliases
5233 Details : Client dcache may become inconsistent in race condition.
5234 In some cases "getcwd" can fail if the current directory is
5240 Description: Inode refcounting problems in NFS export code
5241 Details : link_raw functions used to call d_instantiate without obtaining
5242 extra inode reference first.
5246 Bugzilla : 9942, 9903
5247 Description: Referencing freed requests leading to crash, memleaks with NFS.
5248 Details : We used to require that call to ll_revalidate_it was always
5249 followed by ll_lookup_it. Also with revalidate_special() it is
5250 possible to call ll_revalidate_it() twice for the same dentry
5251 even if first occurence returned success. This fix changes semantic
5252 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
5253 reference on a request referred from the intent.
5254 ll_intent_release() then releases such a request.
5257 Frequency : rare, normally benchmark loads only
5259 Description: unlinked inodes were kept in memory on the client
5260 Details : If a client is repeatedly creating and unlinking files it
5261 can accumulate a lot of stale inodes in the inode slab cache.
5262 If there is no other client load running this can cause the
5263 client node to run out of memory. Instead flush old inodes
5264 from client cache that have the same inode number as a new inode.
5267 Frequency : SLES9 2.6.5 kernel and long filenames only
5268 Bugzilla : 9969, 10379
5269 Description: utime reports stale NFS file handle
5270 Details : SLES9 uses out-of-dentry names in some cases, which confused
5271 the lustre dentry revalidation. Change it to always use the
5275 Frequency : rare, unless heavy write-truncate concurrency is continuous
5276 Bugzilla : 4180, 6984, 7171, 9963, 9331
5277 Description: OST becomes very slow and/or deadlocked during object unlink
5278 Details : filter_destroy() was holding onto the parent directory lock
5279 while truncating+unlinking objects. For very large objects this
5280 may block other threads for a long time and slow overall OST
5281 responsiveness. It may also be possible to get a lock ordering
5282 deadlock in this case, or run out of journal credits because of
5283 the combined truncate+unlink. Solution is to do object truncate
5284 first in one transaction without parent lock, and then do the
5285 final unlink in a new transaction with the parent lock. This
5286 reduces the lock hold time dramatically.
5289 Frequency : rare, 2.4 kernels only
5291 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
5292 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
5293 before unmounting their filesystem in order to avoid deadlock.
5294 The kernel_locked() function in 2.4 kernels only checks whether
5295 the kernel lock is held, not whether it is this process that is
5296 holding it as 2.6 kernels do.
5301 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
5302 Details : The client ptlrpc code may be trying to reconnect to a down
5303 server before a previous connection attempt has timed out.
5304 Increase the reconnect interval to be longer than the connection
5305 timeout interval to avoid sending duplicate connections to
5309 Frequency : echo_client brw_test command
5311 Description: fix echo_client to work with OST preallocated code
5312 Details : OST preallocation code (5137) didn't take echo_client IO path
5313 into account: echo_client calls filter methods outside of any
5314 OST thread and, hence, there is no per-thread preallocated
5315 pages and buffers to use. Solution: hijack pga pages for IO. As
5316 a byproduct, this avoids unnecessary data copying.
5320 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
5321 Description: Client can oops in mdc_commit_close() after open replay
5322 Details : It was possible for the MDS to return an open request with no
5323 transaction number in mds_finish_transno() if the client was
5324 evicted, but without actually returning an error. Clients
5325 would later try to replay that open and may trip an assertion
5326 Simplify the client close codepath, and always return an error
5327 from the MDS in case the open is not successful.
5330 Frequency : rare, 2.6 OSTs only
5332 Description: OST may deadlock under high load on fragmented files
5333 Details : If there was a heavy load and highly-fragmented OST filesystems
5334 it was possible to have all the OST threads deadlock waiting on
5335 allocation of biovecs, because the biovecs were not released
5336 until the entire RPC IO was completed. Instead, release biovecs
5337 as soon as they are complete to ensure forward IO progress.
5339 Severity : enhancement
5341 Description: Support for specifying external journal device at mount
5342 Details : If an OST or MDS device is formatted with an external journal
5343 device, this device major/minor is stored in the ext3 superblock
5344 and may not be valid for failover. Allow detecting and
5345 specifying the external journal at mount time.
5350 Description: Mounting an MDS with pending unlinked files may cause oops
5351 Details : target_finish_recovery() calls mds_postrecov() which returned
5352 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
5353 considers this an error and immediately begins cleaning up the
5354 lov, just after starting the mds_lov process
5356 Severity : enhancement
5358 Description: Implement 'lfs df' to report actual free space on per-OST basis
5359 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
5360 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
5361 usage of objects; '-h' to report in human readable format.
5363 ------------------------------------------------------------------------------
5365 08-26-2005 Cluster File Systems, Inc. <info@clusterfs.com>
5372 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
5373 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
5374 at startup during orphan unlinking. Instead, skip buddy update
5375 before setup, as it will be regenerated after recovery is complete.
5380 Description: If an OST is inactive, its locks might reference stale inodes.
5381 Details : lov_change_cbdata() must iterate over all namespaces, even if
5382 they are inactive to clear inode references from the lock.
5384 Severity : enhancement
5385 Frequency : occasional, if non-standard max_dirty_mb used
5387 Description: Client will block write RPCs if not enough grant
5388 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
5389 then the client will block writes while waiting for another RPC
5390 to complete instead of consuming its dirty limit. With change
5391 we get improved performance when max_dirty_mb is small.
5393 Severity : enhancement
5394 Bugzilla : 3389, 6253
5395 Description: Add support for supplementary groups on the MDS.
5396 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
5397 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
5398 lookups for user supplementary groups into a cache.
5402 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
5403 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
5404 for liblustre O_EXCL is enforced by MDS.
5408 Description: readdir never returns NULL in liblustre.
5409 Details : Corrected llu_iop_getdirentries logic, to return offset of next
5410 dentry in struct dirent.
5414 Frequency : liblustre only, depends on application IO pattern
5415 Description: liblustre clients evicted if not contacting servers
5416 Details : Don't put liblustre clients into the ping_evictor list, so
5417 they will not be evicted by the pinger ever.
5419 Severity : enhancement
5421 Description: Add ability to evict clients by NID from MDS.
5422 Details : By echoing "nid:$NID" string into
5423 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
5424 $NID would be instantly evicted from this MDS and from all active
5425 OSTs connected to it.
5429 Description: Do not query file size twice, somewhat slowing stat(2) calls.
5430 Details : lookup_it_finish() used to query file size from OSTs that was not
5435 Description: service threads change working directory to that of init
5436 Details : Starting lustre service threads may pin the working directory
5437 of the parent thread, making that filesystem busy. Threads
5438 now change to the working directory of init to avoid this.
5442 Frequency : during shutdown only
5443 Description: shutdown with a failed MDS or OST can cause unmount to hang
5444 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
5449 Frequency : 2.6 only
5450 Description: chmod/chown may include an extra supplementary group
5451 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
5452 supplementary group and if none is specified this is used.
5455 Bugzilla : 5479 (6816)
5457 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
5458 Details : If lookup is in progress on a file that is unlinked we might try
5459 to revalidate the inode and fail in revalidate after lookup is
5460 complete and ll_file_open() enqueues the open again but
5461 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
5464 Frequency : always, if lconf --abort_recovery used
5466 Description: lconf --abort_recovery fails with 'Operation not supported'
5467 Details : lconf was attempting to abort recovery on the MDT device and not
5470 ------------------------------------------------------------------------------
5472 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
5477 Frequency : rare (only unsupported configurations with a node running as an
5479 Bugzilla : 6514, 5137
5480 Description: Mounting a Lustre file system on a node running as an OST could
5482 Details : OSTs now preallocates memory needed to write out data at
5483 startup, instead of when needed, to avoid having to
5484 allocate memory in possibly low memory situations.
5485 Specifically, if the file system is mounted on on OST,
5486 memory pressure could force it to try to write out data,
5487 which it needed to allocate memory to do. Due to the low
5488 memory, it would be unable to do so and the node would
5489 become unresponsive.
5491 Severity : enhancement
5493 Description: Addition of lconf --service command line option
5494 Details : lconf now accepts a '--service <arg>' option, which is
5495 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
5497 Severity : enhancement
5499 Description: Failover mode is now the default for OSTs.
5500 Details : By default, OSTs will now run in failover mode. To return to
5501 the old behaviour, add '--failout' to the lmc line for OSTs.
5503 Severity : enhancement
5505 Description: Health checks are now provided for MDS and OSTs
5506 Details : Additional detailed health check information on MSD and OSTs
5507 is now provided through the procfs health_check value.
5510 Frequency : occasional, depends on IO load
5512 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
5513 after numerous create/delete cycles
5514 Details : The ext3 inode allocation policy would not allocate new inodes
5515 very well on the OSTs because there are no new directories
5516 being created. Instead we look for groups with free space if
5517 the parent directories are nearly full.
5522 Description: Network or server problems during mount may cause partially
5523 mounted clients instead of returning an error.
5524 Details : The config llog parsing code may overwrite the error return
5525 code during mount error handling, returning success instead
5531 Description: MDS can fail to allocate large reply buffers
5532 Details : After long uptimes the MDS can fail to allocate large reply
5533 buffers (e.g. zconf client mount config records) due to memory
5534 fragmentation or consumption by the buffer cache. Preallocate
5535 some large reply buffers so that these replies can be sent even
5536 under memory pressure.
5540 Frequency : rare (liblustre)
5541 Description: fsx running with liblustre complained that using truncate() to
5542 extend the file doesn't work. This patch corrects that issue.
5543 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
5544 ATTR_SIZE and lsm use in llu_setattr_raw.
5548 Frequency : rare, only 2.6 kernels
5549 Description: Unusual file access patterns on the MDS may result in inode
5550 data being lost in very rare circumstances.
5551 Details : Bad interaction between the ea-in-inode patch and the "no-read"
5552 code in the 2.6 kernel caused the inode and/or EA data not to
5553 be read from disk, causing single-file corruption.
5557 Frequency : rare, only 2.6 filesystems using extents
5558 Description: Heavy concurrent write and delete load may cause data corruption.
5559 Details : It was possible under high-load situations to have an extent
5560 metadata block in the block device cache from a just-unlinked
5561 file overwrite a newly-allocated data block. We now unmap any
5562 metadata buffers that alias just-allocated data blocks.
5566 Frequency : filesystems with default stripe_count larger than 77
5567 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
5568 Details : lconf specifies an inode size of 4096 bytes when the default
5569 stripe_count is larger than 77. This conflicts with the default
5570 inode density of 1 per 4096 bytes. Allocate smaller inodes in
5571 this case to avoid pinning too much memory for large EAs.
5573 ------------------------------------------------------------------------------
5575 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
5580 Frequency : rare (extremely heavy IO load with hundreds of clients)
5582 Description: Client is evicted, gets IO error writing to file
5583 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5584 caused clients to be evicted for AST timeouts. The fixes in
5585 bug 5192 mean we no longer need to have such short AST timeouts
5586 so ldlm_timeout has been increased.
5589 Frequency : occasional during --force or --failover shutdown under load
5590 Bugzilla : 5949, 4834
5591 Description: Server oops/LBUG if stopped with --force or --failover under load
5592 Details : a collection of import/export refcount and cleanup ordering
5593 issues fixed for safer force cleanup
5596 Frequency : only filesystems larger than 120 OSTs
5597 Bugzilla : 5990, 6223
5598 Description: lfs getstripe would oops on a very large filesystem
5599 Details : lov_getconfig used kfree on vmalloc'd memory
5602 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5603 Bugzilla : 6242, 6243
5604 Description: reading from files that had been truncated to a non-zero size
5605 but never opened returned no data
5606 Details : ll_file_read() reads zeros from no-object files to EOF
5611 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5612 Details : There's little harm in aborting MDS/OSS recovery and letting it
5613 try again, so I removed the LASSERT and return an error instead.
5615 Severity : enhancement
5617 Description: New debugging infrastructure for tracking down data corruption
5618 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5619 (b) cover more of the client-side code path, and (c) try to narrow
5620 down where problems occurred
5624 Bugzilla : 3819, 4364, 4397, 6313
5625 Description: Racing close and eviction MDS could cause assertion in mds_close
5626 Details : It was possible to get multiple mfd references during close and
5627 client eviction, leading to one thread referencing a freed mfd.
5629 Severity: : enhancement
5630 Bugzilla : 3262, 6359
5631 Description: Attempts to reconnect to servers are now more aggressive.
5632 Details : This builds on the enhanced upcall-less recovery that was added
5633 in 1.4.2. When trying to reconnect to servers, clients will
5634 now try each server in the failover group every 10 seconds. By
5635 default, clients would previously try one server every 25 seconds.
5640 Description: After recovery, certain operations trigger a failed
5641 assertion on a client.
5642 Details : Failing over an mds, using lconf -d --failover, while a
5643 client was doing a readdir() call would cause the client to
5644 LBUG after recovery completed and the readdir() was resent.
5646 Severity : enhancement
5648 Description: Default groups are now added by lconf
5649 Details : You can now run lconf --group <servicename> without having to
5650 manually add groups with lmc.
5653 Frequency : occasional
5655 Description: Nodes with an elan id of 0 trigger a failed assertion
5658 Frequency : always when accessing e.g. tty/console device nodes
5660 Description: tty and some other devices nodes cannot be used on lustre
5661 Details : file's private_data field is used by device data and lustre
5662 values in there got lost. New field was added to struct file to
5663 store fs-specific private data.
5666 Frequency : when exporting Lustre via NFS
5668 Description: NFSD failed occasionally when looking up a path component
5669 Details : NFSD is looking up ".." which was broken in ext3 directories
5670 that had grown large enough to become hashed.
5673 Frequency : Clusters with multiple interfaces not on the same subnet
5675 Description: Nodes will repeatedly try to reconnect to an interface which it
5676 cannot reach and report an error to the log.
5677 Details : Extra peer list entries will be created by lconf with some peers
5678 unreachable. lconf now validates the peer before adding it.
5681 Frequency : Only if a default stripe is set on the filesystem root.
5683 Description: Setting a default stripe on the filesystem root prevented the
5684 filesystem from being remounted.
5685 Details : The client was sending extra request flags in the root getattr
5686 request and did not allocate a reply buffer for the dir EA.
5689 Frequency : occasional, higher if lots of files are accessed by one client
5690 Bugzilla : 6159, 6097
5691 Description: Client trips assertion regarding lsm mismatch/magic
5692 Details : While revalidating inodes the VFS looks up inodes with ifind()
5693 and in rare cases can find an inode that is being freed.
5694 The ll_test_inode() code will free the lsm during ifind()
5695 when it finds an existing inode and then the VFS later attaches
5696 this free lsm to a new inode.
5700 Bugzilla : 6422, 7030
5701 Description: MDS deadlock between mkdir and client eviction
5702 Details : Creating a new file via mkdir or mknod (starting a transaction
5703 and getting the ns lock) can deadlock with client eviction
5704 (gets ns lock and trying to finish a synchronous transaction).
5707 Frequency : occasional
5708 Description: While starting a server, the fsfilt_ext3 module could not be
5710 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5711 kernels. Previously, lconf would still use the ext3 name
5712 when trying to load modules. Now, it will correctly use
5713 ext3 on 2.4 and ldiskfs on 2.6.
5715 Severity : enhancement
5716 Description: The default stripe count has been changed to 1
5717 Details : The interpretation of the default stripe count (0, to lfs
5718 or lmc) has been changed to mean striping across a single
5719 OST, rather than all available. For general usage we have
5720 found a stripe count of 1 or 2 works best.
5722 Severity : enhancement
5723 Description: Add support for compiling against Cray portals.
5724 Details : Conditional compiling for some areas that are different
5728 Frequency : occasional
5729 Bugzilla : 6409, 6834
5730 Description: Creating files with an explicit stripe count may lead to
5731 a failed assertion on the MDS
5732 Details : If some OSTs are full or unavailable, creating files may
5733 trigger a failed assertion on the MDS. Now, Lustre will
5734 try to use other servers or return an error to the
5738 Frequency : occasional
5740 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5741 caused lock timeout during readahead (since 1.4.2).
5742 Details : Processes doing readahead might match a lock that hasn't been
5743 granted yet if there are overlapping and conflicting lock
5744 requests. The readahead process waits on ungranted lock
5745 (original lock is CBPENDING), while OST waits for that process
5746 to cancel CBPENDING read lock and eventually evicts client.
5748 Severity : enhancement
5750 Description: Initial enabling of flock support for clients
5751 Details : Implements fcntl advisory locking and file status functions.
5752 This feature is provided as an optional mount flag (default
5753 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5754 locking are implemented yet, and those that are are not guaranteed
5755 to be completely correct in production environments.
5756 mount -t lustre -o [flock|noflock] ...
5759 Frequency : occasional
5761 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5762 assertion in the ext3 JBD (journaling) layer.
5763 Details : The b_committed_data struct is protected by the big kernel lock
5764 in 2.4 kernels, serializing journal_commit_transaction() and
5765 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5766 access to this struct. In 2.6 kernels there is finer grained
5767 locking to improve SMP performance of the JBD layer.
5771 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5776 Description: merge and cleanup kernel patches.
5779 -----------------------------------------------------------------------------
5781 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5783 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5784 versions of Lustre. Please update all servers and clients to
5785 version 1.4.2 or later at the same time. You must also run
5786 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5787 to update the configuration logs.
5789 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5790 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5791 - swab llog records of type '0' so we get proper header size/idx (5861)
5792 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5793 - fix rename of one directory over another leaking an inode (5953)
5794 - avoid SetPageDirty on 2.6 (5981)
5795 - don't re-add just-being-destroyed locks to the waiting list (5653)
5796 - when creating new directories, inherit the parent's custom
5797 striping settings if present parent (3048)
5798 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5799 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5800 - don't hold client locks on temporary worklist from l_lru (5666)
5801 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5802 - automatically evict dead clients (5921)
5803 - Update file size properly in create+truncate+fstat case (6196)
5804 - Do not unhash mountpoint dentries, do not allow removal of
5806 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5807 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5808 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5809 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5810 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5811 - Correctly handle reads of files with no objects (6243)
5812 - lctl recover will also mark a device active if deactivate used (5933)
5814 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5815 - allow --write-conf on an MDS with different nettype than client (5619)
5816 - don't write config llogs to MDS for mounts not from that MDS (5617)
5817 - lconf should create multiple TCP connections from a client (5201)
5818 - init scripts are now turned off by default; run chkconfig --on
5819 lustre and chkconfig --on lustrefs to use them
5820 - upcalls are no longer needed for clients to recover to failover
5822 - add --abort-recovery option to lconf to abort recovery on device
5824 - add support for an arbitrary number of OSTs (3026)
5825 - Quota support protocol changes.
5826 - forward compatibility changes to wire structs (6007)
5827 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5828 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5829 - improved "lustre status" script
5830 - initialize blocksize for non-regular files (6062)
5831 - added --disable-server and --disable-client configure options (5782)
5832 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5833 - Vanilla 2.4.29 support
5834 - increase maximum number of obd devices to 520 (6242)
5835 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5836 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5837 SLES 9 2.6.5 kernels
5838 - limit stripes per file to 160 (the maximum EA size) (6093)
5840 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5843 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5844 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5845 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5846 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5847 - small ext3 extents cleanups and fixes (5733)
5848 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5849 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5850 - avoid needless client->OST connect, fix handle mismatch (5317)
5851 - fix DLM error path that led to out-of-sync client, long delays (5779)
5852 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5853 - fix several locking issues related to i_size (5492,5624,5654,5672)
5854 - don't move pending lock onto export if it is already evicted (5683)
5855 - fix kernel oops when creating .foo in unlinked directory (5548)
5856 - fix deadlock in obdfilter statistics vs. object create (5811)
5857 - use time_{before,after} to avoid timer jiffies wrap (5882)
5858 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5859 - Do not leak request if server was not able to process it (5154)
5860 - If mds_open unable to find parent dir, make that negative lookup(5154)
5861 - don't create new directories with extent-mapping (5909, 5936)
5863 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5864 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5865 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5866 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5867 - support for RHEL4 kernel on i686 (5773)
5868 - provide error messages when incompatible logs are encountered (5898)
5870 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5871 * version 1.4.0.10 (1.4.1 release candidate 1)
5873 - don't keep a lock reference when lock is not granted (4238)
5874 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5875 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5876 - drop import inflight refcount on signal_completed_replay error (5255)
5877 - unlock page after async write error during send (3677)
5878 - handle missing objects in filter_preprw_read properly (5265)
5879 - no transno return for symlink open, don't save no-trasno open (3440)
5880 - don't try to complete elan receive that already failed (4012)
5881 - free RPC server reply state on error (5406)
5882 - clean up thread from ptlrpc_start_thread() on error (5160)
5883 - readahead could read extra page into cache that wasn't ejected (5388)
5884 - prevent races in class_attach/setup/cleanup/detach (5260)
5885 - don't dereference de->d_inode after l_dput of de (5458)
5886 - use "int" for stripe value returned from lock_to_stripe (5544)
5887 - mballoc allocation and error-checking fixes in 2.6 (5504)
5888 - block device patches to fix I/O request sizes in 2.6 (5482)
5889 - look up hostnames for IB nals (5602)
5890 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5891 - don't start multiple acceptors for the same port (5277)
5892 - fix incorrect LASSERT in mds_getattr_name (5635)
5893 - export a proc file for general "ping" checking (5628)
5894 - fix "lfs check" to not block when the MDS is down (5628)
5896 - service request history (4965)
5897 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5898 - create an "evict_client" /proc entry on OSTs, like the MDS has
5899 - fix mount usage message, return errors per mount(8) (5168)
5900 - change grep [] to grep "[]" in tests so they work in more UMLs
5901 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5902 - remove ancient LOV_MAGIC_V0 EA support (5047)
5903 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5904 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5905 - allow readahead allocations to fail when low on memory (5383)
5906 - mmap locking landed again, after considerable improvement (2828)
5907 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5909 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5912 - send OST transaction number in read/write reply to free req (4966)
5913 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5914 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5915 - the watchdog thread now runs as interruptible (5246)
5916 - flock/lockf fixes (but it's still disabled, pending 5135)
5917 - don't use EXT3 constants in llite code (5094)
5918 - memory shortage at startup could cause assertion (5176)
5920 - reorganization of lov code
5921 - single portals codebase
5923 - add extents/mballoc support (5025)
5924 - direct I/O reads in the obdfilter (4048)
5925 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5927 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5930 - send OST transaction number in read/write reply to free req (4966)
5931 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5932 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5933 - the watchdog thread now runs as interruptible (5246)
5934 - handle missing objects in filter_preprw_read properly (5265)
5935 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5936 - drop import inflight refcount on signal_completed_replay error (5255)
5937 - unlock page after async write error during send (3677)
5938 - return original error code on reconstructed replies (3761)
5939 - no transno return for symlink open, don't save no-trasno open (3440)
5941 - add pid to ldlm debugging output (4922)
5942 - bump the watchdog timeouts -- we can't handle 30sec yet
5943 - extra debugging for orphan dentry/inode bug (5259)
5945 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5948 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5949 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5950 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5951 - fix NULL dereference in /proc/sys/portals/routes (4827)
5952 - allow failed mdc_close() operations to be interrupted (4561)
5953 - stop precreate on OST before MDS would time out on it (4778)
5954 - don't send partial-page writes before EOF from client (4410)
5955 - discard client grant for sub-page writes on large-page clients (4520)
5956 - don't free dentries not owned by NFS code, check generation (4806)
5957 - fix lsm leak if mds_create_objects() fails (4801)
5958 - limit debug_daemon file size, always print CERROR messages (4789)
5959 - use transno after validating reply (3892)
5960 - process timed out requests if import state changes (3754)
5961 - update mtime on OST during writes, return in glimpse (4829)
5962 - add mkfsoptions to LDAP (4679)
5963 - use ->max_readahead method instead of zapping global ra (5039)
5964 - don't interrupt __l_wait_event() during strace
5966 - add software watchdogs to catch hung threads quickly (4941)
5967 - make lustrefs init script start after nfs is mounted
5968 - fix CWARN/ERROR duplication (4930)
5969 - return async write errors to application if possible (2248)
5970 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5971 - print NAL number in %x format (4645)
5972 - update barely-supported suse-2.4.21-171 series (4842)
5973 - support for sles 9 %post scripts
5974 - support for building 2.6 kernel-source packages
5975 - support for sles km_* packages
5977 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5980 - ignore -ENOENT errors in osc_destroy (3639)
5981 - notify osc create thread that OSC is being cleaned up (4600)
5982 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5983 - reconstruct ost_handle() like mds_handle() (4657)
5984 - create a new thread to do import eviction to avoid deadlock (3969)
5985 - let lconf resolve symlinked-to devices (4629)
5986 - don't unlink "objects" from directory with default EA (4554)
5987 - hold socknal file ref over connect in case target is down (4394)
5988 - allow more than 32000 subdirectories in a single directory (3244)
5989 - fix blocks count for O_DIRECT writes (3751)
5990 - OST returns ENOSPC from object create when no space left (4539)
5991 - don't send truncate RPC if file size isn't changing (4410)
5992 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5993 - bind to privileged port in socknal and tcpnal (3689)
5995 - rate limit CERROR/CWARN console message to avoid overload (4519)
5996 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5997 - basic mmap support (3918)
5998 - kernel patch series update from b1_4 (4711)
6000 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
6003 - avoid crash during MDS cleanup with OST shut down (2775)
6004 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
6005 - don't use bad inodes on the MDS (3744)
6006 - dynamic object preallocation to improve recovery speed (4236)
6007 - don't hold spinlock over lock dumping or change debug flags (4401)
6008 - don't zero obd_dev when it is force cleaned (3651)
6009 - print grants to console if they go negative (4431)
6010 - "lctl deactivate" will stop automatic recovery attempts (3406)
6011 - look for existing locks in ldlm_handle_enqueue() (3764)
6012 - don't resolve lock handle twice in recovery avoiding race (4401)
6013 - revalidate should check working dir is a directory (4134)
6015 - don't always mark "slow" obdfilter messages as errors (4418)
6017 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
6020 - don't close LustreDB during write_conf until it is done (3860)
6021 - fix typo in lconf for_each_profile (3821)
6022 - allow dumping logs from multiple threads at one time (3820)
6023 - don't allow multiple threads in OSC recovery (3812)
6024 - fix debug_size parameters (3864)
6025 - fix mds_postrecov to initialize import for llog ctxt (3121)
6026 - replace config semaphore with spinlock (3306)
6027 - be sure to send a reply for a CANCEL rpc with bad export (3863)
6028 - don't allow enqueue to complete on a destroyed export (3822)
6029 - down write_lock before checking llog header bitmap (3825)
6030 - recover from lock replay timeout (3764)
6031 - up llog sem before sending rpc (3652)
6032 - reduce ns lock hold times when setting kms (3267)
6033 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
6034 - fix NULL deref and obd_dev leak on setup error (3312)
6035 - replace some LBUG about llog ops with error handling (3841)
6036 - don't match INVALID dentries from d_lookup and spin (3784)
6037 - hold dcache_lock while marking dentries INVALID and hashing (4255)
6038 - fix invalid assertion in ptlrpc_set_wait (3880)
6040 - add libwrap support for the TCP acceptor (3996)
6041 - add /proc/sys/portals/routes for non-root route listing (3994)
6042 - allow setting MDS UUID in .xml (2580)
6043 - print the stack of a process that LBUGs (4228)
6045 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
6048 - don't cleanup request in ll_file_open() on failed MDS open (3430)
6049 - make sure to unset replay flag from failed open requests (3440)
6050 - if default stripe count is 0, use OST count for inode size (3636)
6051 - update parent mtime/ctime on client for create/unlink (2611)
6052 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
6053 - free recovery state on server during a forced cleanup (3571)
6054 - unregister_reply for resent reqs (3063)
6055 - loop back devices mounting and status check on 2.6 (3563)
6056 - fix resource-creation race that can provoke i_size == 0 (3513)
6057 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
6058 - more debugging for page-accounting assertion (3746)
6059 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
6060 - avoid deadlock after precreation failure (3758)
6061 - fix race and lock order deadlock in orphan handling (3450, 3750)
6062 - add validity checks when grabbing inodes from l_ast_data (3599)
6064 - add /proc/.../recovery_status to obdfilter (3428)
6065 - lightweight CDEBUG infrastructure, debug daemon (3668)
6066 - change default OSC RPC parameters to be better on small clusters
6067 - turn off OST read cache for files smaller than 32MB
6068 - install man pages and include them in rpms (3100)
6069 - add new init script for (un)mounting lustre filesystems (2593)
6070 - run chkconfig in %post for init scripts (3701)
6071 - drop scimac NAL (unmaintained)
6073 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
6076 - clean kiobufs before and after use (3485)
6077 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
6078 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
6079 - change init script to fail more clearly if not run as root (1528)
6080 - allow clients to reconnect during replay (1742)
6081 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
6082 - don't do DNS lookups on NIDs too small for IP addresses (3442)
6083 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
6084 - fix cond_resched (3554)
6085 - only evict unfinished clients after recovery (3515)
6086 - allow bulk resend, prevent data loss (3570)
6087 - dynamic ptlrpc request buffer allocation (2102)
6088 - don't allow unlinking open directory if it isn't empty (2904)
6089 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
6090 - remove extraneous obd dereference causing LASSERT failure (3334)
6091 - don't use get_cycles() when creating temp. files on the mds (3156)
6092 - hold i_sem when setting i_size in ll_extent_lock() (3564)
6093 - handle EEXIST for set-stripe, set proper directory name (3336)
6095 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
6096 - fix ksocknal_fmb_callback() error messages (2918)
6098 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
6101 - don't copy lvb into (possibly NULL) reply on error (2983)
6102 - don't deref dentry after dput, don't free lvb on error (2922)
6103 - use the kms to determine writeback rpc length (2947)
6104 - increment oti_logcookies when osc is inactive (2948)
6105 - update client's i_blocks count via lvb messages (2543)
6106 - handle intent open/close of special files properly (1557)
6107 - mount MDS with errors=remount-ro, like obdfilter (2009)
6108 - initialize lock handle to avoid ASSERT on error cleanup (3057)
6109 - don't use cancelling-locks' kms values (2947)
6110 - use highest lock extent for kms, not last one (2925)
6111 - don't dereference ERR_PTR() dentry in error handling path (3107)
6112 - fix thread race in portals_debug_dumplog() (3122)
6113 - create lprocfs device entries at setup instead of at attach (1519)
6114 - common AST error handler, don't evict client on completion race (3145)
6115 - zero nameidata in detach_mnt in 2.6 (3118)
6116 - verify d_inode after revalidate_special is valid in 2.6 (3116)
6117 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
6118 - initialize RPC timeout timer earlier for 2.6 (3219)
6119 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
6120 - print nal/nid for unknown nid (3258)
6121 - additional checks for oscc recovery before doing precreate (3284)
6122 - fix ll_extent_lock() error return code for 64-bit systems (3043)
6123 - don't crash in mdc_close for bad permissions on open (3285)
6124 - zero i_rdev for non-device files (3147)
6125 - clear page->private before handing to FS, better assertion (3119)
6126 - tune the read pipeline (3236)
6127 - fix incorrect decref of invalidated dentry (2350)
6128 - provide read-ahead stats and refine rpc in flight stats (3328)
6129 - don't hold journal transaction open across create RPC (3313)
6130 - update atime on MDS at close time (3265)
6131 - close LDAP connection when recovering to avoid server load (3315)
6132 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
6133 - don't leak open file on MDS after open resend (3325)
6134 - serialize filter_precreate and filter_destroy_precreated (3329)
6135 - loop device shouldn't call sync_dev() for nul device (3092)
6136 - clear page cache after eviction (2766)
6137 - resynchronize MDS->OST in background (2824)
6138 - refuse to mount the same filesystem twice on same mountpoint (3394)
6139 - allow llmount to create routes for mounting behind routers (3320)
6140 - push lock cancellation to blocking thread for glimpse ASTs (3409)
6141 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
6142 - fix rare problem with rename on htree directories (3417)
6144 - allow default OST striping configuration per directory (1414)
6145 - fix compilation for qswnal for 2.6 kernels (3125)
6146 - increase maximum number of MDS request buffers for large systems
6147 - change liblustreapi to be useful for external progs like lfsck (3098)
6148 - increase local configuration timeout for slow disks (3353)
6149 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
6151 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
6154 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
6155 - don't overwrite extent policy data in reply if lock was blocked (2901)
6156 - drop filter export grants atomically with removal from device (2663)
6157 - del obd_self_export from work_list in class_disconnect_exports (2908)
6158 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
6159 - swab reply message in mdc_close, other PPC fixes (2464)
6160 - fix destroying of named logs (2325)
6161 - overwrite old logs when running lconf --write_conf (2264)
6162 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
6163 - fix race in target_handle_connect (2898)
6164 - mds_reint_create() should take same inode create lock (2926)
6165 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
6166 - don't close files for self_export to avoid uninitialized obd (2936)
6167 - allow MDS with the same name as client node (2939)
6168 - hold dentry reference for closed log files for unlink (2325)
6169 - reserve space for all logs during transactions (2059)
6170 - don't evict page beyond end of stripe extent (2925)
6171 - don't oops on a deleted current working directory (2399)
6172 - handle hard links to targets without a parent properly (2517)
6173 - don't dereference NULL lock when racing during eviction (2867)
6174 - don't grow lock extents when lots of conflicting locks (2919)
6176 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
6179 - account for cache space usage on clients to avoid data loss (974)
6180 - lfsck support in lustre kernel code (2349)
6181 - reduce journal credits needed for BRW writes (2370)
6182 - orphan handling to avoid losing space on client/server crashes
6183 - ptlrpcd can be blocked, stopping ALL progress (2477)
6184 - use lock value blocks to assist in proper KMS, faster stat (1021)
6185 - takes i_sem instead of DLM locks internally on obdfilter (2720)
6186 - recovery for initial connections (2355)
6187 - fixes for mds_cleanup_orphans (1934)
6188 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
6189 - block all file creations until orphan recovery completes (1901)
6190 - client remove rq_connection from request struct (2423)
6191 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
6192 - recovery timer race (2670)
6193 - mdc_close recovey bug (2532)
6194 - ptlrpc cleanup bug (2710)
6195 - mds timeout on local locks (2588)
6196 - namespace lock held during RPCs (2431)
6197 - handle interrupted sync write properly (2503)
6198 - don't try to handle a message that hasn't been replied to (2699)
6199 - client assert failure during cleanup after abort recovery (2701)
6200 - leak mdc device after failed mount (2712)
6201 - ptlrpc_check_set allows timedout requests to complete (2714)
6202 - wait for inflight reqs when ptlrpcd finishes (2710)
6203 - make sure unregistered services are removed from the srv_list
6204 - reset bulk XID's when resending them (caught by 1138 test)
6205 - unregister_bulk after timeout
6206 - fix lconf error (2694)
6207 - handle write after unfinished setstripe, stripe-only getstripe (2388)
6208 - readahead locks pages, leaves pending causing memory pressure (2673)
6209 - increase OST request buffers to 4096 on large machines (2729)
6210 - fix up permission of existing directories in simple_mkdir (2661)
6211 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
6212 - don't assign transno to errored transactions (2742)
6213 - don't delete objects on OST if given a bogus objid from MDS (2751)
6214 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
6215 - if rq_no_resend, then timeout request after recovery (2432)
6216 - fix MDS llog_logid record size, 64-bit array alignment (2733)
6217 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
6218 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
6219 - MDS recovery shouldn't delete objects that it has given out (2730)
6220 - if enqueue arrives after completion, don't clobber LVB (2819)
6221 - don't unlock pages twice when trigger_group_io returns error (2814)
6222 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
6223 - don't write pages to disk if there was an error (1450)
6224 - don't ping imports that have recovery disabled (2676)
6225 - take buffered bytes into account when balancing socknal conn (2817)
6226 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
6227 - reconnect unlink llog connection after MDS reconnects to OST (2816)
6228 - remove little-endian swabbing of llog records (1987)
6229 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
6230 - retry reposting request buffers if they fail (1191)
6231 - grow extent at grant time to avoid granting a revoked lock (2809)
6232 - lock revoke doesn't evict page if covered by a second lock (2765)
6233 - disable VM readahead to avoid reading outside lock extents (2805)
6235 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
6236 - updated kernel patches for hp-2.4.20 kernel (2681)
6238 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
6241 - fix truncated write corruption (2366)
6242 - fix for failed assertion in iopen_connect_dentry (1792,2517)
6244 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
6245 - ensure len(uuid) < 37 in lmc (1171)
6246 - fix ia64 OOPS in llog_test (2255)
6247 - zero end of page at obdfilter for partial page writes (2648)
6248 - don't leave stale dentries around after renames (bug 2428)
6249 - fix timeouts when evicting a client with a single lock held (2642)
6250 - set deadline for the initial HELLO message to drain (2634)
6251 - print out dotted-quad IP addresses in the socknal (2302)
6253 - additional debugging for MDS client eviction problem (2443)
6254 - fix mkfsoptions support for osts (2603, 2604)
6256 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
6259 - add series for the vanilla 2.6.0 kernel
6260 - add series for the vanilla 2.4.24 kernel
6261 - add series for a cray x86/64 UL kernel drop
6262 - fix xattr patches for the vanilla 2.4.19 series
6264 - generate true UUIDs in lmc (1171)
6265 - have portals stack dumping break in UML (2466)
6266 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
6267 - allocate with _NOFS in ldlm to avoid deadlock (1933)
6268 - wake callback waiting threads on client eviction (2460)
6269 - Add --ptldebug and --subsystem to lmc (1719)
6270 - update assertion to allow safe interrupt allocation
6271 - set rq_no_resend for cancel requests (2432)
6272 - recalculate ptlrpcd timeout after resend (2494)
6273 - call vfs_rmdir when removing pending directories (2368)
6274 - fix renaming a file to itself (2429)
6275 - lmc creates a default one-stripe lov (2454)
6276 - expand procfs space to handle large clusters (2326)
6277 - increase UML stack to avoid overflow
6278 - update lconf's list of debug and subsystem masks
6279 - fix lfs find --obd (2510)
6280 - /proc tunable for disabling filter read caching (2591)
6281 - stop rpm packages from altering slapd.conf (2301)
6282 - disable nagle in the socknal under 0conf (2578)
6283 - choose mds inode size based on stripe count (2572)
6284 - fix kernel-source rpm problems (2516)
6286 - add --disable-doc to avoid pdf generation (2421)
6287 - update documentation, tests, type-os, comments
6288 - avoid format warnings on ia64
6289 - remove the TOE NAL
6290 - tiny code cleanups by removing unused fields
6292 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
6295 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
6296 - remove the most blatant lies from BUILDING file (2371)
6297 - change default debug level to reasonable production setting
6298 - reduce client side cache size to reduce cache flush time
6299 - reduce max RPCs in flight to avoid unnecessary file fragmentation
6300 - make TCP zerocopy and pinger support enabled by default (2476)
6301 - sync writes completed after process exits caused crashes (2319)
6302 - maintain correct mount count on the MDS (2356)
6303 - backout 1557, because 2316 wasn't really fixed
6304 - better file I/O statistics gathering in /proc
6305 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
6306 - another kernel patch to fix zero-copy TCP function export
6307 - don't take duplicate lock when processing re-sent getattr (2420)
6308 - lctl uses obd_self_export instead of creating new conn (2353)
6309 - MDS/OST recovery case which requires object creation asserted (2425)
6310 - move lfs from /usr/sbin to /usr/bin in packages
6311 - fix race between mds_client_add and mds_client_free (2417)
6312 - use kmalloc instead of slabs in portals (2430)
6313 - don't create duplicate records when a failover MDS is present (2442)
6314 - remove unnecessary mount age check (2332)
6315 - don't remove directory inodes from locks prematurely (2451)
6316 - don't break if MDS service name is the same as hostname (2103)
6317 - fix races in client write RPC generation when cache full (2482)
6319 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
6322 - remove now-unused request->rq_obd (278)
6323 - if an allocation fails, print out how much memory we've used (1933)
6324 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
6325 - add the "configurable stack size" patch to most series files (1256)
6326 - ability to write large log records, for 100+ OST configs (2306)
6327 - fix NULL deref when filter_prep fails (2314)
6328 - fix operator precedence error in filter_sync
6329 - dynamic allocation of socknal TX descriptors (2315)
6330 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
6331 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
6332 - ensure that CWARN messages in Portals always get to the syslog
6333 - __init/__exit are not for prototype decls (ldlm_init/exit)
6334 - x86-64 compile warning fixes
6335 - fix gateway LMC keyword conflict (2318)
6336 - fix MDS lock inversions in getattr/reint paths (1844)
6337 - fix a rare lock re-ordering bug, which caused deadlock (2322)
6338 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
6339 - DLM race condition prevented some lock evictions (2328)
6340 - ENOMEM detection and retry on socknal sends (2230)
6341 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
6342 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
6343 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
6344 - better error messages when a client is rejected during recovery (1505)
6345 - avoid cancelling locks which were never granted, after failure (2330)
6346 - fix i_sem/journal inversion in mds_client_add (2333)
6347 - fix truncate/getattr lock cycle deadlock (2334)
6348 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
6349 - fix two rare exit paths which could leak an l_lock() ref (2321)
6350 - fencepost error in MDS/OST orphan recovery (2226)
6351 - make log record alignment 8 bytes (1988)
6352 - lstripe now fails when requested offset > ost_count (2237)
6353 - ensure that all kernel series have a complete list.h (1607)
6354 - fix crashes in special-file operations (2316)
6355 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
6357 - allow configurable automake binary, for testing new versions
6358 - small update to the lfs documentation
6360 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
6362 * fix negative export reference count in fsfilt_sync (2312)
6364 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
6365 * release candidate 0.9.1
6367 - orphans are moved into the PENDING directory for possible recovery
6368 - replayed opens now open by fid for orphan/rename safety (1042)
6369 - last close of an orphan inode generates a transno (683)
6370 - chdir() and mount() now pin the directory entry (1020)
6371 - avoid CERROR in normal ll_setattr_raw() error case (1500)
6372 - discard very old requests without processing them (1502)
6373 - remove some common, well-understood CERRORs (1505)
6374 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
6375 - clear "grant" flags in OST replies until OST grant code lands (1644)
6376 - fix read performance by not clobbering i_blksize on client (1598)
6377 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
6378 - make LRU size a /proc tunable, clears locks when reduced (707)
6379 - fix some lprocfs rot that prevented ptlbd from loading (1732)
6380 - server locks take references on exports now (1558)
6381 - build fixes for 2.4.20-rh trees (1663)
6382 - return an error from lov_create if all OSCs are inactive (1751)
6383 - fix import levels when a reconnect happens without a timeout (1597)
6384 - exit early from mds_open if we get a lookup error (1749)
6385 - partial page read at EOF wouldn't wait for disk before sending (1642)
6386 - avoid NULL deref in obdfilter when reading page past EOF (1592)
6387 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
6388 - fix LBUG in ll_it_open_error with rc = -2 (1861)
6389 - write/truncate lock inversion (1639)
6390 - Don't auto-load obdclass, portals modules during cleanup (1495)
6391 - fix timestamps from jumping to "now" (1763)
6392 - extra journal assertions (1648)
6393 - add an extra multiunlink test (1771)
6394 - fix read_record/write_record API (1776)
6395 - fix leak of offset_extent, possible incorrect i_size later (1772)
6396 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
6397 - Debugging for the kqswnal_get_idle_tx problems (1820)
6398 - Allow recovery to be attempted multiple times (1536)
6399 - Write out MDS last_rcvd file after it is first created (1600)
6400 - Fix tx_descriptor leak in failed transmit situations (1827)
6401 - ext3 journaling fixes for assertion failure after IO error (1871)
6402 - class_export_put() on freed export after completion AST error (1896)
6403 - Fix revalidate looping in VFS (1322)
6404 - Don't access a freed export during MDS_REINT timeout (1521)
6405 - Add open-unlink recovery support on the MDS (1673,1764)
6406 - Return an error if no MDS data was read from last_rcvd (1946)
6407 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
6408 - Don't setup a disk device that doesn't match exported UUID (317)
6409 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
6410 - avoid committing NULL handle in force close
6411 - local.sh is now a one-stripe LOV configuration
6412 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
6413 - fix ext3 htree duplicate directory entry corruption (1516)
6414 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
6415 - update to new LOV EA format (2097)
6416 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
6418 2003-06-15 Phil Schwan <phil@clusterfs.com>
6421 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
6422 - per-import/export recovery handling (958, 931, 959)
6423 - multiple last-rcvd slots, for serving multiple FSes (949)
6424 - connections are again shared between multiple imp/exports (963, 964)
6425 - "umount -f" would hang if any requests needed to be sent (393, 978)
6426 - avoid pinning large req buffer by copying for queued messages (989)
6427 - add "uuid" to "lctl device" command to help upcalls (991)
6428 - "open" RPCs with transnos would confuse recovery counters (1037)
6429 - do proper endian conversion of all wire messages (288, 340, 891)
6430 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
6431 - call waiting locks callback from LDLM recovery thread (1127, 1151)
6432 - fix ptlrpc_connection leak in target_handle_connect (1174)
6433 - fix import refcounting bug in OST and MDS cleanup (1134)
6434 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
6435 - fix very unlikely obd_types race condition (501)
6436 - remove osc_open hack for echo_client (1187)
6437 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
6438 - a failure in read_inode2 leads to deadlock (1139)
6439 - cancel ack-locks as soon as transaction is committed (1072)
6440 - fix major leaks and crashes in the bulk I/O path (937, 1057)
6441 - make sure to commitrw after any preprw to avoid deadlock (1162)
6442 - failing to execute a file in a lustre FS would lock inode (1203)
6443 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
6444 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
6445 - obd_uuid2tgt would walk off the end of the list (1255)
6446 - on IA64 the osc would give portals incorrect bulk size (1258)
6447 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
6448 - fix lock inversion caused by new llite matching code (1282)
6449 - limit the number of dirty pages on a client to 10MB (1286)
6450 - timed out locks were not being corrected cancelled (1289)
6451 - fix O_DIRECT above 4GB on IA-32 (1292)
6452 * major user-visible changes
6453 - fail out/fail over policy now controlled by the upcall (993)
6455 - add OBD_PING to check server availability and failure (954)
6456 - lustre messages are now sent in sending host order (288, 340, 891)
6457 - add eadatalen to MDS getattr reply (340)
6458 - OST read replies may contain second buffer, with per-page status (593)
6460 2003-03-11 Phil Schwan <phil@clusterfs.com>
6463 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
6464 - failed open()s could cause deadlock; fixed (867, 869)
6465 - stop cancelling OST locks when files are closed (481)
6466 - overlapping XID spaces caused network corruption (851, 853)
6467 - fix unsafe fsfilt counter arithmetic; change to atomic_t
6468 - setattr_raw added, to do single-RPC, server-side setattrs
6469 - lmc/lconf syntax change for OST UUIDs
6470 - fix crashy race condition between ptlrpc_free_req and osc_close
6471 - don't use request in mdc_enqueue if we hit a timeout (889)
6472 - don't set the inode i_size for regular files from the MDS (896)
6473 - handle out of order completion AST (842)
6474 - don't LBUG if a lock request times out after receiving AST (913)
6475 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
6476 - if a bad lock AST arrives, send an error instead of dropping entirely
6477 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
6478 - fix leak in bulk IO when only partially completed (899, 900, 926)
6479 - fix O_DIRECT for ia64 (55)
6480 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
6481 - C-z after timeout could hang a process forever; fixed (977)
6483 - client-side I/O cache (678, 924, 929, 941, 970)
6485 - READPAGE and SETATTRs which don't take server-side locks get
6488 2003-02-11 Phil Schwan <phil@clusterfs.com>
6491 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
6492 - Fix fsfilt_extN_readpage() to read a full page of directory
6493 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
6494 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
6495 - Fix imbalanced LOV object allocation and out-of-bound access (469)
6496 - Most intent operations were removed, in favour of a new RPC mode
6497 that does a single RPC to the server and bypasses most of the VFS
6498 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
6499 - Aggressively cancel local locks on DLM servers
6500 - mds_reint_unlink sends EA to the client if it's the last nlink.
6501 client uses that EA to unlink OST objects.
6502 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
6503 - recursive symlinks were fixed (439)
6504 - fixed NULL deref in DEBUG_REQ
6505 - filter_update_lastobjid no longer calls sync, which annoyed extN
6506 - fixed multi-client small-writes to a single file problem (445)
6507 - fixed mtime updates during file writes (607)
6508 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
6509 - fixed bug in obd_brw_read/write() (under guise of testing 367)
6510 - fixed Linux OST size reporting problem (444, 656)
6511 - OST now updates object mtime with writes or setattr (607, 619)
6512 - client verifies file size before zeroing page past EOF (445)
6513 - OST now writes last allocated objid to disk with allocation (108)
6514 - LOV on echo now works (409)
6516 - mds_reint_unlink sends a new buffer, with the EA included. this
6517 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
6518 set if a regular file was being unlinked, and it was the last link
6519 - use PtlGet from the target for bulk writes (315)
6520 - OST now updates object mtime with writes or setattr (607, 619)
6521 - LDLM now has a grant-time callback to revalidate locked items, if
6523 - Many MDS operations were reorganized to combat race conditions
6525 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
6526 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
6528 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
6531 - Fully reactivate OST imports after reconnection (512, others)
6532 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
6533 - More graceful error handling for truncating on dead OST (515)
6534 - Don't error out unless we're actually accessing dead stripes (474)
6535 - Fix garbage sizes when stripes are missing (410)
6536 - LRU counters were broken, causing constant lock purge (433, 432)
6537 - garbage on read from stripes with failed OSTs (441)
6538 - mark OSCs as active before reconnecting during recovery (438)
6539 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
6540 - lfind did not preserve OST order in output (443)
6541 - symlinks cause hung clients, incorrect data (439)
6542 - stop dereferencing request after dropping refcount (457)
6543 - don't LASSERT(spin_is_locked) on non-SMP (455)
6544 - fixes for many rename() bugs
6545 - fstat didn't correctly synchronize attributes (399)
6546 - server must handle lock cancellation during blocking AST prep (487)
6547 - bulk descriptors were free()d too soon (511)
6548 - fix paths in lconf, which would load incorrect modules (451, 507)
6549 - fix confusing lconf 'host not found' error message (386)
6550 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
6551 - fix race condition in mdc_blocking_ast() for inode access (526)
6552 - fix lov_unpackmd() unpacking wrong number of stripes (537)
6553 - fix lov_set_osc_active() marking wrong OSC inactive (440)
6554 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
6555 - fix multiple writes of stripe MD to MDS (358, maybe 519)
6556 - fix lstripe in several ways (kernel side) (527)
6557 - fix request leak in ldlm_cli_enqueue (262)
6558 - incorrect OSC was marked inactive after OST failure
6559 - call mds_fs_cleanup before unmounting filesystem (524)
6560 - fix races between taking ns_lock and ldlm_lock_change_resource
6561 - fix races updating LOV export open file list
6562 - fix lov_enqueue error path, avoid decref-ing bad lock handle
6563 - fix recovery NULL deref in ldlm_cli_cancel_unused
6564 - fix some DLM races by using new hash table for lock handles (419)
6565 - permit the client to specify desired inodes, at replay
6566 - duplicate requests when we queue them for replay reintegration
6567 - fix last_rcvd offset calculation
6568 - sync after each recovered transaction, so we always make progress
6569 - never, not always, ERESTART requests without transnos
6570 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
6572 - skip replay if the MDS says that the client is already connected
6573 - don't check for a recovery-enabled export to match lctl's UUID
6574 - don't INC_USE_COUNT for phantom exports
6575 - don't crash when cleaning up phantom exports (567)
6576 - don't double-finish or set replay data for errored mdc_open requests
6577 - abort requests when they time out, so we don't get old replies
6578 - send/receive replies for AST messages again
6579 - if the client says that it doesn't have the lock, cancel it on the
6581 - if we timeout during I/O, don't try to cancel an in-use lock; instead
6582 mark it as destroyed, it will all work out when decref is called
6583 - fix module use counts (22, 581)
6585 - ASTs now expect a reply (server cancels lock on error reply)
6587 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6590 - fix many simultaneous client startup (392)
6591 - fix dentry->d_it clobbering
6592 - credentials weren't being shipped for readdir/getattr operations
6593 - remove invalid assertions triggered during some concurrent MD
6595 - proper Lustre versions added (336, 389)
6596 - fix memory leak for create error case (398)
6597 - fix LOV locking bug that would get cli/srv out of sync
6598 - fix echo client over LOV (409)
6599 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6600 - fix double-O_EXCL intent crash (424)
6601 - avoid sending multiple lock CANCELs (352)
6603 - MDS can do multi-client recovery (modulo bugs in new code)
6605 - many updates, edits, cleanups
6607 2002-11-18 Phil Schwan <phil@clusterfs.com>
6610 - fix null d_it dereference (346)
6611 - fix full OST/dbench hang (333)
6612 - fix permission problem with file removal (286)
6613 - fix removal of OSCs from LOV when they fail
6614 - fix NULL deref during bulk timeout (214)
6615 - fix problems related to multiple filesystems on one MDS (241)
6616 - fixed serious subtle metadata locking bugs
6617 - free locks on clients when inodes are removed due to memory
6619 - fix inode pointer in lock data (285)
6620 - partial support for multiple MDS on a single host (241)
6621 - data locks weren't cancelled at clear_inode time (290, 311)
6622 - intent locks could lead to unbounded lock growth (205)
6623 - added a maximum lock count, an LRU list, and a flusher
6624 - fix multiple rename (365)
6625 - properly abstracted the echo client
6626 - OSC locked 1 byte too many; fixed
6627 - rewrote brw callback code:
6628 - fixed recovery bugs related to LOVs (306)
6629 - fixed too-many-pages-in-one-write crash (191)
6630 - fixed (again) crash in sync_io_timeout (214)
6631 - probably fixed callback-related race (385)
6633 - Add capability to MDS protocol
6634 - LDLM cancellations and callbacks on different portals
6636 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6639 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6640 - defer open object destruction to close time (601981, 138)
6641 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6642 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6643 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6644 - don't call obd_setattr in ll_file_release for destroyed objects
6645 * protocol change to lustre_msg: move |version| and add |flags|
6646 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6647 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6648 * added replay of create, unlink, link and rename operations during
6649 MDS failover; recovery should be much more robust now
6650 * remove failed OSCs from LOVs (only lov_create uses this so far)
6651 * the lustre-HOWTO was brought (more) up to date (582544)
6653 2002-10-23 Phil Schwan <phil@clusterfs.com>
6656 - in-use dentries weren't being reused properly (617851)
6657 - prevent multiple LDLM setup (599178)
6658 - fix LOV size calculations for truncate (617853)
6659 - fix client handling of MDS intent errors (POSIX)
6660 - fix permission bug in lovstripe.c test (624321)
6661 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6662 - truncate past end of file could corrupt data
6663 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6664 - a race in recovery could return ETIMEDOUT to apps (623947)
6665 - building outside the source directory was fixed
6666 * the lustre-HOWTO was brought (more) up to date (582544)
6667 * major progress was made on recovery functionality
6669 2002-10-10 Phil Schwan <phil@clusterfs.com>
6672 - recovery deadlock fix
6673 - rm -rf causes LBUG fix (617817)
6674 - file open by multiple tasks fix (618962)
6675 - directory permissions bugs (602707 and 620007)
6676 - journal_stop fixed with locking (611313)
6677 - O_APPEND failures resolved (618273, perhaps 614459)
6678 - lconf PATH fix (619770)
6679 - IA64 build fix (621450)
6680 - RPC buffer sizes scale with amount of memory
6682 2002-10-01 Phil Schwan <phil@clusterfs.com>
6685 - locks would be cancelled without throwing away data pages,
6686 resulting in inconsistent data (605627)
6687 - inode attributes were not always being refreshed (605627, 612449)
6688 - lconf now continues to cleanup after lctl reports an error
6689 - MDS now enforces user permissions (602707)
6690 - lprocfs cleanup fixed, but not yet enabled (614157)
6691 - fixed infinite server hang, should a client not respond to an AST
6692 - avoid going into recovery if user calls readlink() with a buffer
6693 that's too small (613941)
6694 - AST RPCs no longer require replies (614867) -- this may be changed
6695 - don't crash server if client sends an IOV that's too big (611336)
6696 - fixed lock conversion deadlock (611892)
6697 - fixed the following of symlinks (614622)
6698 * recovery: the server can remove locks from a client that dies, other
6699 clients can make progress
6700 * more extN patch fixes
6701 * compile-time configurable ptlrpc buffer allocations
6703 - collaborative read cache document
6704 - Lustre Lite Performance CDR document-in-progress
6706 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6709 - fix typo in patch-2.4.18
6711 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6714 - clear ptlrpc request each time in handle_incoming_request()
6715 - unlink of files now destroys the object on the OST
6717 2002-09-19 Peter Braam <braam@clusterfs.com>
6719 * add hard link support
6720 * change obdfile creation method
6721 * kernel patch changed
6723 2002-09-19 Peter Braam <braam@clusterfs.com>
6726 - stack overflow bug in extN fixed
6728 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6730 * documentation updates
6731 - add man pages for config tools
6732 - update tests/README to describe testing with new config tools
6733 - finish metadata API descriptions
6734 * bug fixes and cleanups
6735 - statfs workaround for 16TB limit
6736 - LOV stripe allocation improved, can stripe on subset of OSTs
6737 - LOV file size/IO offset was wrong for files > 4GB in size
6738 - object EA data was being dropped, caused files to be unreadable
6739 - memory overflow with non-LOV OST caused memory corruption
6740 - fixed regression tests to work with new config tools, obdfilter
6741 - fixed bug when directory size became larger than 1 block
6742 - fixed bug (for single client case) when PWD was deleted
6743 - invalidate local directory pages when doing intent-based ops
6744 - avoid LDLM oops when lock callback contained bad data
6746 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6748 * documentation updates
6749 * bug fixes and cleanups
6750 - configuration tools
6753 - 64-bit compile warnings
6754 - 64-bit internal statfs data
6756 * test_brw on persistent OST devices
6758 * lprocfs (disabled)
6760 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6762 * documentation updates
6763 * bug fixes and cleanups
6764 * configuration tools
6766 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6769 * many small fixes to 0_5_4
6770 * io/network handling
6771 * thinkos in MDS operations
6773 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6776 * crucial basic fixes to 0.5.3
6777 * IOR, Iozone work over Elan
6780 2002-08-07 Phil Schwan <phil@clusterfs.com>
6781 * version 0_5_3, our first alpha
6782 * we use the new Portals iovs
6783 * documentation updates
6784 * bug fixes and cleanups
6785 * small changes in the DLM wire protocol
6787 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6788 * version 0_5_1 with some initial stability,
6789 * locking on MD and file I/O.
6790 * documentation updates
6791 * several bug fixes since 0.5.0
6792 * small changes in wire protocol
6794 2002-07-18 Phil Schwan <phil@clusterfs.com>
6796 * delivered as Lustre Light Alpha
6797 * fixed a crash after handling invalid MDS requests
6798 * fixed directory pages for architectures with non-4k pages sizes
6800 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6801 * release version v0_4_4
6802 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6804 * Fixes a number of interruption problems with OST operations.
6805 * Update documentation for portals header changes
6806 * Move all wire protocol structs/defines to lustre_idl.h
6807 * Fixes symlink length bug.
6808 * Add tcpdump to repository.
6810 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6811 * release version v0_4_3
6812 * Fixes statfs for inodes on extN.
6813 * Fixes bug in runtests which would delete /etc/hosts.
6814 * Use 64-bit object IDs wherever possible (not into VFS though)
6815 Remove ost_get_info, which is unused by lustre, and out of date.
6817 2002-07-03 Peter Braam <braam@clusterfs.com>
6818 * release version v0_4_2 Fixes a lookup error (type not passed)
6819 * move forward to head of Portals
6820 * move forward to latest Lustre kernel
6822 2002-06-25 Peter Braam <braam@clusterfs.com>
6823 * release version v0_4_1. Hopefully stable on single node use.
git://git.whamcloud.com / fs / lustre-release.git / blob